DEVELOPMENT OF AN INTELLIGENT EXERCISE

PLATFORM FOR REHABILITATION IN PARKINSON’S DISEASE

by

HASSAN MOHAMMADI ABDAR

Submitted in partial fulfillment of the requirements

For the degree of Doctor of Philosophy

Dissertation Advisor: Dr. Kenneth Loparo

Department of Electrical Engineering & Computer Science

CASE WESTERN RESERVE UNIVERSITY

August 2014
 CASE WESTERN RESERVE UNIVERSITY
SCHOOL OF GRADUATE STUDIES

We hereby approve the thesis/dissertation of

Hassan Mohammadi Abdar

candidate for the degree of Doctor of Philosophy*

Committee Chair

(signed) Dr. Kenneth A. Loparo

Committee Member

Dr. Vira Chankong

Committee Member

Dr. Evren Gurkan-Cavusoglu

Committee Member

Dr. Farhad Kaffashi

(date) 07 July 2014

*We also certify that written approval has been obtained for any proprietary

material contained therein.

ii
Table of Contents
Table of Contents…………………………………………………………………iii
List of Tables……………………………………………………………………...v
List of Figures…………………………………………………………………….vi
Acknowledgement…..……………………………………………………………ix
Abstract……………………………………………………………………………x
1. Introduction………………………………………………………………………..…1
1.1. Introduction and Motivation..…………………………………………………….1
1.1.1. Parkinson’s Disease and Exercise Therapy……………………………….3
1.1.2. High-Rate Cycling and Rehabilitation in Parkinson’s Disease…………...5
1.1.3. Dance Therapy for PD Patients……………………………………………6
1.1.4. Motor Function Evaluation………………………………………………..7
1.2. Contributions……………………………………………………………………..8
1.3. Thesis Organization……………………………………………………………..10
2. Variability Analysis of PD Patients Data, Forced and Voluntary Cycling……...11
2.1. Introduction……………………………………………………………………...11
2.2. Patients Exercise and Data Recording (Methods and protocol)………………...14
2.2.1. Forced and Voluntary Cycling Data Collection………………………….14
2.2.2. Motor Function Assessment……………………………………………..15
2.3. Variability Data Analysis and Signal Processing Methods……………………..15
2.3.1. Approximate Entropy…………………………………………………….15
2.3.2. Sample Entropy…………………………………………………………..17
2.3.3. Spectral Entropy………………………………………………………….17
2.3.4. Multiple Linear Regression………………………………………………18
2.3.5. Logistic regression……………………………………………………….19
2.3.6. Odds ratio………………………………………………………………...20
2.4. Biomechanical and Physiological Feature Analysis…………………………….20
2.5. Variability Analysis Results…………………………………………………….23
2.6. Session by Session Variability Analysis and Prediction………………………..26
2.7. Conclusion………………………………………………………………………29
3. Design, Development and Validation of a Smart Exercise Bike for Rehabilitation
in Parkinson’s Disease……………………………………………………………...32
3.1. Introduction……………………………………………………………………...32
3.2. System Architecture; Hardware and Electronics………………………………..35
3.2.1. Hardware and Electronics………………………………………………..36
3.2.2. Heart Rate Monitoring System…………………………………………..38
3.2.3. Bike Operational Specifications…………………………………………39
3.3. Software, Programs and Control Algorithms…………………………………...42
3.3.1. Development and Programming Software……………………………….42
3.3.2. Motor Drive Control Modes……………………………………………..43

iii
 3.3.3. Static Mode (Inertia Load)……………………………………………….45
3.3.4. Dynamic Mode (Speed Reference)………………………………………46
3.3.5. User interface and data logging………………………………………….48
3.4. Test and Verifications…………………………………………………………...50
3.5. Data Analysis and Smart Bike Validation with PD Patients……………………54
3.5.1. Materials and Methods…………………………………………………...56
3.5.2. Biomechanical and Physiological Feature Analysis……………………..60
3.5.3. Data Analysis Results……………………………………………………62
3.6. Conclusion………………………………………………………………………69
4. Electronic Coupling of a Tandem Bike……………………………………………73
4.1. Introduction……………………………………………………………………...73
4.2. System Architecture, Mechanical Design and Bike Modifications……………..75
4.2.1. Mechanical Design and Bike Modification……………………………...78
4.2.2. Hardware Design, Electrical and Electronics ……………...……………79
4.3. Software and Control Algorithm………………………………………………..81
4.3.1. Programming Software…………………………………………………..82
4.3.2. Control Algorithm………………… …………………………………….82
4.4. Practical Experiments and test Results………………………………………….85
4.4.1. Mechanical Coupling Experiments and Data Collecting………….……..86
4.4.2. Electronically Coupling Experiments and Test Results…………….……88
4.5. Remote Tandem Cycling……………………………………………………..…92
4.6. Conclusion……………………………………………………………………....95
5. Summary and Future Directions………………………………………………..…97
5.1. Summary………………………………………………………………………...97
5.2. Future Direction and Suggestions……………………………………………...101
5.2.1. Adaptive Dynamic Exercise Program…………………………………..102
5.2.2. Real-Time Dopamine Change Measurement/ Monitoring of Animal Model
During the Exercise……………………………………………………….105
5.2.3. Shaking Measurement for Monitoring the UPDRS and Evaluate the
Quality of Life in PD Patients…………………………………………….106

Bibliography…………………………………………………………………………...107

iv
List of Tables
Table 2.1 Mean and variance for power, heart rate and cadence signals……………….15
Table 2.2 The values of ApEn, SaEn and SpEn for cadence signal of 21 exercise
sessions of one patient (with time delays,   1 and   3 )…………………...21
Table 2.3 ApEn and SpEn for power, heart rate and cadence signals…………………22
Table 2.4 Regression analysis results for single and tandem rider tests………………..26
Table 2.5 Combined single and tandem rider data regression analysis results…………26
Table 3.1 Comparison between the static and dynamic modes with real data signals of
two PD patients exercised with smart bike…………………………………..54
Table 3.2 Subject Demographics……………………………………………………….57
Table 3.3 Exercise Parameters………………………………………………………….57
Table 3.4 Mean and variance for power, heart rate and cadence signals……………….59
Table 3.5 Mean of ApEn and SpEn for power, heart rate and cadence signals over three
sessions………………………………………………………………..……..61
Table 3.6 Regression analysis results for the static group………………………...……67
Table 3.7 Combined static and dynamic group data regression analysis results……….67
Table 3.8 Regression analysis results for the dynamic group…………………………..68

v
List of Figures
Figure 2.1 Representative training data (pedaling rate, HR, and trainer and patient

power) during a 30-minute exercise block of forced exercise…..…………12

Figure 2.2 (a) Real (measured) versus predicted UPDRS values; A. single (voluntary)

exercise data; B. tandem (forced) data; C. combined single and tandem data.

(b) Correlation analysis of real (measured) versus predicted UPDRS values;

A. single (voluntary) exercise data; B. tandem (forced) data; C. combined

single and tandem data……………………………………………………..25

Figure 2.3 Session by session UPDRS prediction of four PD patients. (a) Positive

change with correct prediction, (b) Negative change with inexact prediction,

(c) Negative change with correct prediction, (d) No change with precise

final prediction……………………………………………………………..29

Figure 3.1 Duplicating the tandem cycling with a controllable motorized bike………33

Figure 3.2 Functional diagram of the Smart Bike……………………………………..36

Figure 3.3 Electrical/electronic component diagram of the bike……………………...37

Figure 3.4 Block diagram of the heart rate monitoring system………………………..39

Figure 3.5 Smart exercise bike, completed system……………………………………40

Figure 3.6 Screenshot of the drive parameters setup in RSLogix 5000……………….44

Figure 3.7 Closed loop control block diagram of the samrt bike in static mode………46

Figure 3.8 Closed-loop control block diagram of the samrt bike in dynamic mode…..48

Figure 3.9 PanelView display and control screen……………………….…………….50

Figure 3.10 Smart Bike real test data for two PD patients; (a) Static mode; (b) Dynamic

mode…………………………………………………………………….....53

vi
Figure 3.11 Flow diagram for recruitment and assessment of individuals with PD……56

Figure 3.12 Correlation analysis of real (measured) versus predicted UPDRS values for

linear MLR model. (a): static group. (b): Dynamic group (active assisted

exercise) (c): combined static and dynamic data…………………………..65

Figure 3.13 Correlation analysis of real (measured) versus predicted UPDRS values for

linear MLR model, (a): static group (voluntary exercise), (b): Dynamic

group (active-assisted exercise), (c): combined static and dynamic data….66

Figure 4.1 Block diagram of the electrically coupled tandem bike……………………74

Figure 4.2 (a) Quarq Riken GXP power meter crankset, (b) Garmin USB ANT+ stick

used to receive the power meter signals on the computer…………………76

Figure 4.3 Bike Friday Family Tandem and Kurt Kinetic Road Machine trainer…….77

Figure 4.4 Functional diagram of the electronically coupling tandem bike…………...78

Figure 4.5 Mechanical structure designed for servomotors attachment to the bike…...79

Figure 4.6 Electrical/electronics components diagram of the instrumented tandem

bike…………………………………………………………………………80

Figure 4.7 Electronically coupling tandem bike, completed system…………………..81

Figure 4.8 Block diagram of control algorithm for electronically coupling of the

tandem bike………………………………………………………………...84

Figure 4.9 Screen shot of the torque command equations implemented with ladder

diagram using the RSLogix 5000 software………………………………...85

Figure 4.10 Cadence, power and torque data for a 10-minute experiment …..………...87

Figure 4.11 Real (measured) cadence and simulated model output data……………….88

Figure 4.12 The cadence and power signals of both riders in a test scenario with one

vii
 rider as the trainer/leader…………………………………………………..90

Figure 4.13 The cadence and power signals of both riders in a test scenario with one

rider as the follower………………………………………………………..90

Figure 4.14 The cadence and power signals of both riders in a test scenario with both the

trainer/leader and follower…………………………………………………91

Figure 4.15 A screenshot of the RSLogix 5000 Trends window displaying the velocity

and power of riders in a test scenario with both the leader and follower....91

Figure 4.16 Electronically (remote) connection of two single bikes…………………...92

Figure 4.17 Electronically coupling of two (single) Smart Bikes to behave as a tandem

bike…………………………………………………………………………93

Figure 4.18 The cadence and power signals of both (coupled) Smart Bikes in a test

scenario with one rider as the trainer/leader……………………………….94

Figure 4.19 The cadence and power signals of both riders on two (coupled) Smart Bikes

in a test scenario with both the trainer/leader and follower………………..95

Figure 5.1 Diagram of adaptive/dynamic control within the exercise session……….103

Figure 5.2 Diagram of session by session adaptive exercise control………………...104

viii
 ACHNOWLEDGEMENTS

My special and greatest gratitude goes to Prof. Kenneth Loparo, for his excellent
leadership and all his support during my PhD program as an advisor and, more
importantly, as a friend. My PhD dissertation has been a collaborative effort between the
Case Western Reserve University, where Dr. Kenneth Loparo and I have been affiliated
with, and Kent State University and Rockwell Automation Company, where Dr. Angela
Ridgel and Dr. Fred Discenzo have been our collaborator, respectively. This project was
supported by National Institute of Health under the grant NIH R21HD068846-01A1
(ALR).

I would like to express the words of appreciation to Dr. Angela Ridgel who has done
the clinical and patient tests, and data collecting at her lab in KSU, and Dr. Fred
Discenzo, Director of Advanced Technology Lab, for all his support.

I extend my acknowledgement to the members of dissertation committee for their

time and suggestions to improve this work. I would like also to thank Dr. Brian Fast and
Tomas Tichy from Rockwell Automation for their help with motion control. I have,
furthermore, to thank the faculty, staff and students in the department for their assistance
and kindness, especially to my friends Donald Moore, Jordan Murray, Hanieh Agharazi,
Nahal Geshnizjani, Fei Ding, Ting-Jung Chen and Amirhossein Sajadi for the good times
we shared together.

I would like also to recognize the roles of many people who were influential in my
preparation to the point that I was qualified to start my PhD. In particular, I have to
mention my previous advisors Prof. Ali Khaki Sedigh and Prof. Hamid Taghirad from
K.N. Toosi University of Technology in Tehran, Iran.

Finally, the important role of family in somebody’s success is undeniable. In this

regard, I would love to express my highest esteem and deepest gratitude to my parents
and all my family members.

ix
 Development of an Intelligent Exercise Platform for Rehabilitation in Parkinson’s Disease

Abstract

by

HASSAN MOHAMMADI ABDAR

In this dissertation, some novel approaches have been employed to develop a

platform for studying new and potentially high-impact therapeutic approaches (e.g.

forced exercise or active-assisted cycling) for improving motor function in individuals

with Parkinson’s disease (PD). These therapeutic methods can readily be implemented

and can provide benefits to patients at lower cost and reduced risk.

First, with an innovative approach, the complexity of biomechanical and

physiological features of forced (tandem) and voluntary (single) cycling was studied

using some advanced signal processing methods such as Approximate Entropy (ApEn),

Sample Entropy (SaEn), and Spectral Entropy (SpEn). The variability analysis results

were used to determine which unique temporal features are positively correlated with

measures of the rider’s motor skill level. These features were then correlated to motor

function improvement in PD patients as measured by the Unified Parkinson’s Disease

Rating Scale (UPDRS) Motor III using a multiple linear regression (MLR) model.

Next, using the results of the first stage, a novel instrumented and automatically

controlled cycle (Smart Exercise Bike) was developed to study the associations between

PD rider performance and changes in motor function. Smart Bikes were examined with a

group of forty-seven PD patients, and the results indicated that the bike imitates tandem

cycling in the Dynamic Mode of operation and provides patient benefits similar to

x
tandem cycling. It can also be used to understand how different types of exercise can

provide therapeutic benefit to PD patients. Moreover, the Smart Bike can be transformed

into a commercial or medical device with data logging and remote access capability for

physicians, trainers, and therapists. The data set collected using the Smart Bike will

provide a basis for dynamically prescribing customized optimal exercise regimen

providing superior therapeutic benefit for individuals with PD in future controlled clinical

studies.

Lastly, to study the interaction between the trainer and the patient on the tandem

bike, the pedals of a tandem bike were coupled electronically (instead of mechanically by

a chain) by equipping a tandem bike with servo motors, drives and various sensors

coupled to a Programmable Logic Controller (PLC). The electronic coupled tandem bike

provides an unprecedented ability to explore the specific benefits observed during

exercise sessions with PD riders at high pedaling rate on a tandem bike but without the

assistance from a human trainer.

The outcomes of this study, including variability analysis results, Smart Exercise

Bike and electronically connected tandem bicycle, provide a unique and valuable

foundation for investigations of the operating and control parameters of the exercise

system that provides the optimum benefit for PD riders. It can also be used as an effective

platform for research on the underlying mechanisms for improvements in motor function

and for readily implementing a feedback system that can dynamically optimize the

benefits from accelerated cycling for individual PD riders.

xi
Chapter 1

Introduction

1.1 Introduction and Motivation

Parkinson's disease (PD), which affects approximately one million people in the

US and 7 to 10 million people worldwide, is a chronic, progressive neurological disorder

that is characterized by the loss of dopaminergic neurons in the brainstem [2, 3, 24, 25].

The main symptoms of the disease are movement disorders, and include shaking or

tremor, muscle stiffness and rigidity, and slowness of physical movements

(bradykinesia). As PD progresses, the combined motor and non-motor symptoms often

lead to decreased independence and increased reliance on caregivers and the healthcare

system. The economic impact of PD, including treatment, social security payments, and

lost income from inability to work, is estimated up to $25 billion per year in the United

States [2].

There is no known cure for this degenerative disease that results in progressive

deterioration of motor skills along with other reduced physical and mental functions. The

accepted treatment for PD is medication (e.g. levodopa) and in some cases surgical

intervention (e.g. deep brain stimulation). These treatments only mask the symptoms and

do not slow progression of the disease. Furthermore, they often have undesirable side

effects, are costly and can introduce additional health risks. Considering these

1
deficiencies, there is a need for innovative treatments to prevent, delay disease

progression, or improve symptoms of PD.

Recent studies have shown that exercise and movement therapies have significant

benefits for individuals with PD, but there is little consensus on the optimal mode or

intensity [2, 3, 5, 8]. Several studies have documented the benefits of high-cadence

tandem cycling for motor function improvement in PD riders [1, 3, 5, and 8]. However,

the effective factors of exercise (i.e. RPM, intensity, intervention type, duration of the

exercise) constitute an optimal exercise intervention for PD patients are still unknown.

Each PD patient has different symptoms and skills level which make it challenging to

design a general rehabilitation system that gives the maximum benefit to all. Moreover,

progression of the disease often requires re-assessments and modifications of the motor

rehabilitation programs.

Despite the remarkable results have been reported from accelerated (individual or

tandem) cycling for PD rehabilitation, there has been little study into the cycling

characteristics that provide the most benefits for riders with PD. No study has reported

the investigation of the exercise parameters and biomechanical characteristics of the

beneficial cycling exercises. Likewise, some questions such as optimum RPMs, duration

of exercise period and frequency (exercise sessions per week) need to be answered. For

any exercise, overexertion or improper performance may provide diminished benefits and

may cause damage. Furthermore, the tools needed to support the clinical studies and

research that uncover the fundamental factors of the cycling and to quantify the dynamics

of motor function improvement do not exist. Existing exercise bikes typically operate at a

rate determined by the rider with pedaling resistance preset. Some bikes provide

2
minimum instrumentation such as a cadence counter and energy monitor, but capabilities

like data archiving, dynamic load change and feedback control which are essential for

research and clinical studies are non-existent.

Furthermore, even with the exceptional results reported from tandem cycling,

large-scale use of the tandem cycling paradigm for exercise therapy is not feasible. First,

tandem cycling requires an able-bodied trainer to assist in pedaling that is not reasonable

in large-scale clinical deployment or in-home use. Second, variability in trainer pedaling

speed, stamina, and response to the PD rider’s performance creates variations that make

data analysis and conclusions in clinical studies difficult to generalize. Third, there are a

number of factors, such as cadence, foot position and workload that can affect the

biomechanics of cycling. Many motorized single-rider stationary exercise bikes are

commercially available today that can provide a pre-programmed load profile for the

rider. However, it has not been possible to reproduce the dynamics of the tandem bike

cycling paradigm using currently available motorized cycles.

1.1.1 Parkinson’s Disease and Exercise Therapy

Many studies have documented the positive impact of physical therapy and exercise

as a treatment for Parkinson's disease. Based on published research in both animals and

humans with PD, exercise has been shown to have a positive impact in improving motor

function in PD and to facilitate neuroplasticity [11]. The principle interest in employing

exercise for neuro-rehabilitation in PD is that it incorporates many aspects of the

practices that are effective for motor skill learning. These factors include intensity,

3
repetition, and challenge that together with skill training lead to motor function

improvement in PD patients [11].

Many studies with animal models have revealed that high intensity exercise can

promote neural plasticity and neuroprotection against dopaminergic cell loss [12]. In

addition, several studies in humans have revealed that high intensity treadmill training [9,

10] and high cadence cycling [1, 2, 3, 5, and 8] can improve motor function. Ridgel [3, 5]

and Alberts [17] have shown that high cadence cycling can result in significant

improvement in motor symptoms as measured with the Unified Parkinson’s Disease

Rating Scale (UPDRS) Motor III test. Ridgel has presented Active-Assisted Cycling

(AAC) at fast rates to be well tolerated and to result in immediate reductions in tremor

and bradykinesia in most PD participants [5]. Feodoroff has reported remarkable

improvements in walking ability, hand coordination and quality of life as the primary

results of an AAC study that proposed that AAC might affect central motor control

processes in PD patients [30].

Resistance training has also been reported as a positive exercise program for PD

patients with mild-to-moderate disease levels with demonstrated improvements in

performance or intensity comparable to the level of normal adults of the same age.

Resistance training can provide functional improvements in gait and, could be beneficial

as part of a physical rehabilitation and health support program for PD patients [15]. In

[18], active-assisted cycling and resistance training are combined and introduced as the

multifaceted intervention, and has shown that the older adult and the older adult

diagnosed with PD showing both tolerance and health-related improvements in physical

fitness following the multifaceted intervention.

4
1.1.2 High-Rate Cycling and Rehabilitation in Parkinson’s Disease

Several studies in rehabilitation of PD have shown that (high cadence) cycling can

reduce the symptoms of the disease [1, 2, 3, 5, and 8]. Ridgel [3, 5] and Alberts [17] have

shown that high cadence cycling can result in significant improvement in motor

symptoms as measured with the Unified Parkinson’s Disease Rating Scale (UPDRS)

Motor III test. In [3], a novel approach was introduced to increase exercise intensity in

individuals with PD referred to as forced exercise (FE). This method used a stationary

tandem bicycle and an able-bodied cyclist (trainer) to assist individuals with PD to pedal

at a cadence between 80-90 RPM (revolutions per minute, RPM). This high-cadence

cycling was roughly 30% faster than they were able to pedal on their own at a self-

selected rate that was determined during the baseline evaluation.

High-cadence tandem cycling resulted in a profound reduction in motor symptom

characteristics of people with PD as measured with standard motor performance tests

(e.g. UPDRS). In particular, tremors and bradykinesia were significantly reduced. These

improvements were global in that upper extremity motor function was enhanced after this

lower extremity exercise. The observed global improvement was supported by fMRI data

(functional brain scan) that showed increased levels of activation within cortical

structures such as the supplementary motor area [17]. Alberts and his colleagues [17]

have also compared the effect of FE (off medication) with the effect of medication by

examining the percent of signal change in the side contralateral to the task. They

observed strong correlations, indicating a similar change in BOLD MRI response for FE

and medication, and decrease in UPDRS-III ratings in same patients by 35% and 32%

after FE and on medication compared with off medication, respectively. Imaging data

5
showed a significant correlation between FE and medication for areas in the basal ganglia

and cortex that indicates that FE and medication employ similar pathways to provide

symptomatic cure.

These findings support the hypothesis that high cadence (tandem) cycling can

promote central motor control processes in PD riders. If high-rate cycling therapy proves

to be an effective therapeutic approach, this may permit altering the traditional treatments

prescribed to individuals with PD through reducing the dependency on medications. A

schematic in [17] presents the anticipated effect of forced exercise (FE) on central

nervous center (CNS) structure and function. It is proposed that FE results in an increase

in the quantity (high rate of pedaling) and quality (consistent pedaling pattern) of intrinsic

feedback from the Golgi tendon organs (GTO) and muscle spindles [17].

However, individuals that cycled on a single stationary ergometer at similar aerobic

intensity, but at a self-selected pedaling rate (voluntary exercise) showed no change in

UPDRS Motor III scores.

1.1.3 Dance Therapy for PD Patients

Several researches have reported dance as an effective alternative to traditional

exercise that addresses many of the suggested elements for exercise paradigms designed

for individuals with PD [26-28]. The benefits of dance include improved balance and gait

function as well as improved quality of life. Most studies of dance for PD have included

primarily individuals with mild to moderate PD. While benefits can be obtained with a

short, intensive dance intervention, longer interventions may prove to be more effective

[26].

6
1.1.4 Motor Function Evaluation

The Unified Parkinson’s Disease Rating Scale (UPDRS) Part III motor exam is

administered for all PD patients participated in exercise sessions while individuals are off

anti-Parkinsonian medication for 12 hours. Assessments are performed prior to and after

completion of the desired exercise sessions. The difference in UPDRS Motor III scores

between these two time periods is calculated and is used in analysis and evaluations. A

negative score change represents improvement while a positive change indicates

worsening of motor symptoms.

Motor function of the upper extremity is assessed with the Kinesia

MotorAssessment System (Cleveland Medical Devices, Ohio). Kinesia provides

quantitative scores for the degree of tremor, bradykinesia and quality of upper extremity

movement [8]. Subjects will work through seven upper extremity motor tasks while

wearing the device. Outcome variables will include tremor score, amplitude, speed and

frequency of movement. Motor function of the lower extremity is assessed using the

Functional Assessment of Biomechanics System (FAB, Biosyn Systems, Canada). The

FAB includes inertial sensors that combine accelerometers, gyros and earth’s magnetic

field sensors to allow for real-time detection of any angular displacement. Lower

extremity motor function is assessed using the Timed Up and Go (TUG) and the 6-minute

Walk Test (6MW). During the TUG, the subject will be asked to stand up from a standard

chair and walk a distance of approximately 3 meters, turn around and walk back to the

chair and sit down again. The 6MW is a measure of the distance an individual can walk

in 6 minutes to assess overall locomotion and fatigue. Both tests have been widely used

in PD/older adult exercise literature to assess lower body function. Outcome variables

7
will include time to completion, distance traveled, stride length, cadence, gait velocity,

step width and double limb support time [2].

1.2 Contributions

In this dissertation, some novel approaches have been developed and deployed for

studying new and potentially high-impact therapeutic approaches (e.g. forced exercise or

active-assisted cycling) for improving motor function in individuals with PD. This

therapeutic method can readily be implemented and may provide benefits similar to some

medications but at a lower cost.

First, with an innovative approach, the complexity of biomechanical and

physiological features of forced (tandem) and voluntary (single) cycling was investigated

using some advanced signal processing methods, Approximate Entropy (ApEn), Sample

Entropy (SaEn), and Spectral Entropy (SpEn). The variability analysis results were used

to determine which unique temporal features are positively correlated with measures of

the rider’s motor skill level. These features were then correlated to motor function

improvements in PD patients as measured by the Unified Parkinson’s Disease Rating

Scale (UPDRS) motor III, using a multiple linear regression (MLR) model.

In addition, using the results of the first stage, a novel instrumented and

automatically controlled cycle (Smart Exercise Bike) has been designed, built and used as

a clinical instrument to study the associations between PD rider performance and changes

in motor function. The developed Smart Bike has been examined with a group of forty-

seven PD patients and the results of these experiments showed that the bike imitates

tandem cycling in the Dynamic Mode of operation and this provides benefits to the

8
patients similar to tandem cycling. It can also be used to understand how different types

of exercise can provide therapeutic benefit to PD patients. Moreover, the developed

Smart Bike can be readily transformed into a consumer device with data logging and

remote access capability for physicians, trainers, and therapists. The data set collected

using the Smart Bike will provide a basis for dynamically prescribing a customized

optimal exercise regimen providing superior therapeutic benefit for individuals with PD

in future controlled clinical studies.

Lastly, to study the interaction between the trainer and the patient on the tandem

bike, the pedals of a tandem bike were coupled electronically (instead of mechanically by

the chain) by equipping a tandem bike with servo motors, drives and various sensors

coupled to a Programmable Logic Controller (PLC). The electronically coupled tandem

bike provides an unprecedented ability to explore the specific benefits observed with PD

riders during sessions of high pedaling rate on a tandem bike but without the uncertainties

and variability from using a human trainer.

The outcomes of this study, including variability analysis results, Smart Bike and

electronically connected tandem bicycle, provide a unique and valuable foundation for

investigations of the operation and control of the exercise system that provides the

optimum benefit for the PD rider. These results also provide an effective platform for

further research on the underlying mechanisms for improvements in motor function and

for readily implementing a feedback system that can dynamically optimize the benefits

from accelerated cycling for individual PD riders.

Because high cadence (accelerated) cycling therapy has been proven to be an

effective therapeutic aid, this platform may permit altering the treatment prescribed to

9
individuals with PD to reduce medication levels, delay dosage increases, or potentially

delay or eliminate the need for surgery. Furthermore, this work will support future

research studies to develop an adaptive exercise program to optimize the therapeutic

benefit for riders with PD or other neurological disorders such as stroke or spinal cord

injury.

1.3 Thesis Organization

In this dissertation, the next chapter discusses the variability analysis of collected data

of two groups of PD patients who had exercised in tandem and single cycling groups,

using some advanced signal processing techniques. Chapter 3 presents the design,

development and validation of a smart exercise bike based on the variability analysis

results obtained in Chapter 2. Electronically coupling of the riders on a tandem bike

while removing the mechanical link (i.e. shared chain) is discussed in Chapter 4. Lastly,

Chapter 5 presents the summary of the research as well as some ideas for the future

directions of the study.

10
Chapter 2

Variability Analysis of PD Patient Data,

Forced Exercise and Voluntary Cycling

2.1 Introduction

Exercise and movement therapies have been shown to benefit individuals with PD

by improving their motor function, but there is little consensus on the optimal mode or

intensity [1, 2, 5, 8]. In [5] a novel approach was developed to increase exercise intensity

in individuals with PD called forced exercise. This approach used a stationary tandem

bicycle (Figure 2.1) and an able bodied cyclist to assist individuals with PD to pedal at a

cadence [revolutions per minute (rpm)] between 80–90 rpm. This cadence was roughly

30% faster than they were able to pedal on their own at a self-selected rate that was

determined during baseline evaluation. Crank-based power-meters (Schoberer Rad

Meßtechnik [SRM]) measured the work of the individuals with PD during each exercise

session. Forced exercise results in a significant improvement in motor symptoms as

measured with the Unified Parkinson’s Disease Rating Scale (UPDRS) Motor III. This

clinical scale evaluates the degree of tremor, bradykinesia, rigidity, and posture/gait

difficulties in individuals with PD. However, individuals that cycled on a single

stationary ergometer at similar aerobic intensity, but at a self-selected pedaling rate

(voluntary exercise) showed no change in UPDRS Motor III scores.

11
 Heart Rate

bpm 100

50

0
0 200 400 600 800 1000 1200 1400 1600 1800
Time (sec)
Cadence
100
rpm

50

0
0 200 400 600 800 1000 1200 1400 1600 1800
Time (sec)
Power Patient Trainer
200

150
watts

100

50

0
0 200 400 600 800 1000 1200 1400 1600 1800
Time (sec)

Figure 2.1: Representative training data (cadence, HR, and trainer and patient power) during a 30-minute
exercise block of forced exercise.

In this chapter, the complexity of biomechanical and physiological features of

forced and voluntary cycling is examined and the relationship between these features and

improvements in motor function as measured by the UPDRS Motor III scale is studied.

We hypothesize that temporal variability or lack of predictability in cadence during

forced cycling can be used to accurately predict improvements in UPDRS Motor III

scores. Subject and trainer data (power, heart rate, and cadence), collected previously

from exercise training sessions using a stationary tandem bicycle published in [3], were

examined using sophisticated signal processing techniques: approximate entropy (ApEn),

sample entropy (SaEn) , and spectral entropy (SpEn). ApEn is a regularity statistic that

12
quantifies the unpredictability of temporal fluctuations in a time series such as an

instantaneous heart rate time series. The presence of repetitive temporal patterns in a time

series renders it more predictable than a time series in which such patterns are absent.

ApEn quantifies the likelihood that “similar” patterns of observations will not be followed

by additional “similar” observations. A time series containing many repetitive patterns

has a relatively small ApEn; a less predictable (i.e., more random or less time-correlated)

time series will have a greater ApEn. SaEn is a modification of ApEn [6] that removes the

potential bias in ApEn [8], eliminates self-matches in the computation, reduces

computational complexity, and can be applied to short time series data. Both quantify the

predictability (or regularity) in a time series, and are useful in quantifying differences in

health and disease [6, 7, 31, 32]; whereas SpEn regularity or lack of regularity in the

distribution of frequency content in a time series.

More sophisticated analysis of the exercise performance variables across groups

and throughout exercise sessions will aid in identifying and quantifying time and

frequency domain features that may be responsible for the improved motor performance

observed after forced exercise. Previous studies have shown that behavioral effects of

forced and voluntary exercise were dramatically different [3]. A precise understanding of

specific and differentiating characteristics between forced and voluntary exercise will

provide important guidance in the development of more cost and time effective methods

of delivering forced exercise than tandem cycling, such as motorized single bikes that

will not require a trainer.

13
2.2 Patient Exercise and Data Collection (Methods and Protocols)

2.2.1 Forced and Voluntary Cycling Data Collection

Ten individuals with idiopathic Parkinson’s disease were assigned to one of two

groups: forced (tandem) or voluntary (single) cycling [3]. Both groups completed 24 one-

hour exercise sessions (3 per week) over an 8 week period1. In the forced cycling group,

an able-bodied cyclist/certified personal trainer riding on the front of a stationary tandem

bike assisted the PD subject. The trainer had the objective of maintaining bike operation

at an accelerated cadence rate between 80 and 90 rpm. The voluntary cycling group

pedaled a stationary single bike (SRM indoor trainer, Jülich, Germany) at a self-selected

cadence of roughly 60 rpm. The pedaling cadence and power performed by the subject

and the trainer on the tandem and on the single bicycle were measured using SRM

power-meters. A Polar heart rate monitor (Polar Electro, Lake Success, NY) was used to

collect heart rate data. Approximately 20 or more data sets for each person were collected

for exercise sessions across the 8-week intervention. Heart rate (HR), cadence and power

variables were measured with sampling rate of one Hz. In order to examine the raw data

for single (voluntary) and tandem (forced) rider exercise tests, the mean and standard

deviation of the “heart rate”, “power” and “cadence” signals were averaged for 20-24

sessions per subject (table 2.1). Additional methodological details can be found in Ridgel

et al 2009 [3]. Data were collected under an IRB study approved by the Cleveland Clinic

Institutional Review Board.

1
The experiment conducted at Cleveland Clinic Foundation

14
2.2.2 Motor Function Assessment

The Unified Parkinson’s Disease Rating Scale (UPDRS) Part III motor exam was

administered while individuals were “off” anti-Parkinsonian medication for 12 hours.

Assessments were performed prior to and after completion of the eight week exercise

intervention. The difference in UPDRS Motor III scores between these two time periods

was calculated and was used in the analysis and models (described later). A negative

score represents improvement while a positive score indicates worsening of motor

symptoms.

Table 2.1: Mean and variance for power, heart rate and cadence signals.
Patient Group UPDRS Heart rate Power Cadence
ID Pre Post Change Mean Stdev Meanb Stdev
a b
Meanb Stdev
1 Single 73 64 -9 100.86 3.24 42.07 9.27 58.35 6.95
2 Single 45 45 0 116.20 2.46 73.77 7.01 43.60 2.80
3 Single 48 52.5 +4.5 124.84 6.55 105.26 22.81 75.72 6.81
4 Single 49 59 +10 110.43 2.96 41.36 5.42 57.38 3.94
5 Single 30 45 +15 156.10 3.92 72.83 9.27 69.27 3.91
MEAN 49 53.1 +4.1 121.68 67.05 60.86
6 Tandem 58 35 -23 113.86 4.16 45.81 14.45 86.64 1.83
7 Tandem 65 42 -23 119.54 4.93 24.46 13.64 85.67 2.80
8 Tandem 45 28 -17 121.67 4.41 63.30 16.00 84.72 1.73
9 Tandem 36 24.5 -11.5 122.51 7.51 50.63 24.45 80.49 2.70
10 Tandem 38 29.5 -8.5 108.94 4.37 68.07 16.80 85.31 1.99
MEAN 41.4 31.8 -16.6 117.30 50.45 84.56
a
Negative change in UPDRS represents improvements in motor function.
b
Mean values were calculated over 20-24 exercise sessions per patient.

2.3 Variability Data Analysis and Signal Processing Methods

Before applying the signal processing techniques, each of the methods are briefly

described.

2.3.1 Approximate Entropy

Approximate Entropy (ApEn ) is used to quantify the degree of temporal

regularity or predictability in the time series data for power, heart rate and cadence. The

15
algorithm for computing ApEn has been published elsewhere [6, 7]. Here, a brief

summary of the calculations is presented as applied to a time series of instantaneous heart

rate measurements, HR(i). Given a sequence SN, consisting of N instantaneous heart rate

measurements HR(1), HR(2), …, HR(N), and a time delay  1, we chose values for two

input parameters, m and r, to compute the Approximate Entropy, ApEn(SN,m,r), of the

time series. The parameter m specifies the patternlength, and the parameter r defines the

criterion of similarity. We denoted a subsequence (or pattern) of m heart rate

measurements, beginning at measurement i within SN, by the vector pm(i). Two patterns,

pm(i) and pm(j), are similar if the difference between any pair of corresponding

measurements in the patterns is less than the tolerance r, i.e., if

| HR(i  k )  HR( j  k ) | r for 0  k  m (2.1)

Now consider the set Pm of all patterns of length m [i.e., pm(1), pm(2), …, pm(N-

m+1)], within SN. We now define

(2.2)

Here nim(r) is the number of patterns in Pm that are similar to pm(i) (given the

similarity criterion r), and the quantity is the fraction of patterns of length m that

resemble the pattern of the same length that begins at interval i. We can calculate

for each pattern in Pm, and define as the mean of these values. The

quantity expresses the prevalence of repetitive patterns of length m in SN. Finally,

the ApEn of SN for patterns of length m and similarity criterion r, is given by

[ ] (2.3)

Note, ApEn(SN,m,r) is the natural logarithm of the relative prevalence of repetitive

patterns of length m compared with those of length m+1.

16
 ApEn estimates the logarithmic likelihood that the next intervals after each of the

patterns will differ (i.e., that the similarity of the patterns is mere coincidence and lacks

predictive value). Smaller values of ApEn imply a greater likelihood that similar patterns

of measurements will be followed by additional similar measurements. If the time series

is highly irregular, the occurrence of similar patterns will not be predictive for the

following measurements, and ApEn will be relatively large. It should be noted that ApEn

has significant weaknesses, notably its strong dependence on sequence length and its

poor self-consistency (i.e., the observation that ApEn for one data set is larger than ApEn

for another with a given choice of m and r, does not necessarily hold true for other

choices of m and r).

2.3.2 Sample Entropy

Sample entropy (SaEn) is very similar to ApEn and both methods calculate the

probability that epochs of window length m that are similar within a tolerance r remain

similar at the next point. SaEn is a modification of ApEn (defined by Pincus [6])

introduced by Richman and Moorman [31] to reduce the bias caused by including self-

matches and to also simplify the calculation somewhat.

2.3.3 Spectral Entropy

Spectral entropy (SpEn) is used to measure the complexity of time series data in the

frequency domain and is used to quantify frequency domain variability in power by

computing the power-spectral-density (PSD) of the time series. The PSD is normalized to

produce a probability-like density function and transformed with the Shannon function as

follows:

17
 1. Compute the Power Spectral Density (PSD) of the signal, P(f).

2. Normalize the power spectrum:

∑
(2.4)

3. Compute the Shannon function:

(2.5)

4. Compute the Spectral Entropy (SpEn ):

∑ (2.6)

Where N is the number of data points.

2.3.4 Multiple Linear Regression

Multiple linear regression (MLR) is a method used to model the linear

relationship between a dependent variable and two or more independent variables. The

dependent variable is sometimes also called the predictand, and the independent variables

called the predictors. MLR is based on least squares where the model is fit such that the

sum-of-squares of differences of observed and predicted values is minimized. The MLR

model is:

yi = b0 + b1xi,1 + b2xi,2 + … + bkxi,k + ei i = 1, 2, …, n (2.7)

Where:
n: Number of predictand (or observations)
k: Total number of predictors
b0: Regression constant
bk: Coefficient of the kth predictor
xi,k: Value of the kth predictor in observation i
yi: Predictand number i
ei: Error term

Here, the parameters of the model are estimated using a least squares approach (there are

some other methods), and the resulting prediction equation is:

18
 ̂ ̂ ̂ ̂ (2.8)

Where “^” denotes estimated values.

The error term in equation (2.7) is unknown because the real model is unknown.

Once the model has been estimated, the regression residuals are defined as:

̂ ̂ (2.9)

yi: Observed value of predictand i

̂ : Predicted value of predictand i

The residuals measure the closeness of fit of the predicted values and actual

predictand in the calibration period. The algorithm for estimating the regression equation

(solution of the normal equations) guarantees that the residuals have a mean of zero for

the calibration period. In this work the MATLAB function “regress” has been used to

find the model coefficients and other parameters.

2.3.5 Logistic regression

Logistic regression (also called the logistic model or logit model) is used for

prediction of the probability of occurrence of an event by fitting data to a logistic curve.

It is a generalized linear model used for binomial regression. Similar to many forms of

regression analysis, it makes use of several predictor variables that may be either

numerical or categorical.

The logistic curve relates the independent variable X, to the rolling mean of the

dependent variable (DV), P (Y). The formula may be written as:

or (2.10)

19
 Where P is the probability of a 1 (the proportion of 1s is the mean of Y), e is the base

of the natural logarithm (about 2.718), and a and b are the parameters of the model. The

value of a yields P when X is zero, and b adjusts how quickly the probability changes

with changing X a single unit. Because the relation between X and P is nonlinear, b does

not have a straightforward interpretation in this model as it does in ordinary linear

regression.

2.3.6 Odds ratio

The odds ratio is the ratio of the odds of an event occurring in one group to the odds

of it occurring in another group. The term is also used to refer to sample-based estimates

of this ratio. These groups might be men and women, an experimental group and a

control group, or any other binary classification. If the probability of a 1 is P, then the

probability of a 0 will be (1-P) and the odds will be:

or (2.11)

2.4 Biomechanical and Physiological Feature Analysis

In this step, three main parameters; Approximate Entropy (ApEn), Sample Entropy

(SaEn ) and Spectral Entropy (SpEn ) for heart rate, power and cadence signals were

computed in all data sets for each person. Table 2.2 shows the values of ApEn, SaEn and

SpEn for the cadence signal of all 21-exercise sessions for one sample patient (with time

delays,   1 and   3 ). These parameters were used to quantify the variability in terms of

both temporal and frequency domain patterns. The values for ApEn and SaEn that were

computed for the dataset are approximately the same, so further analysis focused on the

20
ApEn computations. The mean of these parameters for all exercise sessions were

computed for the single (voluntary) and tandem (forced) groups (Table 2.3).

Table 2.2: The values of ApEn, SaEn and SpEn for cadence signal of 21 exercise sessions of one
patient (with time delays,   1 and   3 ).

ApEn SaEn SpEn

Session ID  1  3  1  3 Subject Trainer
P020507_DP 1.2291 1.4561 1.1707 1.5078 0.0718 0.0717
P070507_DP 1.1209 1.3695 1.0246 1.4042 0.0716 0.0716
P090507_DP 1.1893 1.4794 1.1085 1.4981 0.0714 0.0714
P110507_DP 0.1250 0.1908 0.1102 0.1691 0.0711 0.0711
P140507_DP 1.1154 1.3704 1.0280 1.3382 0.0711 0.0711
P160507_DP 1.0656 1.3387 0.9788 1.3411 0.0715 0.0715
P180507_DP 1.1662 1.4107 1.1138 1.4337 0.0714 0.0714
P210507_DP 1.1152 1.3969 0.9732 1.3440 0.0710 0.0710
P230407_DP 0.4482 0.6814 0.3993 0.6031 0.0714 0.0712
P230507_DP 1.1634 1.4127 1.0875 1.4209 0.0713 0.0712
P250407_DP 0.0426 0.0713 0.0348 0.0567 0.0718 0.0715
P250507_DP 1.1816 1.4581 1.0536 1.4564 0.0712 0.0713
P270407_DP 1.2066 1.4977 1.1065 1.5643 0.0716 0.0716
P300407_DP 0.3787 0.6001 0.3377 0.5469 0.0713 0.0713
P300507_DP 1.1470 1.4335 1.0442 1.4061 0.0714 0.0714
P040407_DP 1.2894 1.4300 1.3336 1.6352 0.0718 0.0717
P060407_DP 1.0507 1.2200 0.9719 1.2804 0.0716 0.0716
P090407_DP 1.1808 1.3661 1.1731 1.5081 0.0714 0.0714
P110407_DP 1.1588 1.3103 1.2453 1.4947 0.0711 0.0711
P130407_DP 1.0950 1.3161 0.9975 1.3473 0.0711 0.0711
P180407_DP 1.1256 1.3503 1.1315 1.4750 0.0715 0.0715
P200407_DP 1.0755 1.2444 1.0503 1.4486 0.0714 0.0714

Statistical analysis, including linear regression, logistic regression and computation

of the odds ratio, were used to investigate the relationship between the measures of

21
exercise variability and the exercise related change in motor performance as measured by

UPDRS Motor III scores. The data in Tables 2.1 and 2.3 were used to develop a multiple

linear regression model (MLR) and apply logistic regression to determine the odds ratio

for achieving a negative change (improvement) in the UPDRS Motor III scores in each

group. Each regression model has four independent variables: ApEn(heart rate),

ApEn(power), ApEn(cadence) and SpEn (power), and one dependent variable: UPDRS

score.

Table 2.3: ApEn and SpEn for power, heart rate and cadence signals.
Patient Group UPDRS Power Heart rate Cadence Power
ID change (ApEn) (ApEn) (ApEn) (SpEn )
1 Single -9 0.1379 0.4017 0.2048 0.1053
2 Single 0 0.2292 0.5366 0.5653 0.0782
3 Single +4.5 0.2600 0.0924 0.1658 0.0854
4 Single +10 0.4080 0.6235 0.4417 0.0913
5 Single +15 0.2024 0.2974 0.5128 0.0843
MEAN +4.1 0.2475 0.3903 0.3780 0.0889
6 Tandem -23 0.0563 0.1932 1.0540 0.1072
7 Tandem -23 0.0450 0.0961 1.0411 0.2190
8 Tandem -17 0.0347 0.1997 1.2030 0.1050
9 Tandem -11.5 0.0106 0.0878 0.7739 0.1663
10 Tandem -8.5 0.0133 0.1498 0.9850 0.0990
MEAN -16.6 0.0319 0.1453 1.011 0.1393

The data were examined in two ways: first two separate MLR models were built, one

model for the single sessions and one model for the tandem sessions (Table 2.4) and then

the data were combined into a single dataset and a single MLR model was built (Table

2.5). The residual values in Tables 2.4 and 2.5, and the predicted UPDRS values in

Figure 2.1 were obtained from MLR modeling using the “regress” function in

MATLAB. Logistic regression values and the odds ratio values were calculated using

equations (2.10) and (2.11). The predicted UPDRS Motor III scores were obtained using

MLR as described above. A Pearson product-moment correlation coefficient was

22
computed to assess the relationship between the real and predicted values of UPDRS

Motor III scores.

Dependent variables (e.g. heart rate, power, cadence, power ApEn, heart rate ApEn,

Cadence ApEn, Power SpEn ) were compared between single and tandem groups using a

One-way ANOVA (SPSS, Inc, version 18). Significance was set at P ≤0.05.

2.5 Variability Analysis Results

There were no significant differences in heart rate and power between the two groups

(Table 2.1). However, the pedaling cadence showed a significant difference (F1,8= 17.8,

P=0.003) between the raw values in voluntary (60.8 ± 12.3 rpm) and forced (84.5 ± 2.4

rpm) groups. Cadence for the forced exercise sessions was higher than the voluntary

sessions with less variability as quantified by the standard deviation.

For SpEn, only the power signals are distinguishable between the single and tandem

groups. Comparison of each variable shows clear differences. ApEn for the power in the

single sessions (0.247 ± 0.10) is significantly greater (F1,8= 22.2, P=0.002) than the ApEn

for the power in the tandem sessions (0.032 ± 0.02). ApEn for the heart rate in single

sessions (0.390 ± 0.09) is also significantly greater (F1,8= 6.51, P=0.034) than the ApEn

for the heart rate in the tandem group (0.145 ± 0.02). This suggests that the power and

heart rate signals in the voluntary (single) group have greater variability (are less

predictable) than the signals in forced (tandem) group. The results for the cadence signal

are opposite; that is, the cadence for the single group (0.378 ± 0.18) show less variability

(are more predictable) and are significantly less (F1,8=35.05, P< 0.0001) than the cadence

for tandem sessions (1.01 ± 0.15). Spectral Entropy of the power (SpEn) in the tandem

23
group (0.139 ± 0.05) showed slightly greater, but not significant, variability (F1,8=4.48,

P= 0.67) than in the single sessions (0.089 ± 0.01).

Predicted values of UDPRS, using MLR analysis, are plotted against real (measured)

values in Figure 2.2. Four out of the five individuals who completed single (voluntary)

sessions showed no improvement or worsening of UPDRS Motor III scores while one

individual showed a slight improvement (Figure 2.2(b)). The MLR model is less accurate

in individuals whose scores worsened. All participants who completed forced exercise

sessions showed improvement in the UPDRS Motor III scores (Figure 2.2(b)) and the

model shows more accurate predictions in this group. The combined model results in

greater differences between the predicted and real scores for most subjects (Figure

2.2(b)).

24
 (a) (b)
Figure 2.2: (a) Real (measured) versus predicted UPDRS values; A. single (voluntary) exercise
data; B. tandem (forced) data; C. combined single and tandem data. (b) Correlation analysis of
real (measured) versus predicted UPDRS values; A. single (voluntary) exercise data; B. tandem
(forced) data; C. combined single and tandem data.

There was a positive but not significant correlation between the real and predicted

UPDRS Motor III scores in the voluntary exercise (single) group (figure 2.2(a), r = 0.875,

N = 5, P = 0.052). However, there was a significant correlation in the forced exercise

(tandem) group (Figure 2.2(a), r = 0.997, N = 5, P < 0.001). When the two groups were

combined, there was also a significant and positive correlation between real and predicted

UPDRS values (Figure 2.2(a), r = 0.858, N = 10, P = 0.001).

25
 Table 2.4: Regression analysis results for single and tandem rider tests.
Patient Group UPDRS Residuals P (Logistic P/(1-P)
ID change regression) Odds Ratio
1 Single -9 -0.78 0.991 1.76e+04
2 Single 0 5.99 0.001 2.50e-03
3 Single +4.5 3.37 0.001 3.83e-04
4 Single +10 -2.77 0.00 7.24e-04
5 Single +15 -4.99 0.00 4.52e-05
MEAN +4.1 0.16 -
6 Tandem -23 -0.28 1.00 1.29e+10
7 Tandem -23 -0.05 1.00 1.02e+10
8 Tandem -17 0.73 1.00 1.15e+07
9 Tandem -11.5 0.20 1.00 8.07e+04
10 Tandem -8.5 -0.69 0.99 9.86e+03
MEAN -16.6 -0.02 -

Table 2.5: Combined single and tandem rider data regression analysis results.
Patient Group UPDRS Residuals P(Logistic P/(1-P)
ID change regression) Odds Ratio
1 Single -9 5.66 0.96 2.82e+01
2 Single 0 0.21 0.45 8.11e-01
3 Single +4.5 5.36 0.00 5.21e-05
4 Single +10 -0.03 0.00 4.68e-05
5 Single +15 -13.72 0.22 2.78e-01
MEAN +4.1 -0.50 -
6 Tandem -23 10.30 1.00 3.26e+05
7 Tandem -23 1.07 1.00 3.32e+09
8 Tandem -17 1.93 1.00 3.51e+06
9 Tandem -11.5 -5.73 1.00 3.03e+07
10 Tandem -8.5 -5.06 1.00 7.74e+05
MEAN -16.6 0.50 -

The MLR models also were used to compute the logistic regression (P) and the odds

ratio (P/(1-P) ) given in tables 2.4 and 2.5. In fact, the logistic regression values here

represent the probability of getting improvement after finishing the exercise sessions for

each patient.

2.6 Session by Session Variability Analysis and Prediction

Earlier in this chapter the data for two groups of PD patients which had had been

assigned to forced (tandem) or voluntary (single) exercise groups, were analyzed using

26
entropy signal processing techniques (ApEn, SaEn, and SpEn). These methods were

applied to all exercise sessions for each patient and the average values of these

parameters for each patient for all sessions were used to compare the groups and develop

the MLR model to predict the UPDRS changes based on the exercises and physiological

parameters. The models were shown to predict the UPDRS change after the completion

of all exercise sessions. It would be more appropriate from a therapeutic perspective if it

would be possible to predict the UPDRS change after each exercise session.

Here, the MLR model previously obtained from the variability data analysis

(combined single and tandem data sets) is applied to the variability analysis data of each

session to predict the UPDRS change after each exercise session. The regression model

has four independent variables: ApEn(heart-rate), ApEn(power), ApEn(cadence) and

SpEn (power), and one dependent variable: UPDRS score. So, the independent variables

for every exercise session of each patient are used with the model to predict the

dependent variable, UPDRS change for that session. The results of the session by session

predictions for the exercise sessions of four patients are shown in figure 2.3. The real

UPDRS change was measured only after finishing all sessions, but the predicted UPDRS

change was calculated for each exercise session. The results show that the session by

session prediction yields the final prediction which in most cases is close to the real

UPDRS change. In future work, these session by session predictions will be used in the

development of adaptive exercise strategies that provide more beneficial rehabilitation for

PD patients through the adjustment of related exercise parameters for each session.

27
 Session by Session UPDRS Prediction
25
Real
Predicted

20

15
UPDRS Change

10

0
0 2 4 6 8 10 12 14 16 18 20
Sessions

(a)
Session by Session UPDRS Prediction
2
Real
Predicted
0

-2

-4
UPDRS Change

-6

-8

-10

-12

-14

-16
0 5 10 15 20 25
Sessions

(b)

28
 Predicted UPDRS by Session
2
Session by Session Predicted
Real UPDRS Change
0

-2

UPDRS Change
-4

-6

-8

-10

-12
0 2 4 6 8 10 12 14 16 18 20
Sessions

(c)
Session by Session UPDRS Prediction
4
Real
Predicted
3.5

2.5
UPDRS Change

1.5

0.5

-0.5
0 2 4 6 8 10 12 14 16 18 20
Sessions

(d)
Figure 2.3: Session by session UPDRS prediction of four PD patients. (a) Positive change with
correct prediction, (b) Negative change with imprecise prediction, (c) Negative change with
correct prediction, (d) No change with precise final prediction.

2.7 Conclusion

Although many studies have documented the benefits of exercise, it is unclear

what elements (i.e., dosage, intensity, intervention type) constitute an optimal exercise

program for a given PD patient. Each individual with PD has different symptoms and

29
capabilities that make it challenging to design a single rehabilitation program that is

optimal for all. Furthermore, progression of the disease often requires reassessments and

changes to the rehabilitation programs.

This study reveals that pattern irregularity in HR and power is greater in the single

sessions when compared to the tandem sessions, indicating that the trainer provides a

“stabilizing” influence on the patient’s exercise intensity, while maintaining elevated

cadence. The single PD rider has a tendency to introduce greater variability (less regular

patterns) in power output, inducing greater fluctuations in HR, when compared to tandem

PD riders. In contrast, the cadence signal shows greater variability during the tandem

sessions. This variability is likely due to the inability of individuals with PD to maintain a

constant high-speed pedaling cadence. Furthermore, variability was also introduced when

the able-bodied trainer was required to increase or decrease pedal speed to maintain the

desired cadence. The single PD riders rode at a self-selected cadence and thus showed

lower variability during exercise bouts. This supports our hypothesis that temporal

variability or lack of predictability (not quantified by conventional statistical parameters

such as variance or coefficient of variation) in cadence during forced exercise can be used

to accurately predict resulting improvements in UPDRS Motor III scores. Lastly,

predicted UPDRS Motor III scores using a MLR model are highly correlated to measured

scores in the tandem sessions. These data provide insight into how times series analysis

methods can be applied to uncover potential features in the measured variables and how

this information can be used to correlate exercise parameters with improved motor

function. Furthermore, applying the calculated MLR model to the variability analysis

data of each session indicates that such a model can be used to predict the UPDRS

30
change after each exercise session so that the exercise parameters could be adjusted more

precisely to achieve the maximum benefit.

Typically, researchers use the mean and standard deviation or standard error of

the mean to define variability in a dataset. These measures provide a description of the

magnitude of the variability around a central point. However, the presence of certain

patterns or shifts in patterns can often provide important insight into health status or

motor performance [33], [34]. Assessment of irregularities of serial data using entropy

statistics has been shown to reveal subtle disruptions in movement patterns prior to

changes in mean and variance. Previous work has postulated that aging and disease are

accompanied by reduction in the complexity of physiological and behavioral control [35],

[36]. This loss of complexity can reduce the body’s ability to adapt to physiological

stress. For example, Vaillancourt et al. [37] examined the hand tremor during a grip force

task in individuals with PD and healthy age-matched controls. They showed that tremor

is less variable in PD than healthy controls and that there was a negative correlation

between variability of tremor and severity of the disease, as measured by the UPDRS

Motor III. This suggests that progression of PD results in decreased variability of motor

output. In light of these findings, it is possible that exercise or movement training, that

emphasizes complex and variable movements, could promote motor improvement in PD.

These results are used in next steps of the research to design and develop an

automated exercise cycle (Smart Bike) that will be used for clinical studies and research

in PD rehabilitation, which is described in next chapter.

31
Chapter 3

Design, Development and Validation of a Smart

Exercise Bike for Rehabilitation in Parkinson’s

Disease

3.1 Introduction

As stated previously in chapters 1 and 2, forced exercise resulted in improvements

of 40% in motor function of both the upper and lower extremity for a group of 5 patients

diagnosed with PD [1, 3]. These results are remarkable and suggest that significant

physiological changes occur as a result of cycling at an accelerated cadence. Despite

these remarkable results, large-scale use of the tandem cycling paradigm for exercise

therapy is not feasible. First, tandem cycling requires an able-bodied trainer to assist in

pedaling that is not reasonable in large-scale clinical deployment or in-home use. Second,

variability in trainer pedaling speed, stamina, and response to the PD rider’s performance

creates variations that make data analysis and conclusions in clinical studies difficult to

generalize. Third, there are a number of factors, such as cadence, foot position and

workload that can affect the biomechanics of cycling. Many motorized single-rider

stationary exercise bikes are commercially available today that can provide a pre-

programmed load profile for the rider. However, it has also been proven difficult to

reproduce the dynamics of the tandem bike riding paradigm using currently available

motorized cycles.

32
 Controller
Trainer Tandem Bike Patient with Motorized Bike Patient
Trainer
Model

Figure 3.1: Duplicating the tandem cycling with a controllable motorized bike.

An ideal solution to this problem would be designing a controllable motorized

bike that could reproduce the tandem cycling for PD patients (Figure 3.1). Such a smart

motorized bicycle could assess individual effort, performance, skill level, and therapeutic

value in order to dynamically alter motor resistance, speed and riding time. The preferred

instrumented cycle would be constructed using a commercially available exercise bike

chassis that is augmented with high-performance motor, sensors, control and automation

equipment that employ an open architecture with components, training and support

services readily available from distributors around the world. Advanced control

techniques and high performance motor-drives can be integrated with a single-rider

stationary bike to permit the rider to experience similar conditions as on a stationary

tandem bike with two riders. The objective of this program is to establish a motorized

bike that can be readily adapted to accommodate different riders and different riding

experiences including the imitation of a tandem bike riding experience. The bike with

embedded controls and servo-motor drive system is termed a Smart (Intelligent) Bike.

The Smart Bike continually monitors the mechanical and electrical characteristics

of the cycle along with the rider’s physical and neuromuscular response during operation.

The captured data may be analyzed to determine which unique temporal features of the

sampled data are correlated with measures of the rider’s motor skill level. The correlation
33
measures established will permit real-time assessment of the performance of the rider

during a cycling session.

In chapter 2, we have studied the complexity of biomechanical and physiological

features of assisted (tandem) and voluntary (single) cycling and related these features to

improvements in motor function as measured by UPDRS Motor III scale. This study

showed that temporal variability or lack of predictability in cadence during active-

assisted cycling is a predictor of improvements in UPDRS Motor III scores [1]. Based on

these findings, one of the key features of the proposed Smart Bike is the capability of

running at higher speed with the appropriate amount of variability in cadence.

The novel motorized bike developed in this work has been instrumented to

include: (a) sensors for feedback control and for high speed sampling of bike data, (b)

sensors that capture real-time rider data, (c) a motor speed controller capable of

dynamically changing motor speed and torque, (d) a programmable controller that

integrates sensor data, communicates with the motor speed controller, provides feedback

control and communicates with the display system, (e) a user interface, termed HMI

(human machine interface) that is capable of displaying the bike and patient data and for

entering the required control parameters, and (f) a data logging and data acquisition

system to support subsequent historical reporting and data analysis.

The objective is that the instrumented exercise bike will be used as a clinical tool

to examine the associations between rider performance and changes in bike control. The

dataset obtained from this study will provide a basis for future development and testing of

customized optimal exercise regimens for individual with PD. Furthermore, this research

platform will permit automatically tailoring an exercise regimen for individuals with

34
different skill levels, disease severity levels and improvement profiles. Exercise programs

may be readily optimized for each individual with minimum effort.

The bike can accurately control the rider’s experience at an accelerated pedal rate

while capturing real-time performance information. Two main control algorithms have

been developed for the bike, static mode (inertia load) and dynamic mode (speed

reference). In the speed reference mode, the bike has the capability to run at a specified

speed while providing the required variability in pedal speed. Speed variability has been

proven to be an important factor in UPDRS improvement based on our previous findings

[1].

3.2 System Architecture; Hardware and Electronics

A commercially available bike chassis from the RECK Company (Motomed

Viva21) served as the mechanical platform for the Smart Bike. The factory-installed

operator display, motor, power supply and controller board were removed from the bike

and were replaced with a high-speed industrial automation control system, rugged touch

screen, high-performance servomotor and motor drive. The servomotor (TLY-A230P-

BJ62AA from Rockwell Automation) is rated at ½ hp and supports a pedaling rate of up

to 120 rpm. It is capable of both driving and absorbing torque from the rider.

Instrumentation permits continuous monitoring of bike operation (e.g. cadence, pedal

torque, and motor current) and rider condition (e.g. heart rate). The programmable logic

controller (PLC) on the bike is a commercial controller from Rockwell Automation

(ControlLogix, 1768-L18ERM) that runs the control algorithm and dynamically operates

1
http://www.motomed.com/en/models/motomed-viva2.html

35
all the systems and components in the bike. Figure 3.2 shows the functional block

diagram of the modified bike.

Figure 3.2: Functional diagram of the Smart Bike.

3.2.1 Hardware and Electronics

As noted previously we were unable to re-use any of the existing MOTOmed

Viva2 electrical components. A matching high-performance drive and low-inertia

servomotor were integrated with the mechanical power components and a commercial

PLC was used to control bike operation and manage a bike riding session. A PanelView

graphical touch-screen display was mounted on the bike and used to capture setup

parameters and to display real-time bike operating data. The following section describes

the electronic systems implemented with this exercise bike.

Most of the electrical/electronic components such as the motor drive,

36
programmable controller (PLC), network adapter, and power supplies are mounted in a

rugged enclosure separate from the bike chassis, and are connected to the bike via cables

for motor power and control, operator interface, and sensor feedback. Other electronic

components integrated within the bike chassis include the operator touch-screen display,

emergency stop button, heart rate monitor interface board, TTL to serial level converter

board for the heart rate monitor, and the motor coupled to the pedal crank.

The overall block diagram and the physical location of the electrical/electronic

components as well as the communication network between different parts of the system

are presented in Figure 3.3 [4].

Figure 3.3: Electrical/electronic component diagram of the bike [4].

The control platform for the intelligent bike is a commercially available

programmable logic controller (PLC). This is a versatile platform that is currently used

across a broad range of automation and robotic applications. The control algorithms run

37
on the programmable controller, and the PLC also communicates with the computer and

bike-mounted operator interface using Ethernet to send and receive commands and data.

The PLC determines the appropriate motor speed and load values and sends motor

control information to the motor drive (Kinetix 350). The motor drive implements a high-

speed inner loop controller that provide the appropriate voltage and current to the motor

to continually maintain the motor operating state specified by the PLC. Motor feedback is

for the drive to maintain proper motor speed and torque in spite of load disturbances

introduced by the rider.

The operator display and control input device is a rugged touch screen device

(PanelView™ Plus graphic display) mounted on the front of the bike near the rider’s

hand grips. This serves as the HMI for the bike and communicates with the PLC through

Ethernet to send the parameters entered by the user to the PLC and to receive and display

the required data from the programmable logic controller. The display can also provide a

graphical plot showing historical values for bike and rider operation.

3.2.2 Heart Rate Monitoring System

It is important to monitor the condition of the rider during the exercise sessions,

and real-time heart rate information can also be used to enhance the safety of the rider

during a bike riding session. Rider heart rate levels and changes in heart rate can signal

excessive rider exertion or indicate potential health problems with the rider. In addition,

heart rate data is also stored in a rider history database for use in subsequent data

analysis.

A Polar Heart Rate monitor from Polar USA (Polar Wearlink+™ Coded

Transmitter) has been selected to monitor the rider heart rate. The Polar Heart Rate

38
monitor includes a wearable chest strap with a battery-operated integral heart rate sensor.

The sensor is connected to a radio transmitter in the chest strap. Typically a wrist-

mounted radio receiver is used to receive the radio signal from the chest strap and display

the heart rate data from the Polar Heart Rate monitor. Because we want to capture the

heart rate in the PLC and monitor real-time heart rate, display this data, and archive the

real-time data values for later analysis we use a separate radio receiver board, a Heart

Rate Monitor Interface Board (HRMI). This board receives the radio signal from the

Polar Heart Rate monitor chest strap and captures the heart rate signals (Part Number

SEN-08661from SparkFun). The output of the HRMI board is a TTL level voltage (0V to

5V) that is converted to a standard ASCII serial communications voltage level using a

level shifting board (MAX3232 Breakout Board from SparkFun). Figure 3.4 shows the

block diagram of the heart rate monitoring system.

Figure 3.4: Block diagram of the heart rate monitoring system [4].

3.2.3 Bike Operational Specifications

The completed bike system is shown in Figure 3.5. Here, we summarize the main

operational specifications of the bike.

Speed: We retained use of the slotted belt pulley on the motor (motor pulley

diameter =0.825”) and the large slotted pulley connected to the pedal crank (crank pulley

39
diameter = 10.9375”). This gives a pulley ratio of 13.2576. The servomotor used is

Rockwell Automation TLY-A230P-BJ62AA with a rated speed of 5,000 rpm. This motor

can operate the pedals at over 300 rpm. At the nominal pedal speed of 95 rpm the motor

speed will be 1260 rpm.

Touchscreen
Display Controller, drive,
and electronics

Integrated
Servo-motor

Motomed Viva2
Bike Chassis

Figure 3.5: Smart exercise bike, completed system [4].

Load: The servomotor used is a 230V motor rated at 1.3 Nm of torque. The motor

is coupled to the pedals through a 13.2576 pulley ratio. This motor provides the torque

necessary to move the pedals without rider assistance and overcome the weight of each

leg. Most of the time the motor will be absorbing load from the pedals and providing a

braking action to increase the pedal resistance felt by the rider. In order to accommodate

prolonged periods of bike operation under pedal loading by the rider, a separate braking

resistor was added to each bike to dissipate the energy provided by the rider.

Safety: Insuring rider safety was of paramount importance in the design and

development of the Smart Exercise Bike. Redundant software controls and integrity

40
checks are implemented to insure the components are connected properly and

communicating reliably. Additional checks are implemented to insure the operator setup

parameters are valid. The control algorithm limits the maximum speed the pedals can

attain. A heart rate monitor attached to the rider will signal excessive fatigue or stress

levels of the rider. There is a large red pushbutton on the bike display console that can

readily be pressed by the rider or therapist and provides an emergency stop (E-stop)

capability that immediately removes power to the motor and other power components.

Additionally, feedback checks are made every millisecond to insure proper motor

feedback and that accurate communications exists between the various system functions.

Lastly, in the event of an overcurrent condition such as due to the pedals hitting an

obstacle, power will immediately be removed from the motor.

Data logging: During each bike riding session data is captured by the PLC and

provided to the graphical display for operator viewing. This data is also routed to a PC

running a data display and logging program (FactoryTalk View from Rockwell

Automation). This program is configured to log captured data to files on the PC hard

drive. Optionally, the PanelView display located on the bike can also log data to a

memory stick or to the display memory for later access.

Reliability: The bike system employs reliable, commercially available drive and

control components that are used in critical applications throughout the world. The design

and implementation of the bike automation system employs good engineering practices to

further insure safe and reliable operation. Over current and over voltage limits protect the

electronic equipment and a cooling fan on the electronics enclosure helps prevent

overheating.

41
3.3 Software, Programs and Control Algorithms

In this section, software and main control algorithms developed to run and control

the bike are described. The control algorithms that operate the bike have been developed

using RSLogix 5000 software from Rockwell Automation running on a PC for software

development. After the control algorithms have been developed, they are then

downloaded to the PLC. Once in the PLC, they are run on this platform to provide real-

time control for the bike system. The control code in the PLC implements both the static

mode (inertia load) and the dynamic mode (speed reference) of bike operation. These two

control modes are described later in this section. New control algorithms can be readily

implemented using the same development and operating platform (i.e. PC, PLC, motor,

and drive) provided with the Smart Exercise Bike.

3.3.1 Development and Programming Software

Software development for the Smart Exercise Bike consists of bike control

software developed to run on the PLC and software resident in the PanelView

touchscreen to accept user input and display bike information and operating data. Other

microprocessor-based devices in the Smart Bike such as the drive and heart rate interface

board were programmed by setting up parameters or command line strings. The suite of

PLC software development tools from Rockwell Automation have been used to develop

the algorithms and routines, and establish communication with the devices to download

the code as well as transmit and display the data. These software programs also provide

for control of the operator interface and data logging. Here is a list of software tools that

have been used to develop the code, establish the communication and transfer the data.

 RSLinx and RSLinx Enterprise: Used for communication with PLC and
42
 PanelView

 RSLogix 5000: Used for developing the algorithms and programming the PLC

 FactoryTalk View Studio Machine Edition(ME): Used to develop the PanelView

program

 ME Transfer Utility: Used to program the PanelView

 FactoryTalk View Studio: Used to log data on the computer

 FactoryTalk Activation Manager: Used for activating the other software

3.3.2 Motor Drive Control Modes

The motor drive (Rockwell Kinetix 350 drive: 2097-V31PR2-LM) includes

different motor control modes. In one mode, position control is the highest level of

dynamic control [4]. As such, position control requires velocity control, and velocity

control requires acceleration control. Acceleration is directly related to torque or force

through the inertia or mass of the load, so acceleration control requires torque control.

Moreover, motor torque is related to motor current through the torque constant of the

motor, and torque control thus requires current control.

The motor drive may be programmed using the following integrated control

modes:

1. No Control Modes

2. Position Control Mode

3. Velocity Control Mode

4. Torque Control Mode

5. Velocity Control Mode with Frequency Control Method

For the Smart Bike control algorithms, the two primary control modes are torque

43
control and velocity control. In torque control mode, the application control program

(ladder logic code) provides torque set-point values to the drive controller via the

Ethernet/IP interface. Because motor current and motor torque are related by a torque

constant, Kt, torque control is analogous to current control. In velocity control mode, the

application control program provides a set-point speed value to the drive Ethernet/IP

interface. Closed-loop velocity control implies an inner torque/current control loop and

implements a field-oriented control or vector-control inner loop generating a PWM

(pulse-width modulation) pulse train to control the motor flux or magnetic field.

Figure 3.6: Screenshot of the drive parameters setup in RSLogix 5000.

44
3.3.3 Static Mode (Inertia Load)

In static mode, the bike works like a typical commercial motorized exercise bike

with the ability to specify a programmable resistance (load) for the rider. In order to

duplicate the inertia load mode of operation, the drive is set in torque control mode and

the control function provides the torque set-point to the drive based on the velocity and

acceleration. In fact, based on the velocity and acceleration, the drive sets a torque

command for the motor so that it resists pedal movement by the rider. The sampling

frequency of the system is 1 kHz, so the control program cycles through the control logic

once a millisecond.

Figure 3.7 shows the block diagram of the closed-loop control system of the bike in

static mode. The lower section of the loop is inside the Motion Control block which is a

toolbox for motor control in RSLogix 5000, and we have access to the parameters of each

block to tune them based on the requirements and specifications of our application. KT is

the motor torque constant which is 0.373 N-m/Amps(RMS). The upper section of the

diagram is implemented with ladder code in the PLC using RSLogix software. Velocity

and acceleration signals are measured and transferred to the programmable logic

controller by motor feedback. The moving average filter applied to velocity and

acceleration signals is an FIR filter with n=50 samples (50 ms time range). KV and KA are

adjustable coefficients that are programmed by the user on the PanelView screen to

determine the amount of load for the rider. Finally, the output of the sum block is passed

through a low pass filter to make the torque command (set point). The low pass filter

transfer function is shown in equation 3.1, and equation 3.2 shows the actual

45
implementation of the torque command for servo drive based on the block diagram in

Figure 3.7.

0.2 z (3.1)
G LPF ( z ) 
z  0.8

Torque_Command = 0.8*Old_Torque_Command +

0.2*(-1/360*(Accel_Factor*Average_Accel+Velocity_Factor*Average_Velocity)) (3.2)

Figure 3.7: Closed loop control block diagram of the Smart Bike in static mode.

3.3.4 Dynamic Mode (Speed Reference)

In dynamic mode, the bike runs at a reference speed set by the user. The pedal

speed (cadence) is increased linearly by the drive to reach the set point value in

approximately 20 seconds. Then, a pre-set resistance (load) is applied as needed to resist

pedal speed changes. For this mode, the speed (cadence) set point is defined by the user

and input through the PanelView, with the drive programmed in velocity control mode. A

torque command function (similar to static mode) is applied to the drive based on

46
velocity and acceleration. In fact, based on the velocity and acceleration, the drive applies

a torque command to the motor so that it resists velocity changes. The Jog block structure

in motion control (RSLogix 5000) is used to increase and decrease the speed gradually

with start and stop commands, in approximately 20 seconds, to control the rate of speed

changes and for stopping the bike.

Figure 3.8 shows the block diagram of the control loop in dynamic mode. The

lower part of the diagram is the velocity control loop that includes a PI controller

implemented inside the Motion Control toolbox in the RSLogix Software. Aside from the

PI control loop there is a feedforward path that includes a low pass filter with adjustable

KVP and KNFF gains. KVP is the proportional gain of the PI controller. KNFF is the

feedforward (negative) gain used to adjust the time response of the velocity regulator

aside from the normal PI control elements [5]. The effect of the (negative) feedforward

signal is to eliminate backup of the motor shaft, and the selection of KNFF gain setting has

no effect on the stability of the speed regulator [5]. KVI is the gain in the integral loop.

Another feature of this velocity control loop is the velocity droop function. The velocity

error input to the integral term is reduced by a fraction of the velocity regulator output, as

controlled by the droop gain setting, KDR. As torque loading on the motor increases,

actual motor speed is reduced in proportion to the droop gain. The output of the velocity

control loop is applied to the drive as the velocity command.

The upper part of the diagram is the torque control loop that has already been

described in the static mode. The only difference between the torque loop in static and

dynamic modes is that in the dynamic mode, the difference between the reference

velocity and feedback velocity is used to shape the torque command equation. The low

47
pass filter used in the torque command loop is presented in equation 3.3, and equation 3.4

displays the actual implementation of the torque command for the servo drive in dynamic

mode based on the block diagram in Figure 3.8.

0.6 z (3.3)
G LPF ( z ) 
z  0.4

Torque_Command = 0.4*Old_Torque_Command + 0.6*(-1/360*(Accel_Factor*Average_Accel

+Velocity_Factor*(Average_Velocity- Velocity_Reference))) (3.4)

Figure 3.8: Closed-loop control block diagram of the Smart Bike in dynamic mode.

48
3.3.5 User interface and data logging

The operator interface code that operates the PanelView touch screen has been

developed using FactoryTalk View Studio Machine Edition (ME) from Rockwell

Automation. The program receives data from the PLC through the Ethernet/IP network

interface and displays real-time information on the screen. The waveform graph displays

the signals in a specified time scale. For this application we set a default of a 30 second

moving window to graphically display data on the operator screen. Data logged by the

program are saved on the hard drive of a PC in Excel format. Bike operating parameters

such as cadence/velocity set-point, velocity and acceleration factors (KA and KV) can be

set by the operator using the touch-screen monitor mounted on the bike.

The operating procedure requires that the user first select an operating mode

(static or dynamic) for the exercise session. The control screen will then appear (Figure

3.9) that includes a composite graph showing cadence, power, torque and heart rate

signals as well as the instantaneous values for each of these signals. There are also

buttons for setting parameters like desired cadence, acceleration and velocity factor for

example.

49
 Figure 3.9: PanelView display and control screen.

3.4 Test and Verifications

An IRB approved (Institutional Review Board approved) study was conducted

that began with a comprehensive test and validation procedure for the bike and data

acquisition system1. Following validation of the bike platform, a series of exercise

sessions with concurrent data acquisition using the Smart Bike was conducted for 47 PD

riders. Details of the test and validation procedure along with the results of the data

analysis are presented later in this chapter. Here, as an example, the experimental test

data from two of the test subjects that participated in the study and exercised in the static

1
Clinical and patient tests were done at Kent State University (School of Health Science) under the
direction of Dr. Angela Ridgel.

50
and dynamic modes of the Smart Bike are presented. Data from each patient were

recorded for forty-minute exercise sessions including five minutes of warm up cycling,

thirty minutes of exercise, and five minutes of cool down.

Figure 3.10(a) shows captured data for heart rate, cadence, and power signals for

one exercise session for a rider in static mode. The session started with five minutes of

warm up, continued with thirty minutes of exercise, and finished with five minutes of

cool down. The power level is smaller during warm up and cool down compared to 30-

minute exercise session as expected. It is worth noting that there is not a significant

difference in the cadences during the three riding phases. Because the load is adjustable,

the trainer conducting this test session chose reduced load during the warm up and cool

down sessions.

Figure 3.10(b) shows experimental data for heart rate, cadence, and power for one

exercise session of a rider in dynamic mode. The session started with five minutes of

warm up during which the cadence set-point is at 50 rpm, then continued with thirty

minutes of exercise at 80 rpm speed set-point, and finished with five minutes of cool

down at 50 rpm cadence set-point. Any change (increase and decrease) in speed is done

linearly in 20 second time periods to avoid any shock or injury to the rider. Fig 3.10(b)

shows that the cadence in the 30-minute exercise period is roughly stable around 80 rpm

while the power in this period is variable from 0 to 60 Watts. These results are very

similar to performance observed with PD riders on a tandem bike with a trainer that we

have previously analyzed (chapter 2). The sample results presented for the PD rider using

dynamic bike control are consistent with the data observed from a PD rider on a tandem

bike. These results suggests that a PD rider operating the Smart Exercise Bike in dynamic

51
mode may see improvements in motor skill levels consistent with that observed from PD

riders operating a tandem bike. The following section provides a detailed analysis of the

data captured from multiple cycling sessions under conditions of static and dynamic

cycling that seeks to validate this assumption.

Heart Rate
Heart Rate (bpm)

100

50

0
0 500 1000 1500 2000
Time (Sec)
Cadence
100
Cadence (rpm)

50

0
0 500 1000 1500 2000
Time (Sec)
Power
100
Power (Watts)

50

0
0 500 1000 1500 2000
Time (Sec)

(a)

52
 Heart Rate

Heart Rate (bpm)

100

50

0
0 500 1000 1500 2000
Time (Sec)
Cadence
100
Cadence (rpm)

50

0
0 500 1000 1500 2000
Time (sec)
Power
80
60
Power (Watts)

40
20
0
-20
0 500 1000 1500 2000
Time (sec)

(b)
Figure 3.10: Smart Bike real test data for two PD patients; (a) Static mode; (b) Dynamic mode.

Table 3.1 summarizes the main characteristics of the recorded signals from the

two PD patient bike riding experiments in static and dynamic modes. There is no

significant difference in heart rate signals between the two modes. However, the pedaling

cadence and power signals showed a significant difference between the raw values in

static (68.3 ± 7.1 rpm, 42.5 ± 9.7 W) and dynamic (82.4 ± 1.5 rpm, 26.9 ± 12.5 W)

groups. Certainly no conclusions can be drawn from looking at only two samples,

however, the trend shown here is consistent with data captured from the other 45 test

subjects. Variability analysis has shown to indicate another significant difference

between the two modes. While there is no significant difference between the Sample

53
Entropy1 (SaEn) values of heart rate and power signals, there is a significant difference in

SaEn of the cadence for static and dynamic modes. SaEn for the cadence in the dynamic

mode session (1.47) is significantly greater than the SaEn for the cadence in the static

mode (0.26). This indicates that the cadence signals in the dynamic mode have greater

variability (are less predictable) than the signals in static mode. This feature is consistent

with the analysis results presented in Chapter 2 and previously published for single and

tandem cycling [1] and is an important feature of the Smart Bike design. Complete test

details, including validation and data analysis of the Smart Bike with forty seven PD

patients is presented next.

Table 3.1: Comparison between the static and dynamic modes with real data signals of two PD
patients exercised with smart bike.
Bike Heart Rate Cadence Power
Mode Mean SaEn Mean SaEn Mean SaEn
Static 96.3 0.31 68.3 0.26 42.5 0.11
Dynamic 92.8 0.28 82.4 1.47 26.9 0.06

3.5 Data Analysis and Smart Bike Validation with PD Patients

In previous sections of this chapter, the details of the design and development of

the Smart Exercise Bike including hardware, electronics, software, and control

algorithms have been presented. The bike can accurately control the rider’s experience at

an accelerated pedal rate while capturing real-time performance data. Two main control

algorithms have been developed for the bike, static mode (inertia load) and dynamic

mode (speed reference). In the speed reference mode, the bike has the capability to run at

a specified speed while providing the required variability in pedal speed. Speed

1
Sample entropy (SaEn) is a technique used to quantify the amount of regularity or unpredictability of
fluctuations in time series data and was discussed in chapter 2.

54
variability has been shown to be an important factor in UPDRS improvement based on

the results and findings in Chapter 2.

In order to test the effects of exercise with these two types of control methods,

forty-seven individuals with PD were recruited in an IRB-approved study at Kent State

University (School of Health Science) under the direction of Dr. Angela Ridgel.

Individuals selected for the study were randomly assigned to either the static or dynamic

cycling group. Each test subject completed three 40-minute exercise sessions every other

day over a period of one week. Heart rate, cadence and power data were captured and

recorded during each exercise session. The UPDRS Motor III test was administered in

double-blind tests by a trained neurologist. This test was administered to each patient

before and after the three exercise sessions in order to evaluate the effect of the exercise

regimen.

To distinguish the static and dynamic cycling groups by biomechanical and

physiological features, we have studied the complexity of the recorded signals (power,

heart rate, and cadence) using variability analysis techniques such as Approximate

Entropy (ApEn), Sample Entropy (SaEn) and Spectral Entropy (SpEn) which were

described in Chapter 2. Results from these computations were used in a multiple linear

regression (MLR) model to correlate these features to improvements in motor function as

measured by the UPDRS Motor III scale.

The variability analysis results are consistent with the results we obtained

previously in chapter 2 describing the changes in motor function observed in PD riders

following voluntary (single) and forced (tandem) cycling experiments [1].

55
3.5.1 Materials and Methods

To test and validate the effectiveness of Smart Bike, forty-seven individuals

diagnosed with Parkinson’s disease completed three cycling sessions over a one week

period riding smart bikes, and were evaluated for changes in motor function after the

exercise sessions were concluded1 (Figure 3.11). All individuals were randomly assigned

to use either the static or dynamic control mode during bike operation and were able to

successfully complete three 40-minute cycling sessions. Demographic data was analyzed

using independent sample t-tests and there were no significant differences between the

static and dynamic cycling groups, demonstrating an acceptable level of homogeneity

across the groups (Table 3.2).

Figure 3.11: Flow diagram for recruitment and assessment of individuals with PD

1
Clinical and patient tests were done at Kent State University (School of Health Science) under the
direction of Dr. Angela Ridgel.

56
 Table 3.2: Subject Demographics
Static (n=23) Dynamic (n=24) p-value
Age (years) 67.26± 0.97 67.17 ± 1.66 0.962
Hoehn and Yahr Scale 1.83 ± 0.14 2.13 ± 0.16 0.151
Height (inch) 67.71 ± 0.74 68.15 ± 0.76 0.681
Weight (lbs) 165.17 ± 6.0 175.08 ± 8.14 0.336
BMI 25.08 ± 0.71 26.64 ± 0.91 0.186
PD duration (months) 77.74 ± 9.73 83.46 ± 11.17 0.702
Levodopa Equivalent 153.32 ± 23.9 178.80 ± 29.4 0.507
Dose

Table 3.3: Exercise Parameters

Static (n=23) Dynamic (n=24) p-value
Cadence (rpm) 66.00 ± 3.23 78.63 ± 1.13 0.000
Power (watts) 31.17 ± 4.09 8.79 ± 4.33 0.000
Torque (N-m) 0.38 ± 0.038 0.106 ± 0.052 0.000
Heart Rate (bpm) 103.26 ± 3.12 91.08 ± 2.56 0.004
RPE (6-20 scale) 13.65 ± 0.40 12.71 ± 1.06 0.417

Pedaling cadence and power exerted by the patients in each group was measured

and recorded by the programmable logic controller (PLC) that also controlled the bike

operation and display screen. A wireless chest-worn heart rate monitor (Polar Electro,

Lake Success, NY) transmitted rider heart rate data to the PLC during bike operation for

subsequent analysis. In order to examine the captured raw data for the static and dynamic

control modes, the mean and standard deviation of the “heart rate”, “power” and

“cadence” signals for all tests were calculated (Table 3.4). Dynamic and static cycling

modes resulted in similar subject self-assessments of RPE (Rating of Perceived Exertion).

However, there were significant differences in cadence, power, torque and heart rate

observed (Table 3.3). Interestingly, individuals in the static cycling group showed a lower

cadence but higher power, torque and heart rate than riders in the dynamic cycling group.

57
Motor Function Assessment

The Unified Parkinson’s Disease Rating Scale (UPDRS) Motor III exam was

administered to each participant by a blinded neurologist (Dr. B. Walter). Assessments

were performed just prior to the start of the first exercise session and two days after

completion of the third exercise session (Table 3.4). The difference in UPDRS Motor III

scores between these two time periods was calculated and used in later analysis. A

negative change in score represents improvement while a positive change indicates

worsening of motor symptoms.

58
 Table 3.4: Mean and variance for power, heart rate and cadence signals
Patient Group UPDRS Heart Rate Power Cadence
ID Pre Post Changea Meanb Stdev Meanb Stdev Meanb Stdev
SMB0 Static 27 29 2 * * 15.5 4.77 51.9 6.42
SMB0
1 Static 28 28 0 * * 10.7 2.16 73.1 9.45
2
SMB0 Static 29 41 12 * * 16.0 2.54 66.4 5.19
3
SMB0 Static 30 30 0 121.3 9.6 17.3 4.17 60.1 7.11
4
SMB0 Static 17 22 5 130.3 7.7 61.6 11.53 77.5 4.16
6
SMB0 Static 28 36 8 87.4 4.7 18.5 4.08 74.4 6.38
7
SMB0 Static 26 21 -5 105.4 22.3 31.5 5.75 80.3 4.57
9
SMB1 Static 41 30 -11 * * 27.6 4.80 69.9 6.06
0
SMB1 Static 14 18 4 123.6 9.9 25.0 2.93 79.1 4.94
2
SMB1 Static 38 35 -3 * * 7.70 2.56 50.6 6.29
3
SMB1 Static 18 26 8 60.0 0.0 48.0 10.40 81.1 4.41
7
SMB2 Static 18 20 2 93.6 4.7 39.9 13.21 65.2 6.99
0
SMB2 Static 35 38 3 98.3 6.8 42.5 8.60 80.6 3.89
5
SMB2 Static 41 45 4 90.7 4.8 30.7 6.50 43.9 7.03
6
SMB2 Static 24 19 -5 119.2 6.4 22.1 6.92 70.0 10.4
8
SMB2 Static 14 17 3 84.7 2.7 42.5 8.18 48.6 7.38
9
SMB3 Static 27 22 -5 132.6 7.4 19.6 3.73 69.2 5.43
1
SMB3 Static 27 24 -3 88.1 2.1 10.2 4.13 43.8 9.86
4
SMB3 Static 16 20 4 92.2 1.8 43.1 12.74 72.1 8.36
5
SMB3 Static 34 25 -9 112.7 9.9 48.5 9.53 78.4 6.32
6
SMB3 Static 22 22 0 102.0 5.9 45.2 9.38 73.0 5.61
8
SMB4 Static 11 5 -6 110.7 8.8 70.4 14.85 82.7 3.45
0
SMB4 Static 15 16 1 128.5 5.9 34.2 3.66 79.8 3.96
3
MEAN 25.2 25.6 +0.4 104.5 6.7 31.7 6.83 68.3 6.2
SMB0 Dynamic 15 21 6 89.0 1.2 18.1 10.46 81.4 1.41
5
SMB0 Dynamic 24 27 3 119.0 21.3 20.1 9.05 79 3.73
8
SMB1 Dynamic 47 36 -11 * * -17.7 9.73 72 2.89
1
SMB1 Dynamic 22 18 -4 112.4 12.1 54.1 10.02 88.1 1.47
4
SMB1 Dynamic 35 43 8 80.1 2.6 -10.2 8.05 74.8 2.58
5
SMB1 Dynamic 37 22 -15 115.8 7.1 26.1 9.51 84.6 1.97
6
SMB1 Dynamic 20 11 -9 83.8 5.8 23.4 18.44 80.6 1.82
8
SMB1 Dynamic 54 57 3 91.1 3.4 -16.1 7.23 71.6 6.38
9
SMB2 Dynamic 30 28 -2 88.5 3.8 30.8 9.34 82.3 3.18
1
SMB2 Dynamic 14 9 -5 105.5 16.5 45.7 13.42 84.6 1.41
2
SMB2 Dynamic 35 36 1 80.3 1.9 -14.6 6.97 73.4 2.05
3
SMB2 Dynamic 23 15 -8 89.7 3.1 20.0 10.45 81.6 1.86
4
SMB2 Dynamic 18 9 -9 90.4 3.1 24.3 14.13 81.7 1.42
7
SMB3 Dynamic 19 19 0 108.2 3.9 4.10 14.15 78.6 3.59
0
SMB3 Dynamic 48 44 -4 84.2 1.3 -17.50 4.70 71.3 1.98
2
SMB3 Dynamic 29 29 0 94.6 1.3 0.40 6.95 78.8 1.15
3
SMB3 Dynamic 27 20 -7 71.7 2.9 -6.90 5.50 76.6 1.38
7
SMB3 Dynamic 34 34 0 110.4 20.0 26.30 6.47 79.8 4.57
9
SMB4 Dynamic 22 16 -6 100.1 4.8 8.00 10.96 78.9 4.30
1
SMB4 Dynamic 21 15 -6 86.5 3.6 3.00 8.77 79.1 1.43
2
SMB4 Dynamic 52 27 -25 96.4 4.2 -6.80 14.69 72.3 5.51
4
SMB4 Dynamic 55 47 -8 73.3 1.93 -24.50 6.65 72.9 4.28
5
SMB4 Dynamic 4 1 -3 111.3 21.0 -9.20 7.98 75.1 4.51
6
SMB4 Dynamic 45 45 0 110.4 2.86 4 18.10 6.28 79.8 1.82
7
MEAN 30.4 26.2 -4.21 95.3 6.52 8.30 9.58 78.3 2.8
a
Negative change in UPDRS represents improvements in motor function
b
Mean values were calculated over 3 exercise sessions per patient
*Due to heart rate sensor failure, the data was inaccurate

59
3.5.2 Biomechanical and Physiological Feature Analysis

To study the difference between the static and dynamic cycling groups, the

complexity of the recorded signals (power, heart rate, and cadence) have been studied

using the variability analysis techniques; ApEn, SaEn and SpEn which were introduced

earlier in chapter 2. Both ApEn and SaEn quantify the predictability (or regularity) in a

time series, and are useful in quantifying differences in health and disease [6, 7]. The

values for ApEn and SaEn computed for each dataset are nearly the same, so further

analysis focused on the SaEn computations. SpEn is another variability measure that

quantifies the distribution of frequency content in a time series. The interpretation of

SpEn (in the frequency domain) is very similar to ApEn and SaEn (in the time domain).

ApEn, SaEn and SpEn for “heart rate”, “power” and “cadence” signals were

computed in all data sets for each person. The computed values were used to quantify the

variability of both temporal and frequency domain patterns. The mean of these

parameters for three exercise sessions were computed for the static and dynamic groups

(Table 3.5) and have been used to distinguish between the two groups. These data were

also used in a multiple linear regression (MLR) model to connect these features to

improvements in motor function as measured by the UPDRS Motor III scale.

The variability analysis results are consistent with the results we previously

presented in chapter 2 describing the results from voluntary (single) and assisted

(tandem) cycling by riders with PD [1].

60
Table 3.5: Mean of ApEn and SpEn for power, heart rate and cadence signals over three sessions
Patient Group UPDRS Heart Rate Power Cadence
ID Changea SaEn SpEn SaEn SpEn SaEn SpEn
SMB01 Static 2 * * 0.51 0.09 0.33 0.08
SMB02 Static 0 * * 0.72 0.08 0.04 0.08
SMB03 Static 12 * * 1.00 0.08 0.51 0.07
SMB04 Static 0 0.0 0.1 0.45 0.09 0.24 0.07
SMB06 Static 5 0.1 0.1 0.05 0.07 0.72 0.07
SMB07 Static 8 0.0 0.1 0.23 0.08 0.17 0.07
SMB09 Static -5 0.0 0.1 0.29 0.07 0.54 0.07
SMB10 Static -11 * * 0.54 0.08 0.49 0.07
SMB12 Static 4 0.08 0.07 1.00 0.08 0.58 0.07
SMB13 Static -3 1.15 0.07 0.89 0.10 0.38 0.08
SMB17 Static 8 * * 0.08 0.07 0.59 0.07
SMB20 Static 2 0.18 0.07 0.07 0.09 0.29 0.07
SMB25 Static 3 0.02 0.07 0.12 0.07 0.74 0.07
SMB26 Static 4 0.62 0.07 0.96 0.15 0.43 0.10
SMB28 Static -5 0.05 0.07 0.12 0.08 0.10 0.07
SMB29 Static 3 0.35 0.07 0.86 0.15 0.26 0.11
SMB31 Static -5 0.02 0.07 0.67 0.08 0.54 0.07
SMB34 Static -3 0.95 0.07 0.64 0.13 0.14 0.10
SMB35 Static 4 0.01 0.07 0.26 0.08 0.13 0.07
SMB36 Static -9 0.01 0.07 0.06 0.08 0.29 0.07
SMB38 Static 0 0.11 0.07 0.08 0.08 0.40 0.07
SMB40 Static -6 0.03 0.07 0.02 0.07 0.81 0.07
SMB43 Static 1 0.02 0.07 0.71 0.07 0.73 0.07
MEAN +0.4 0 0.07 0.45 0.09 0.42 0.08
SMB05 Dynamic 6 0.78. 0.07 0.14 0.28 1.55 0.07
SMB08 Dynamic 3 0.012 0.07 0.07 0.31 0.66 0.07
SMB11 Dynamic -11 *1 * 0.09 0.13 0.66 0.07
SMB14 Dynamic -4 0.511 0.07 0.11 0.09 1.40 0.07
SMB15 Dynamic 8 0.66 0.07 0.24 0.18 0.96 0.07
SMB16 Dynamic -15 0.02 0.07 0.22 0.17 1.02 0.07
SMB18 Dynamic -9 0.57 0.07 0.10 0.32 1.38 0.07
SMB19 Dynamic 3 0.94 0.07 0.31 0.12 0.99 0.07
SMB21 Dynamic -2 0.26 0.07 0.17 0.12 1.09 0.07
SMB22 Dynamic -5 0.04 0.07 0.03 0.10 1.08 0.07
SMB23 Dynamic 1 1.21 0.07 0.33 0.12 1.20 0.07
SMB24 Dynamic -8 0.58 0.07 0.12 0.26 1.16 0.07
SMB27 Dynamic -9 0.44 0.07 0.06 0.21 1.33 0.07
SMB30 Dynamic 0 0.10 0.07 0.07 0.36 0.60 0.07
SMB32 Dynamic -4 1.01 0.07 0.52 0.08 1.13 0.07
SMB33 Dynamic 0 0.58 0.07 0.34 0.29 1.92 0.07
SMB37 Dynamic -7 0.38 0.07 0.39 0.13 1.32 0.07
SMB39 Dynamic 0 0.38 0.07 0.34 0.12 0.86 0.07
SMB41 Dynamic -6 0.08 0.07 0.10 0.42 0.63 0.07
SMB42 Dynamic -6 0.86 0.07 0.19 0.28 1.55 0.07
SMB44 Dynamic -25 0.07 0.07 0.24 0.11 1.08 0.07
SMB45 Dynamic -8 1.10 0.07 0.22 0.08 0.92 0.07
SMB46 Dynamic -3 0.20 0.07 0.33 0.08 1.14 0.07
SMB47 Dynamic 0 0.50 0.07 0.57 0.34 1.52 0.07
MEAN -4.21 0.49 0.07 0.22 0.20 1.13 0.07
*Due to heart rate sensor failure, the data was inaccurate

61
3.5.3 Data Analysis Results

UPDRS Motor III assessment shows a significant difference between the static

and dynamic groups (Table 3.4). The average of UPDRS change in the static group is

+0.4 (1.6 % worsening) compared to the dynamic group with an average UPDRS change

of -4.21 (13.84 % improvement). There were slight differences (PHeart-Rate=0.441,

PPower=0.289) in heart rate and power signals between the two groups (Table 3.4).

However, the pedaling cadence showed a significant difference between the raw values in

the static and dynamic groups. Cadence for the dynamic group (78.3± 2.8 rpm) was

higher than the static group (68.3 ± 6.2 rpm) with less variability as quantified by the

standard deviation. Power for the dynamic group (8.3 ± 9.6 W) was lower than the static

group (31.7 ± 6.8 W).

Variability analysis reveals the hidden differences in signals between the two

groups (Table 3.5). Comparison of each variable shows clear differences. Sample

Entropy (SaEn) for the cadence in the dynamic group (1.13) is significantly greater and

different than the SaEn for the cadence signal in the static group (0.42). This suggests

that the cadence signals in the dynamic group have greater variability (are less

predictable) than the cadence signals in static group. The results for the power signal are

opposite; that is, the power signals for the dynamic group (SaEn=0.22) show less

variability (are more predictable) and are significantly different from the power signals of

the static group (SaEn=0.45).

Only the Spectral Entropies (SpEn) of the power signals are distinguishable

between the static and dynamic groups. SpEn of the power in the dynamic group (0.2)

showed more variability as compared to the static group (SpEn=0.10).

62
 The data in Table 3.4 and Table 3.5 have been used to develop a multiple linear

regression (MLR) model and to compute the odds ratio for achieving a positive change in

the UPDRS Motor III scores in each group using logistic regression. Based on the

distinguishable parameters in Table 3.4 and Table 3.5 we selected six independent

variables for the MLR model. The selected parameters are: Mean, StDv, and SaEn of

cadence, and Mean, SaEn, and SpEn of power. The dependent variable of the MLR

model is the change in UPDRS score. The MLR model development has been done in

two ways: first two separate MLR models were built, one model for the static group

(Table 3.6) and one model for the dynamic group (Table 3.7), and then the data were

combined into a single dataset, and another MLR model was built (Table 3.8). The

residual values and the predicted UPDRS scores in Tables 3.6 and 3.7 were obtained

from MLR modeling using the “LinearModel” function in MATLAB. Two different

models were built for each group of data: linear MLR model (model 1), and linear MLR

with interactions1 (model 2). Figures 3.12 and 3.13 show the correlation between the real

and predicted UPDRS changes for three MLR models (static group, dynamic group, and

combined static and dynamic groups). The linear model shows a small correlation

between the real and predicted data for all three cases (Figure 3.12). However, the linear

model with interactions shows positive and significant correlation between real and

predicted UPDRS change (Figure 3.13).

Tables 3.6, 3.7 and 3.8 also contain the values of logistic regression and odds ratio

computations for the two groups with different models. Here, the logistic regression

shows the probability of improvement in motor function as quantified by a negative

1
MLR model with interactions includes the main variables and two-way interaction terms.

63
change in UPDRS score. The higher values of logistic regression and odds ratio for the

dynamic group in each of the tables shows that the MLR model predicts a greater

UPDRS improvement in the dynamic group, which is consistent with the real UPDRS

data from the study.

Correlation of Real and Predicted UPDRS Change for Static Group

15

10

5
Predicted

-5

-10

-15
-15 -10 -5 0 5 10 15
Real

(a)
Correlation of Real and Predicted UPDRS Change for Dynamic Group
25

20

15
Predicted

10

-5
-5 0 5 10 15 20 25
Real

(b)

64
 Correlation of Real and Predicted UPDRS Change for Combined Static and Dynamic Groups
25

20

15

10
Predicted

-5

-10

-15

-15 -10 -5 0 5 10 15 20 25
Real

(c)
Figure 3.12: Correlation analysis of real (measured) versus predicted UPDRS values for linear
MLR model. (a): static group. (b): Dynamic group (active assisted exercise) (c): combined static
and dynamic data. There was a positive, but small correlation of the real and predicted UPDRS
Motor III scores in both the static (r=0.39) and dynamic (r=0.30) groups. Combined scores also
maintained a positive, but not significant correlation(r=0.39) between real and predicted scores.

Correlation of Real and Predicted UPDRS (Static Group, Interactions Model)

15

10

5
Predicted UPDRS

-5

-10

-15
-15 -10 -5 0 5 10 15
Real UPDRS

(a)

65
 Correlation of Real and Predicted UPDRS (Dynamic Group, Interactions Model)
25

20

15

Predicted UPDRS
10

-5

-10
-10 -5 0 5 10 15 20 25
Real UPDRS

(b)
Correlation of Real and Predicted UPDRS (Combined Groups, Interactions Model)
25

20

15

10
Predicted UPDRS

-5

-10

-15
-15 -10 -5 0 5 10 15 20 25
Real UPDRS

(c)
Figure 3.13: Correlation analysis of real (measured) versus predicted UPDRS values for linear
MLR model, (a): static group (voluntary exercise), (b): Dynamic group (active-assisted exercise),
(c): combined static and dynamic data. There was a positive, correlation of the real and predicted
UPDRS Motor III scores in both the static (r=0.81) and dynamic groups (r=0.87). Combined
scores also maintained a positive and significant correlation between real and predicted scores
(r=0.63).

66
 Table 3.6: Regression analysis results for the static group
Patient UPDRS Change Residuals P (Logistic ratio) Odds Ratio (P/(1-P))
ID Predicted
Real Model 1 Model 2 Model 1 Model 2 Model 1 Model 2
Model 1 Model 2
SMB01 2 1.3 0.4 0.7 1.6 0.22 0.40 0.3 0.7
SMB02 0 -0.2 0.3 0.2 -0.3 0.55 0.42 1.2 0.7
SMB03 12 3.2 11.2 8.8 0.8 0.04 0.00 0.0 0.0
SMB04 0 0.9 1.9 -0.9 -1.9 0.28 0.13 0.4 0.1
SMB06 5 -0.9 3.4 5.9 1.6 0.71 0.03 2.5 0.0
SMB07 8 2.6 5.9 5.4 2.1 0.07 0.00 0.1 0.0
SMB09 -5 0.8 2.4 -5.8 -7.4 0.31 0.09 0.5 0.1
SMB10 -11 -0.5 -10.1 -10.5 -0.9 0.63 1.00 1.7 23349.0
SMB12 4 2.7 2.4 1.3 1.6 0.06 0.08 0.1 0.1
SMB13 -3 2.2 -2.4 -5.2 -0.6 0.10 0.92 0.1 11.5
SMB17 8 0.2 -3.7 7.8 11.7 0.46 0.98 0.9 42.0
SMB20 2 -0.5 2.3 2.5 -0.3 0.62 0.09 1.6 0.1
SMB25 3 -1.7 3.7 4.7 -0.7 0.85 0.02 5.7 0.0
SMB26 4 1.1 4.2 2.9 -0.2 0.24 0.01 0.3 0.0
SMB28 -5 -6.4 -5.4 1.4 0.4 1.00 1.00 593.6 232.1
SMB29 3 3.8 2.9 -0.8 0.1 0.02 0.05 0.0 0.1
SMB31 -5 0.2 -5.0 -5.2 0.0 0.44 0.99 0.8 152.5
SMB34 -3 -3.3 -3.1 0.3 0.1 0.96 0.96 26.1 22.5
SMB35 4 0.5 2.7 3.5 1.3 0.38 0.06 0.6 0.1
SMB36 -9 1.4 -5.0 -10.4 -4.0 0.19 0.99 0.2 150.4
SMB38 0 1.0 1.3 -1.0 -1.3 0.26 0.21 0.4 0.3
SMB40 -6 -0.6 -2.7 -5.4 -3.3 0.65 0.94 1.9 14.9
SMB43 1 1.1 1.5 -0.1 -0.5 0.25 0.18 0.3 0.2
Table 3.7: Regression analysis results for the dynamic group
Patient UPDRS Change Residuals P (Logistic ratio) Odds Ratio (P/(1-P))
ID Predicted
Real Model 1 Model 2 Model 1 Model 2 Model 1 Model 2 Model 1 Model 2
SMB05 6 -4.1 -5.1 10.1 11.1 0.98 0.99 61.3 167.3
SMB08 3 -3.4 1.2 6.4 1.8 0.97 0.23 29.2 0.3
SMB11 -11 -6.3 -10.7 -4.7 -0.3 1.00 1.00 571.0 44287.0
SMB14 -4 -6.5 -2.7 2.5 -1.3 1.00 0.94 674.9 14.7
SMB15 8 -4.1 9.2 12.1 -1.2 0.98 0.00 59.5 0.0
SMB16 -15 -3.9 -15.6 -11.1 0.6 0.98 1.00 47.7 6.11E+6
SMB18 -9 -3.6 -6.8 -5.4 -2.2 0.97 1.00 35.1 887.4
SMB19 3 -6.0 0.2 9.0 2.8 1.00 0.46 402.4 0.8
SMB21 -2 -5.7 -5.1 3.7 3.1 1.00 0.99 304.7 169.7
SMB22 -5 -6.6 -6.4 1.6 1.4 1.00 1.00 772.5 580.5
SMB23 1 -4.2 -4.7 5.2 5.7 0.99 0.99 67.0 109.3
SMB24 -8 -4.0 -3.1 -4.0 -4.9 0.98 0.96 53.1 22.1
SMB27 -9 -5.6 -3.8 -3.4 -5.2 1.00 0.98 267.8 43.3
SMB30 0 -2.9 -2.9 2.9 2.9 0.95 0.95 18.1 18.4
SMB32 -4 -2.1 -2.5 -1.9 -1.5 0.89 0.92 8.2 12.2
SMB33 0 -2.5 2.4 2.5 -2.4 0.92 0.08 11.9 0.1
SMB37 -7 -3.2 -6.6 -3.8 -0.4 0.96 1.00 24.5 726.3
SMB39 0 -3.6 1.5 3.6 -1.5 0.97 0.18 36.7 0.2
SMB41 -6 -1.8 -3.3 -4.2 -2.7 0.86 0.97 6.3 28.5
SMB42 -6 -3.8 -4.5 -2.2 -1.5 0.98 0.99 45.4 94.3
SMB44 -25 -6.6 -22.7 -18.4 -2.3 1.00 1.00 725.4 7.14E+9
SMB45 -8 -7.0 -6.3 -1.0 -1.7 1.00 1.00 1131.7 518.9
SMB46 -3 -5.9 -2.4 2.9 -0.6 1.00 0.91 347.7 10.5
SMB47 0 2.5 -0.4 -2.5 0.4 0.08 0.59 0.1 1.4

67
 Table 3.8: Combined static and dynamic group data regression analysis results
Patient UPDRS Change Residuals P (Logistic ratio) Odds Ratio (P/(1-P))
ID Predicted
Real Model 1 Model 2 Model 1 Model 2 Model 1 Model 2
Model 1 Model 2
SMB01 2 -0.2 -0.4 2.2 2.4 0.54 0.60 1.2 1.5
SMB02 0 0.5 2.1 -0.5 -2.1 0.37 0.11 0.6 0.1
SMB03 12 3.0 5.1 9.0 6.9 0.05 0.01 0.1 0.0
SMB04 0 -0.5 -0.3 0.5 0.3 0.63 0.57 1.7 1.3
SMB06 5 -1.4 -0.4 6.4 5.4 0.80 0.60 4.0 1.5
SMB07 8 -1.3 6.0 9.3 2.0 0.78 0.00 3.6 0.0
SMB09 -5 -0.8 0.7 -4.2 -5.7 0.70 0.34 2.3 0.5
SMB10 -11 0.1 -5.9 -11.1 -5.1 0.47 1.00 0.9 376.4
SMB12 4 3.3 5.3 0.7 -1.3 0.03 0.00 0.0 0.0
SMB13 -3 1.9 0.7 -4.9 -3.7 0.13 0.33 0.2 0.5
SMB17 8 -1.5 -0.9 9.5 8.9 0.82 0.71 4.4 2.5
SMB20 2 -2.0 4.5 4.0 -2.5 0.88 0.01 7.5 0.0
SMB25 3 -1.9 -2.1 4.9 5.1 0.87 0.89 6.7 8.0
SMB26 4 3.4 1.9 0.6 2.1 0.03 0.13 0.0 0.1
SMB28 -5 -3.8 -3.9 -1.2 -1.1 0.98 0.98 46.8 48.7
SMB29 3 3.9 6.1 -0.9 -3.1 0.02 0.00 0.0 0.0
SMB31 -5 0.7 -0.8 -5.7 -4.2 0.33 0.69 0.5 2.2
SMB34 -3 -0.4 -4.0 -2.6 1.0 0.60 0.98 1.5 56.6
SMB35 4 -0.6 -6.3 4.6 10.3 0.65 1.00 1.9 549.3
SMB36 -9 -1.3 0.5 -7.7 -9.5 0.78 0.38 3.6 0.6
SMB38 0 -1.4 1.9 1.4 -1.9 0.81 0.13 4.2 0.1
SMB40 -6 -1.1 0.0 -4.9 -6.0 0.75 0.51 3.0 1.0
SMB43 1 1.7 -1.2 -0.7 2.2 0.16 0.77 0.2 3.3

SMB05 6 -4.4 -5.7 10.4 11.7 0.99 1.00 78.9 313.1

SMB08 3 -1.8 -2.2 4.8 5.2 0.86 0.90 6.1 9.2
SMB11 -11 -4.6 -6.6 -6.4 -4.4 0.99 1.00 95.4 711.8
SMB14 -4 -3.3 -2.4 -0.7 -1.6 0.96 0.92 26.1 11.4
SMB15 8 -3.9 2.1 11.9 5.9 0.98 0.11 51.7 0.1
SMB16 -15 -2.1 -8.2 -12.9 -6.8 0.89 1.00 7.8 3826.4
SMB18 -9 -3.5 -2.5 -5.5 -6.5 0.97 0.92 34.0 12.1
SMB19 3 -6.5 -4.5 9.5 7.5 1.00 0.99 698.6 88.8
SMB21 -2 -3.5 -4.5 1.5 2.5 0.97 0.99 33.4 86.9
SMB22 -5 -2.7 -6.5 -2.3 1.5 0.94 1.00 15.1 684.7
SMB23 1 -4.8 -7.6 5.8 8.6 0.99 1.00 119.9 1954.8
SMB24 -8 -3.0 -7.6 -5.0 -0.4 0.95 1.00 20.5 2089.8
SMB27 -9 -4.1 -8.9 -4.9 -0.1 0.98 1.00 58.5 7410.0
SMB30 0 -2.0 0.1 2.0 -0.1 0.88 0.49 7.3 0.9
SMB32 -4 -3.6 -0.7 -0.4 -3.3 0.97 0.68 35.4 2.1
SMB33 0 -5.4 1.8 5.4 -1.8 1.00 0.15 232.1 0.2
SMB37 -7 -4.0 -1.7 -3.0 -5.3 0.98 0.85 56.6 5.6
SMB39 0 -2.3 -1.4 2.3 1.4 0.91 0.80 10.4 3.9
SMB41 -6 -1.7 -3.8 -4.3 -2.2 0.84 0.98 5.4 43.9
SMB42 -6 -4.9 -1.7 -1.1 -4.3 0.99 0.84 131.2 5.3
SMB44 -25 -6.5 -11.3 -18.5 -13.7 1.00 1.00 659.6 84882.0
SMB45 -8 -6.4 -10.6 -1.6 2.6 1.00 1.00 593.3 40443.0
SMB46 -3 -5.9 -4.7 2.9 1.7 1.00 0.99 370.5 111.9
SMB47 0 -1.1 -1.3 1.1 1.3 0.75 0.79 3.0 3.8

68
3.6 Conclusion

Recent research results in the field of rehabilitation for people diagnosed with

Parkinson’s disease indicate that a person may realize significant improvements in motor

skills by pedaling a bike under unique conditions of speed and load dynamics. In order to

investigate this important research area, a novel exercise bike has been designed and

fabricated based the operating paradigm of a tandem bike. The framework for the smart

single-rider exercise bike is a commercial bike chassis that has been augmented with high

performance commercially available motor and control equipment. This novel design

incorporates high-performance drives and controls and a low-inertia, power-dense

servomotor to form a flexible and adaptive platform to support clinical research studies of

exercise for people with Parkinson’s disease.

The Smart Exercise Bike has been programmed with two resident bike control

algorithms that provide the ability to operate the bike in either the static (inertial load)

mode, or dynamic (speed reference) mode while capturing operating data such as rider

heart rate, cadence, and power at a high sampling rate. The static mode operates the bike

as a regular exercise bike with a programmable resistance (load). In dynamic mode, the

bike operates at a user defined cadence set-point with a programmable load influencing

cadence changes. The bike is equipped with a user friendly HMI employing an easy to

read color touch screen. This integrated control and display system records critical rider

and bike conditions and allows the rider to set required riding session parameters such as

cadence set point and load.

The Smart Bike has been used since September 2012 to conduct clinical trials of

riders with Parkinson’s disease. Data has been captured during multiple riding sessions

69
and analyzed to confirm the proper operation of the bike and the validity of the data

acquisition system. The Smart Bike has been shown to be an effective platform for

conducting a wide range of bike riding exercise tests for different riders with Parkinson’s

disease. In addition, the flexible and extensible design of the bike permits readily

changing the control system and incorporating additional I/O as needed to provide a wide

range of riding experiences. The network-enabled controller also permits real-time

remote access and remote data logging. Research platforms like this can be very effective

tools in evaluating the impact of new control paradigms for improving the motor skills of

riders with Parkinson’s. Validated clinical studies using high performance control and

data acquisition systems may provide a basis for transitioning high-impact exercise

regimens from a clinical setting to broad scale deployment in therapy centers and

eventually in the home.

To test and validate the effectiveness of the Smart Bike, forty-seven individuals

diagnosed with Parkinson’s disease completed three cycling sessions over a one week

period riding two Smart Bikes, and were evaluated for changes in motor function after

the exercise sessions were concluded. All individuals were randomly assigned to use

either the static or dynamic control mode during bike operation and were able to

successfully complete three 40-minute cycling sessions. Individuals who completed three

sessions of dynamic cycling showed a 13.8% improvement in the Unified Parkinson’s

Disease Rating Scale (UPDRS), while individuals in the static group worsened by 1.6%.

Comparing these results to previous findings in chapter 2 shows that the dynamic control

mode plays a role similar to tandem cycling used in previous PD bike exercise studies.

70
 The variability analysis study reveals that pattern irregularity in power is greater

in the static group compared to the dynamic group, indicating that the bike might provide

a stable influence on the patient’s exercise intensity, while maintaining elevated cadence

in the dynamic mode, which is believed to be of significant value from previous studies.

In contrast, the cadence data shows greater variability for the dynamic group than for the

static group. This variability is likely due to the inability of individuals with PD to

maintain a constant high-speed pedal cadence. Furthermore, variability was also

introduced when the bike in dynamic mode was required to increase or decrease pedal

speed to maintain the desired cadence. The PD patients rode the bike in static mode at a

self-selected cadence and thus showed lower variability during exercise bouts.

The observed variability in the sampled signals is likely not an artifact introduced

by the specific control algorithm used but is primarily due to the rider’s performance in

interacting with a bike under different speeds and loads. This is substantiated by the

correlation between signal variability and motor skill changes for PD riders from the

tandem bike study in chapter 2 (no controller used). Furthermore, the linear regression

model results demonstrate a positive correlation between the degrees of variability in

cadence with improvement in motor skill performance. This suggests a causal link with

variability and motor skill performance rather than an anomaly introduced by the specific

control algorithm used.

The statistical analysis supports the previous results on tandem cycling presented

in chapter 2, that temporal variability or lack of predictability (not quantified by

conventional statistical parameters such as variance or coefficient of variation) in cadence

in dynamic mode (assisted exercise) can be used to predict resulting improvements in

71
UPDRS Motor III scores. Lastly, using the MLR model, the predicted UPDRS Motor III

scores are highly correlated to measured UPDRS values for test subjects in the dynamic

group. These data provide insight into how times series analysis methods can be applied

to uncover potential features in the measured variables and how this information can be

used to correlate exercise parameters with improved motor function.

72
 Chapter 4

Electronic Coupling of a Tandem Bike

4.1 Introduction

The variability analysis results from ten patients who had exercised in two groups of

single and tandem cycling patients presented in chapter 2 showed that variability in

cadence and power (and heart rate) signals in tandem cycling is an important factor in

motor function improvement for PD patients as measured by UPDRS. Additionally,

analysis of data collected from patients who exercised using the Smart Bike discussed in

chapter 3 confirmed the relationship between variability in cadence and power signals

and improvement in motor function.

In this chapter, the interaction and power sharing between trainer (leader) and rider

(follower) on a tandem bike is studied by removing the mechanical coupling of the pedals

of a tandem bike and replacing it with electronic coupling. A tandem bike has been

equipped with servomotors, drives and various sensors coupled to a data acquisition

system, which are all connected to a Programmable Logic Controller (PLC). The

servomotors are capable of providing gear-like resistance to the cyclist, and since the

system imitates two-person tandem bike behavior, each of the two servos will service a

separate rider. These motors along with the motor drives are wired to the PLC, which will

force the motor to react to the cyclists’ increase or decrease in pace.

73
Figure 4.1: Block diagram of the electrically coupled tandem bike.

The system has been designed to operate in two modes: 1) data acquisition, and 2)

real-time bike control. Operating mode 1 is used to collect real-time performance data

from the trainer and rider using the sensors and devices connected to the bike pedals. The

synchronized data samples can be analyzed to determine the drag, elasticity, and backlash

in the coupling. Subsequent data analysis will then examine the response of the trainer to

disturbances (from the rider) and develop a model of how the trainer interacts with the

rider.

In operating mode 2, the common chain that mechanically connects the two riders is

removed and the model and information obtained from operating mode 1 are used to

74
connect the trainer and rider electronically in the tandem bike. In this case, the two

cyclists (trainer and rider) are electronically linked as if they were mechanically

connected thorough a standard tandem bicycle drivetrain (i.e. chain-coupled sprockets).

The motor/control system will be able to dynamically alter the cadence and torque

experienced by the trainer and rider through a real-time power management control

algorithm. In this mode, the system operates with a trainer and rider in both acquisition

and closed-loop control modes, or with a rider and no trainer where inputs to the rider are

provided by an input reference trajectory to the tandem bike controller model.

Moreover, the trainer model could be integrated with existing controller of a single

bike and serve to provide the feel and experience of a tandem bike to a rider on a single

automated bike. The testing, data analysis, and model development is intended to validate

the test platform for use in subsequent clinical trials and to further investigate its

effectiveness in providing customizable motor function benefits for PD riders.

4.2 System Architecture, Mechanical Design and Bike Modifications

In selecting the appropriate devices for the system, a variety of factors were

considered. The bike frame must be commercially available and modifiable. The tandem

bike needs to be rack-mounted to enable stationary cycling. The modifications consist of

removing the mechanical coupling (the shared chain) and connecting the servomotors

directly to the crank assemblies through auxiliary sprockets and chains, as well as

replacing the bikes’ cranksets with the power-meter crankset. The bicycle system requires

a device to provide resistance to the riders when the mechanical linkage is removed.

Servomotors will provide that resistance, and will be attached to the bike pedals with two

75
additional chains, one per crankset. This attachment design needs to involve minimal

modification of the bike in a simple, easily reproducible way.

A commercially available tandem bike from Bike Friday (Family Tandem Traveler,

figure 4.2) was selected and has been outfitted with servomotors, drives, and control and

data acquisition systems. At first step, to provide the bike with the ability to measure

power (torque) and cadence (rpm), the cranksets were replaced with SRM power-meter

cranksets (Quarq Riken GXP, Figure 4.2) which are equipped with wireless power and

cadence sensors, as well as the wireless ANT+ network devices. The wireless link

enables the transfer of data to the computer or another device for monitoring, recording,

and processing purposes. Power/torque and cadence (rpm) signals are transmitted to the

computer or other devices at a rate of one Hz. Two sets of Garmin USB ANT+ sticks

(model number: 010-10999-00, figure 4.2) were used to receive the power and cadence

signals of both riders on the computer. The data are displayed and logged for further

processing using the TrainerRoad1 software. Compatible receiver devices can be used to

display and record the data on a head unit installed on the bike.

(a) (b)

Figure 4.2 (a) Quarq Riken GXP power meter crankset, (b) Garmin USB ANT+ stick used to
receive the power meter signals on the computer.

1
http://www.trainerroad.com

76
 In next step, for operating mode 2, the mechanical coupling (chain) was removed,

and servomotors, drives, PLC and other devices have been used to link the riders

electronically. The Kurt Kinetic Road Machine trainer1 was selected for rack-mounting

the bike to provide stationary operation (figure 4.3).

Figure 4.3: Bike Friday Family Tandem and Kurt Kinetic Road Machine trainer.

The motors selected to electronically couple the tandem bike are the servomotors

which were used previously for the Smart Bike described in chapter 3 (TLY-A230P-

BJ62AA). These servos are rated at ½ hp (1.3 N-m) and are capable of both driving and

absorbing torque from the riders. Installed instrumentation permits continuous monitoring

of bike operation (e.g. cadence, pedal torque, and motor current) and riders’ condition

(e.g. heart rate). The chosen motor drives and PLC are devices that were also used

previously in the Smart Bike. The motor drive (Kinetix 350) implements a high-speed

inner loop controller that provides the appropriate voltage and current to the motor to

continually maintain the motor operating state specified by the PLC. The PLC

(CompactLogix Controller) runs the control algorithms and dynamically operates all the

systems and components in the system. Figure 4.4 shows the functional block diagram of

the electronically coupled tandem bike.

1
http://www.kurtkinetic.com

77
Figure 4.4: Functional diagram of the electronically coupling tandem bike.

4.2.1 Mechanical Design and Bike Modification

The next step toward the electronic coupling of the tandem bike after selecting the

required instruments/devices is the mechanical installation of the servomotors. After the

assessment of different scenarios, the decision was made to install the servomotors on the

bike chassis, to retain the mobility of the modified tandem bike with all installed

equipment.

To connect the motor shaft to the bike sprocket, two chain-rings have been designed

and built for the two servomotor/sprocket pairs. The designed chain-ring on the motor

shaft has 8 teeth and the sprocket has 39 teeth which gives a power/torque transmission

ratio of 39/8 between the servomotors and sprockets.

78
 The mechanical structure that supports mounting the servomotors on the bike chassis

is shown in figure 4.5. Figure 4.5 also shows the completed mechanical design of the bike

modification including the mechanical design of the servomotor mounting brackets,

installed servomotors, chain-rings and auxiliary chains.

Figure 4.5 Mechanical structure designed for servomotors attachment to the bike.

4.2.2 Hardware Design, Electrical and Electronics

Electrical/electronic components used for the electronic coupling of the tandem bike

are mounted in three rugged enclosures (figure 4.6) and are connected to the bike via

cables for motor power and control, operator interface, and sensor feedback. Most of the

components, such as the programmable controller (PLC), network adapter, safety relays,

79
and power supplies are mounted in the main enclosure and connected to the other two

enclosures containing the motor drives through the cables for power (voltage), control

signals and data transmission.

The overall block diagram and the location of the electrical/electronic components as

well as the communication network between different parts of the system are presented in

figure. 4.6.

Figure 4.6: Electrical/electronics components diagram of the instrumented tandem bike.

The control platform for the system is a commercially available programmable logic

controller (PLC) from Rockwell Automation (CompactLogix) which is a versatile

platform used across a broad range of automation and robotic applications. The PLC

communicates with the computer and motor drives through an Ethernet network to send

and receive commands and data. The PLC determines the appropriate motor speed and

load (torque) values and transmits the motor control information to the motor drives

(Kinetix 350) that are housed in two separate enclosures. The motor drive implements a

high-speed inner loop controller that provide the appropriate voltage and current to the

80
motor to continually maintain the motor operating state specified by the PLC. Motor

feedback to the drive is used to maintain proper motor speed and torque in spite of load

disturbances introduced by the rider. A photo of the completed system with all electrical

and modified mechanical parts is presented in figure 4.7.

Drive Box 1 Controller

Electrical
Drive Box 2 Electronics

Figure 4.7: Electronically coupling tandem bike, completed system.

The system is also equipped with an operator display (a rugged touch screen device,

PanelView™ Plus). The panel view will serve as the HMI for the bike and communicates

with the PLC through Ethernet to send the parameters entered by the trainer (captain) to

the PLC and to receive and display the required data from the PLC. The display can also

provide a graphical plot showing historical values for bike and rider operation.

4.3 Software and Control Algorithm

In this section, software programs and control algorithms developed to run and

81
control the system are described. The control algorithms that operate the bike have been

developed using RSLogix 5000 software from Rockwell Automation. The developed

algorithms were then downloaded to the PLC to provide real-time control of the bike

operation. New control algorithms can be readily implemented using the development

and operating platform (i.e. PC, PLC, motors, and drives) provided with the tandem bike.

4.3.1 Programming Software

Software development for the system mainly includes the bike control algorithms

developed to run on the PLC. Other microprocessor-based devices in the system, such as

the drives, were programmed by setting up parameters in the software functions. Similar

to the software package used in chapter 3 to develop the control system of the Smart

Bike, a suite of PLC software development tools from Rockwell Automation have been

used to develop the algorithms and routines, establish communication with the devices to

download the code, and to transmit and display the data. The software tools used for

programming and control of the system were listed previously in chapter 3.

4.3.2 Control Algorithm

In this section the main control algorithm used to connect the riders electronically

through the PLC, motor drives and servomotors is discussed. The code has been

developed to run on CompactLogix PLC using the RSLogixTM 5000 software and the

Ladder Programming Language, which is a common language in the development of

PLC programs for industrial control applications. Programming and control of the

servomotors have been done using the Motion Control toolbox integrated into the

RSLogixTM 5000 software.

82
 Two Motion Control functions were used in the ladder diagram code for the two

servomotors. The servomotors could be controlled and run separately. From different

control modes of the motor drives which were discussed earlier in chapter 3, two main

control modes have been used: Torque Control mode, and Velocity Control mode. The

general block diagram of the control system for electronically coupling two riders on the

tandem bike was previously shown in figures 4.1 and 4.4.

The first rider is the trainer/leader who commands the exercise session for the second

rider. Therefore, the algorithm is developed such that the second rider tracks the speed of

the first rider, while they share the power required to pedal the bike. Hence, drive 1

corresponding to the rider 1 (trainer/leader) is programmed in Torque Control mode

while the drive 2 (follower) is programmed in Velocity Control mode. Then the

appropriate control equations are used to produce the Torque and Velocity reference

signals for the servo drives based on the set-points and the feedback signals received

from the servomotors.

83
Figure 4.8: Block diagram of control algorithm for electronically coupling of the tandem bike.

Figure 4.8 shows the block diagram of the control system. The load (torque) on

servomotor 1, the resistance against the pedaling for the rider 1, is computed based on the

velocity and acceleration information of the riders that are measured and transmitted to

the PLC from motor feedback. The torque command (trim) equation of Axis 1 is very

similar to the torque command equation of the Static Mode in the Smart Bike (chapter 3).

Here, the torque command is a function of the velocity and acceleration of servo 1, as

well as the velocity and output torque of servo 2. The actual implemented equation for

the torque command of servo 1 is presented in equation 4.1.

Totque_Command1 =

0.4*Old_Torque_Command1+0.6*(Accel_Factor1*Axis_1_AverageAccel+Velocity_Factor1*Axi

s_1.AverageVelocity+Velocity_Factor1b*(Axis_1.AverageVelocity-Axis_2.AverageVelocity)-

0.5*Axis_2.TorqueReference) (4.1)

84
 Axis 2 corresponding to servomotor 2 is programmed in Velocity Control mode to

track the speed of servo 1. However, a torque command is given to the drive based on the

velocity and acceleration signals to define the resistance/load applied to the second rider.

This makes servo 2 operate in a very similar manner to the dynamic mode of the Smart

Bike in chapter 3. Equation 4.2 shows the actual implementation of the torque command

for servo 2.

Totque_Command2 =
0.4*Old_Torque_Command2+0.6*(Accel_Factor2*Axis_2_AverageAccel+Velocity_Factor2*(Ax
is_1.AverageVelocity-Axis_2.AverageVelocity)) (4.2)
Figure 4.9 shows a screen shot of the ladder diagram code implementing the torque

command equations for both servos.

Figure 4.9: Screen shot of the torque command equations implemented with ladder diagram using
the RSLogix 5000 software.

4.4 Practical Experiments and Test Results

In this section the practical test results for both modes of the system are

presented. First sets of experiments were done with mechanical coupling

85
(chain) and data for power and cadence signals were collected using the

power-meter sprockets. The next sets of experiments were performed when

the riders are connected electronically and data are collected using the motor

feedback, sensors, and electronic system developed for electronically

coupling of the tandem bike.

4.4.1 Mechanically Coupling Experiments and Data Collecting

In these tests, the power and cadence signals of both riders are measured by Quarq

SRM power-meter sprockets before removing the mechanical coupling (chain). The

power and cadence signals are transferred to the computer with the sampling rate of one

Hz through the ANT+ wireless network using a Garmin USB ANT Stick. Torque signals

are computed using the power and cadence data. Data for both riders on the tandem bike

were recorded using the power meters and wireless data recording system over several

sessions. Figure 4.10 shows 10 minutes of sampled data consisting of cadence, power and

torque signals. Cadence is measured in rpm, power is measured in watts and torque is

calculated in N-m. In this experiment we have a load change caused by changing the

bike’s gear around 200 sec.

86
 Cadence
100

rpm 80

60

40

20
0 100 200 300 400 500 600
Time (sec)
Total Torque
60

40
N-m

20

0
0 100 200 300 400 500 600
Time (sec)
Power
400
Total Pow er Rider1's Pow er Rider2's Pow er
300
watts

200

100

0
0 100 200 300 400 500 600
Time (sec)

Figure 4.10: Cadence, torque and power data for a 10-minute experiment session.

Here, as an example the recorded data were used to estimate the parameters of the

dynamical model between the total torque and cadence (speed). To find the model

transfer function, system identification methods were applied to the real data collected

from the practical tests. The model was found using the system identification GUI tool in

MATLAB, for which the input signal is the total torque and the output signal is the

cadence. The 1st and 2nd order transfer function models reasonably fit the data in most

instances, but the 2nd order model was better at matching/predicting the output. The final

models are as below:

87
 ( s) 1
G1 ( s)  
( s) 6.14s  0.29

( s) 0.43(s  0.027) (4.3)

G2 ( z )  
( s) (s  0.2)(s  0.017)

Figure 4.11 shows the real (measured) data and the simulated output data of the

identified models.

Measured and simulated model output

90
2nd order model
1st order model
80 measured data

70

60
Cadence (rpm)

50

40

30

20

10

0
0 100 200 300 400 500 600 700
Time (sec)

Figure 4.11: Real (measured) cadence and simulated model output data.

4.4.2 Electronically Coupling Evaluation and Test Results

The control algorithm for electronically coupling has been developed to control the

servomotors in such a way that the system emulates the mechanical coupling of the

tandem bike. The ideal situation occurs when the riders have exactly the same cadence

(velocity) and power is shared between the riders. In practice, there are many limitations

88
that can cause the system to move away from the ideal situation. Different scenarios have

been tested to evaluate the functionality of the system as an electronically coupled

tandem bike, and the results are presented in figures 4.12 to 4.14.

Figure 4.12 shows the cadence and power signals for a test session with only one

rider operating as the trainer/leader. As this figure shows, the pedals of second rider track

the speed of the first rider accurately in this situation, but since there is no second rider,

the power/torque signal of motor 2 is very small compared to motor 1, and its negative

sign shows that the motor consumes that power to rotate the pedal. Another scenario is

when only rider 2 (follower) is on the bike, for which case the cadence and power signals

are presented in figure 4.13. Although the developed algorithm has only considered the

situation in which the first rider is leading the exercise and the second rider tracks the

speed of the leader, figure 4.13 shows that the first rider also tracks the speed of the

second when there is only the second rider on the bike. Figure 4.14 shows the situation

when both riders are on the bike. At the beginning, the larger portion of the power/torque

comes from the first rider and second rider just follows the pedal cycles, resulting in good

cadence tracking. Then, the second rider starts to exert more force to the pedals and

contributes effort to the tandem cycling, so the first rider reduces their power/torque to

adjust the cadence at the 80 rpm level. Figure 4.15 displays a screenshot of the of

RSLogix 5000 Trends window for this scenario.

89
 Cadence
80
Leader (Rider1)
Follower (Rider 2)
60

Cadence (rpm)
40

20

-20
0 20 40 60 80 100 120 140 160 180
Time (sec)

Power
80
Leader (Rider 1)
60 Follower (Rider 2)

40
Power (watts)

20

-20

-40
0 20 40 60 80 100 120 140 160 180
Time (sec)

Figure 4.12: The cadence and power signals of both riders in a test scenario with one rider as the
trainer/leader.
Cadence
80 Leader (Rider 1)
Follower (Rider 2)
60
Cadence (rpm)

40

20

0 20 40 60 80 100 120 140 160 180

Time (sec)

Power
100
Leader (Rider 1)
80 Follower (Rider 2)

60
Power (watts)

40

20

-20

0 20 40 60 80 100 120 140 160 180

Time (sec)

Figure 4.13: The cadence and power signals of both riders in a test scenario with one rider as the
follower.

90
 Cadence
100

80

Leader (Rider 1)

Cadence (rpm)
60
Follower (Rider 2)
40

20

-20
0 20 40 60 80 100 120 140 160 180
Time (sec)

Power
150 Leader (Rider 1)
Follower (Rider 2)
100
Power (watts)

50

-50

-100
0 20 40 60 80 100 120 140 160 180
time (sec)

Figure 4.14: The cadence and power signals of both riders in a test scenario with both the
trainer/leader and follower.

Figure 4.15: A screenshot of the RSLogix 5000 Trends window displaying the velocity and
power of both riders in a test scenario with both the trainer/leader and follower.

91
4.5 Remote Tandem Cycling

The concept and development for electronically linking two riders on a tandem bike

could be extended to couple two single-rider bikes to behave like a tandem bike, thereby

allowing cyclists to bike together in different locations and feel like they are cycling on a

tandem bike (Figure 4.16). The desired signals can be transferred via internet or other

data networks. The most advanced prototypes could even replicate the behavior of one of

the tandem riders, allowing single PD riders to benefit from the effects of tandem cycling,

even within their own homes.

Figure 4.16: Electronically (remote) connection of two single bikes.

One step of remote tandem cycling was tested using the hardware and software

structures developed for the electronically coupling tandem bike to connect two sets of

single Smart Bike (developed in chapter 2) to work together as a tandem bike (Figure

4.17). Experimental test results of the system are very similar to the test results of the

electronically linked tandem bike presented in Figures 4.12 to 4.15.

92
 Drive Box 1 Drive Box 2 Controller
Electrical
Electronics

Smart Bike 1

Smart Bike 2

Figure 4.17: Electronically coupling of two (single) Smart Bikes to behave as a tandem bike.

Figure 4.18 shows the cadence and power signals for a test session with only one

rider operating on the first Smart Bike as the trainer/leader which is very similar to the

situation in Figure 4.12. As this figure shows, the second bike tracks the speed of the first

bike accurately, but since there is no second rider, the power/torque signal of second

Smart Bike is very small compared to Smart Bike 1, and its negative sign shows that the

motor consumes that power to rotate the pedal

Figure 4.19 shows the situation when two electronically connected Smart Bikes is

simulating the tandem cycling with two riders. As this figure shows, the speeds are

following each other while the power is shared between the riders. When the first bike’s

rider takes the larger portion of the power/torque, the second bike’s rider provides the

93
lower amount of power, and vice versa, resulting in good cadence tracking, which is the

same situation as Figure 4.14.

Cadence

80 Bike 1 (Leader)
Bike 2(Follower)
60
Cadence (rpm)

40

20

-20

0 20 40 60 80 100 120 140 160 180

Time (sec)

Power
120
Bike 1(Leader)
100 Bike 2(Follower)
80
Power (watts)

60

40

20

-20
0 20 40 60 80 100 120 140 160 180
Time (sec)

Figure 4.18: The cadence and power signals of both (coupled) Smart Bikes in a test scenario with
one rider as the trainer/leader.

94
 Cadence

80 Bike 1(Leader)
Bike 2(Follower)
60

Cadence (rpm)
40

20

-20

0 20 40 60 80 100 120 140 160 180 200

Time (sec)

Power
100
Bike 1(Leader)
80 Bike 2(Follower)

60
Power (watts)

40

20

-20

0 20 40 60 80 100 120 140 160 180 200

Time (sec)

Figure 4.19: The cadence and power signals of both riders on two (coupled) Smart Bikes in a test
scenario with both the trainer/leader and follower.

4.6 Conclusion

Many studies, including the research presented in this and previous chapters have

established the relationship between tandem cycling and improvements in motor function

for people with Parkinson’s disease. Currently there are no standalone clinical tools

capable of emulating the benefits of trainer assisted tandem bicycling while acquiring

real-time data from each rider. Here, the first step in developing such an instrument by

replacing the mechanical coupling on a tandem bike with an electronic connection has

been made. The electronically linked tandem bike is used to study the dynamics of

interaction and power sharing between the riders on the bike. This tool could also be used

95
in a clinical setting for isolating the relationship between tandem cycling and improving

the motor abilities of Parkinson’s patients.

Tandem cycling requires two participants, and in the case of Parkinson’s treatment

there is a patient and a trainer. The need for an additional, able-bodied cycling partner

(trainer) makes small scale and in-home application of tandem cycling impossible with

current technology. The trainer is required to contribute to pedaling speed, monitor

patient reaction, and consider the capabilities of the patients before and during an

exercise session. Foot position and resistance also affect the cyclists’ biomechanics.

Identifying which elements of exercise are optimal for PD individuals will aid in creating

better custom exercise plans by helping to prioritize types of exercise.

The electronically coupled tandem bike provides a foundation for further

rehabilitation system development. The developed system with additional sensors such as

respiration, heart rate, and other vitals, could allow the cycle to be used in a variety of

different clinical settings. Because previous experiments have shown a connection

between tandem cycling and motor skill improvement in PD patients, this prototype will

support testing different variables in tandem cycling. By identifying the unique

characteristics of tandem cycling, the bike may eventually help in a greater understanding

of the relationship between exercise and motor development in PD patients, as well as

providing the capability for customized in-home treatment with increased effectiveness.

96
Chapter 5

Summary and Future Directions

5.1 Summary

Recent studies have revealed that exercise and movement therapies have

significant benefits for individuals with Parkinson’s disease (PD). Several studies have

shown that high cadence cycling (forced exercise [1, 3, 8, 17] or active assisted cycling

[5, 30]) reduces the symptoms of the disease and improves the motor function in

individuals with PD. Variability in severity and progression of Parkinson’s disease

symptoms makes it challenging to design therapeutic interventions that provide maximal

benefit. Several studies have shown that high cadence cycling (on a tandem bike), results

in greater improvements in motor function than voluntary cycling. The precise

mechanism for differences in function following different types of exercise is unknown.

Moreover, the tools needed to support the clinical studies and research that uncovers the

fundamental factors of the cycling and to quantify the dynamics of motor function

improvements in PD patients do not exist.

In this research some novel approaches have been used to study the relationship

between high cadence cycling and motor function improvement in PD patients. First,

some advanced signal processing methods such as Approximate Entropy (ApEn), Sample

Entropy (SaEn), and Spectral Entropy (SpEn) were applied to the a set of collected data

97
of two groups of PD patients who had exercised in tandem and single cycling groups. The

complexity of biomechanical and physiological features of forced and voluntary cycling

was examined and the relationship between these features and the improvement in motor

function as measured by the Unified Parkinson’s Disease Rating Scale (UPDRS) was

studied.

Heart rate, cadence and power signals were analyzed using entropy signal

processing techniques. The variability analysis revealed that the pattern variability in

heart rate and power signals were greater in the voluntary group when compared to

forced group (tandem cycling). In contrast, variability in cadence was higher during

forced cycling. UPDRS Motor III scores predicted from the pattern variability data using

a multiple linear regression (MLR) model were highly correlated to measured scores in

the forced group. The study shows how time series analysis methods of biomechanical

and physiological parameters of exercise can be used to predict improvements in motor

function. This understanding is very important in the development of optimal exercise-

based rehabilitation programs for Parkinson’s disease.

In next step, the variability analysis outcomes were used to design and develop a

smart motorized bicycle for assisting Parkinson’s patients to improve motor function. The

developed Smart Bike can accurately control the rider’s experience at an accelerated

pedaling rate while capturing real-time test data. The design and development of the

electronics and hardware as well as the software and control algorithms were presented.

Two control algorithms have been developed for the bike; one that implements an inertia

load (static mode) and one that implements a speed reference (dynamic mode). In static

mode the bike operates as a regular exercise bike with programmable resistance (load)

98
that captures and records the required signals such as heart rate, cadence and power. In

dynamic mode the bike operates at a user-selected speed (cadence) with programmable

variability in speed that has been shown to be essential to achieving the desired motor

performance benefits. In addition, the flexible and extensible design of the bike permits

readily changing the control algorithm and incorporating additional I/O as needed to

provide a wide range of riding experiences. Furthermore, the network-enabled controller

will provide remote access to bike data during a riding session.

To evaluate the effectiveness of the Smart Bike, forty-seven individuals with PD

were randomly assigned to either the static or dynamic cycling group, and completed

three 40-minute exercise sessions every other day over a period of one week1. Heart rate,

cadence and power data were captured and recorded for each patient during exercise.

Motor function for each subject was assessed with the UPDRS Motor III test before and

after the three exercise sessions to evaluate the effect of exercise on functional abilities.

Individuals who completed three sessions of dynamic cycling showed the average of

13.8% improvement in the UPDRS, while individuals in the static cycling group

worsened by 1.6% in UPDRS.

To distinguish the static and dynamic cycling groups by biomechanical and

physiological features, the complexity of the recorded signals (power, heart rate, and

cadence) were examined using variability analysis techniques: ApEn, SaEn and SpEn. A

MLR model was used to associate these features to changes in motor function as

measured by the UPDRS Motor III scale. Pattern variability in cadence was greater in the

1
Clinical tests were done at School of Health Science, Kent State University by Dr. Angela Ridgel

99
dynamic group when compared to the static group. In contrast, variability in power was

greater for the static group. UPDRS Motor III scores predicted from the pattern

variability data were correlated to measured scores in both groups. These results are

consistent with the previous analysis results of forced (tandem) and voluntary (single)

cycling groups.

Lastly, to find more information about the interaction between the trainer and

patient on the tandem cycling, the pedal of a tandem bike were coupled electronically

(instead of mechanically by chain). A tandem bike has been equipped with servo motors,

drives and various sensors coupled to a data acquisition system, which are all connected

to a Programmable Logic Controller (PLC). The system has been designed to operate in

two modes: 1) data acquisition, and 2) real-time bike control. In operating mode 1, the

system is used to collect real-time performance data from the trainer and rider using the

sensors and devices connected to the bike pedals. The synchronized data samples can be

analyzed to determine the drag, elasticity, and backlash in the coupling. Subsequent data

analysis will then examine the response of the trainer to disturbances (from the rider) and

develop a model of how the trainer interacts with the rider.

In operating mode 2, the model and information obtained from operating mode 1 are

used to remove the mechanical coupling (common chain) and connect the trainer and

rider electronically in the tandem bike. In this case, the two cycles (trainer and rider) are

electronically linked as if they were mechanically connected thorough a standard tandem

bicycle drivetrain (i.e. chain-coupled sprockets). The motor/control system will be able to

dynamically alter the cadence and torque experienced by the trainer and rider through a

real-time power management control algorithm.

100
 Smart Bike and the electronically coupled tandem bike provide a foundation that can

be used for further development of rehabilitation systems for PD patients. The developed

system with additional sensors such as respiration, heart rate, and other vital devices,

could allow the cycle to be used in a variety of clinical settings. Because previous

experiments have shown a connection between tandem cycling and motor skill

improvement in PD patients, these prototypes will support testing different variables in

tandem cycling. By identifying the unique characteristics of tandem cycling, the system

may eventually help in a greater understanding of the relationship between exercise and

motor development in PD patients, as well as providing the capability for customized in-

home treatment with increased effectiveness.

This dissertation has provided a novel, fully instrumented, automatically controlled

cycling test-bed to support research and clinical studies to investigate the factors causing

the significant improvement in motor skills from high cadence (accelerated) cycling. This

platform provides a rigorous test-bed for correlating cycle operation with the rider’s

physical state and resulting improvement in motor skills for PD patients. Furthermore,

this research platform can be used for automatically tailoring an exercise regimen for

individuals with different skill levels and improvement profiles. Exercise programs may

be readily optimized for each individual with minimum effort.

5.2 Future Direction and Suggestions

This dissertation has identified critical elements in high-intensity (forced or

active-assisted) exercise that may predict improved motor function in riders diagnosed

with Parkinson’s disease. However, there exists variability in each individual’s response

101
to exercise and the mechanism causing improvement is not well understood. Future work

could focus on the development of patient-specific algorithms and methods that can adapt

to changing patient conditions during a given exercise session as well as across multiple

exercise sessions. Future studies may also examine the role of proprioceptive input

during a bout of high cadence cycling on alterations of motor function using

neuroimaging and some other techniques.

Here, more details about some ideas on the future direction and extension of this

research. These ideas briefly discuss the future extension of the research and the other

related methods that can be used to design the optimal exercise regimen for the PD

patients as well as evaluating the exercise programs and patient condition more precisely.

Although many studies have documented the benefits of exercise, it is unclear what

elements (i.e. dosage, intensity, intervention type) constitute an optimal exercise

intervention for people with PD. Furthermore, progression of the disease often requires

re-assessments and changes to motor rehabilitation programs.

5.2.1 Adaptive Dynamic Exercise Program

Earlier in this dissertation, the relationship between variability in exercise signals

(cadence, power and heart rate) and UPDRS change in PD patients was studied, and the

results were used to design and develop the Smart Bike (Chapter 2 and 3). Primary test of

the Smart Bike on a group of twenty four PD patients indicated an average of 13.8%

improvement in motor function as measured by UPDRS change after three sessions of

exercise within a week. Better results could be reached with an adaptive exercise system

which uses the real-time data signals as well as previously recorded data (changes in

102
motor function, bike signals, and physiological data) to compute adaptive exercise

parameters within the exercise session as well as session by session.

Figure 5.1 shows the proposed block diagram for the closed loop adaptive control

of the exercise parameters within a session. Based on the previous data results, the

exercise planning center exports the appropriate control parameters for the bike

considering target UPDRS changes. During the exercise, bike and patient signals are

being processed in real-time to evaluate the exercise quality and calculate the expected

UPDRS change. The feedback loop applies the results to the planning system to modify

the control parameters. These data are also logged to be used in future sessions.

Moreover, the system provides a platform for automatically controlling rider-bike

interaction to optimize the benefit from accelerated dynamic cycling.

Figure 5.1: Diagram of adaptive/dynamic control within the exercise session.

Figure 5.2 shows the block diagram of the session by session control scheme

which extends the adaptation exercise algorithm to subsequent exercise sessions. The

exercise planning center provides the appropriate data and instructions to the adaptive

exercise control system within the session based on the target UPDRS and previous data.

103
Session planning center runs and controls the bike within the session during the exercise

based on the information received from the exercise planning center. After each session

actual UPDRS will be measured and expected UPDRS will be computed based on the

recorded signals from the bike and patient. These data will be used by exercise planning

center to define the next session exercise parameters.

Figure 5.2: Diagram of session by session adaptive exercise control.

This adaptive-dynamic exercise system will provide the optimal exercise program

for individuals with different skill levels and improvement profiles. Exercise programs

could be optimized for each patient based on the individual conditions and skill level to

provide the most benefit for the patient. Moreover, online data analysis permits rapid

identification of problems, rider fatigue, or unusual behavior and allows for corrective

control action and provides superior rider safety. Furthermore, data logging and remote

access capability could be used by physicians, trainers, and therapists interested in

monitoring in-home progress of PD patient exercise.

104
5.2.2 Real-Time Dopamine Change Measurement/ Monitoring of Animal Model

During Exercise

Animal models provide a critical tool to investigate the molecular mechanisms of

exercise-induced improvement in motor behavior. Several researches have documented

the dopamine release during the exercise in animal models [21, 22]. In [21], a study has

been presented which examines the dopamine change in a mouse model of Parkinson’s

disease subjected to high-intensity treadmill exercise. In that research the magnetic

resonance (MR) image of the mouse brain was used to investigate the dopamine change.

Besides that, several studies have reported the use of miniaturized wireless

devices for monitoring the chemical gradients such as dopamine level in the brain of

ambulatory subjects [20, 24]. In [20], a miniaturized polyimide-encased push–pull probe

was developed and used to measure basal neurotransmitter spatial gradients within brain

of live animals with 0.004 mm3 resolution. In [24], a miniaturized device has been

developed and used for wireless monitoring of extracellular dopamine levels in the brain

of an ambulatory rat using fast-scan cyclic voltammetry at a carbon-fiber microelectrode.

A future research can employ these methods, using a miniaturized wireless circuit

to provide the real-time measurement and monitoring of the dopamine changes in the rat

brain during (and after) the sets of defined exercise sessions. Such a research can study

the effectiveness of different exercise types (such as high intensity treadmill and high

cadence cycling) as well as the features of the exercise (speed, time, number of sessions,

and so on).

105
5.2.3 Shaking Measurement for Monitoring UPDRS and Evaluating the Quality of

Life in PD Patients

Through the advent of miniature sensors such as accelerometers, many studies

have focused on the application of wearable technology to monitor older adults and

subjects with chronic conditions in the home and community settings [23]. Advancement

of enabling technologies such as sensor technology, communication technology, and data

analysis techniques has allowed researchers to implement wearable systems in different

applications and devices.

Moreover, the advent of the miniaturized accelerometers has brought the

capability of evaluating Parkinson’s disease by tremor characteristics and temporal

attributes using wireless 3D MEMS accelerometers [21, 22]. Accelerometer systems have

also been tested and evaluated for ascertaining general status, drug therapy efficacy, and

amelioration of Parkinson’s disease based on deep brain stimulation parameter settings

[23].

Taking these new technologies to account, another research can be focused on

using the wearable accelerometers to measure and monitor the tremor (hand

vibration/shaking) in PD patients and use it for monitoring the UPDRS change. The first

step is using the signal processing methods and real data experiments to find a

relationship between UPDRS and measured shaking/ vibration signals. In next step, the

wearable accelerometer can be used for real-time monitoring of the UPDRS change in

PD patients during (and after) the exercise session. Such a system can even be used for

following up the UPDRS changes in PD patients in daily activities to track the status of

the disease and evaluate the quality of life in PD patients.

106
Bibliography

[1] A. L. Ridgel, H. M. Abdar, J. L. Alberts, F. M. Discenzo, K. A. Loparo, “Correlation

of motor skill changes with variability in cadence during forced and voluntary

cycling in individuals with Parkinson’s disease”, IEEE Transactions on Neural

Systems and Rehabilitation Engineering, May 2013.

[2] A. L. Ridgel, K. A. Loparo, F. M. Discenzo, “Development of an intelligent bicycle

for rehabilitation in Parkinson’s disease”, NIH R21 (R21HD068846).

[3] A. L. Ridgel, J. Vitek, J. L. Alberts., “Forced, not voluntary, exercise improves

motor function in Parkinson's disease patients”, Neurorehabilitation and Neural

Repair. 2009; 23(6):600-8.

[4] Technical report for “Development a smart bike for rehabilitation in Parkinson’s

disease”, Rockwell Automation Co., Dec 2012.

[5] Ridgel AL, Peacock CA, Fickes EJ, Kim CH. “Active-assisted cycling improves

tremor and bradykinesia in Parkinson's disease”, Arch Phys Med Rehabil. 2012 Nov;

93(11):2049-54.

[6] S. M. Pincus, “Approximate entropy as a measure of system complexity”, Proc Natl

Acad Sci U S A, vol. 88, (no. 6), pp. 2297-301, Mar 15 1991.

[7] F. Kaffashi, R. Foglyano, C. G. Wilson, K. A. Loparo, “The effect of time delay on

Approximate & Sample Entropy calculations”, Physica D 237, 3069–3074 (2008).

[8] A. L. Ridgel, L. Vitek, M. J. Phillips, M. L. Lowe, M. Hutson, J. L. Alberts,

“Improved motor function and cortical activation in Parkinson’s disease patients

following acute forced-exercise”, Medicine & Science in Sports & Exercise,

2009;41(5):148.

107
[9] Herman, N. Giladi, L. Gruendlinger, “Six weeks of intensive treadmill training

improves gait and quality of life in patients with Parkinson’s disease: a pilot study”,

Arch Phys Med Rehabil 2007; 88:1154–1158.

[10] I. Miyai, Y. Fujimoto, H. Yamamoto, “Long term effect of body weight-supported

treadmill training in Parkinson's disease: a randomized controlled trial”, Arch Phys

Med Rehabil 2002; 83: 1370–1373.

[11] G. M. Petzinger, B. M. Fisher, S. McEwen, J. A. Beeler, J. P. Walsh, M. W.

Jakowec. “Exercise-enhanced neuroplasticity targeting motor and cognitive circuitry

in Parkinson's disease”, Lancet Neurol 2013;12:716-726.

[12] D. M. Huse, K. Schulman,L. Orsini, J. Castelli-Haley, S. Kennedy, G. Lenhart.

“Burden of illness in Parkinson's disease”, Mov Disord, 2005 Nov, 20(11):1449-54.

[13] Schenkman M, Hall DA, Baron AE, Schwartz RS, Mettler P, Kohrt WM. “Exercise

for People in Early- or Mid-Stage Parkinson Disease: A 16-Month Randomized

Controlled Trial”, Phys Ther. 2012; 92 (11):1395–410.

[14] Wilson, KA, Phillips, RS, Abdar, HM, Discenzo, FM., Loparo, KA, Ridgel, AL.

“Dynamic Cycling Promotes Upper Extremity Motor Improvements in Parkinson’s

disease”. Medicine and Science in Sports and Exercise, 2013, 45 (5), 355-355

[15] Scandalis TA, Bosak A, Berliner JC, Helman LL, Wells MR, “Resistance training

and gait function in patients with Parkinson's disease”, Am J Phys Med Rehabil.

2001;80:38–43.

[16] Crizzle AM, Newhouse IJ, “Is Physical Exercise Beneficial for Persons with

Parkinson’s disease?”, Clin J Sport Med. 2006;16:422–425.

108
[17] Alberts J., Linder S., Penko A., Lowe1 M., Philips M., “It Is Not About the Bike, It

Is About the Pedaling, Forced Exercise and Parkinson's Disease”, Exerc Sport Sci

Rev. 2011; 39(4):177-186. (http://www.medscape.com/viewarticle/751998)

[18] Peacock C.A., Sanders G.J., Wilson K.A., Fickes-Ryan E.J., Corbett D.B., Ridgel

A.L. “Introducing a multifaceted exercise intervention particular to older adults

diagnosed with Parkinson’s disease: A preliminary study Aging and Clinical

Experimental Research”. Springer Int. Pub., Dec 2013

[19] M.G. Vuckovic, Q. Li, B. Fisher, A. Nacca, R.M. Leahy, J.P. Walsh, J. Mukherjee, C.

Williams, M.W. Jakowec, G.M. Petzinger, “Exercise elevates dopamine D2 receptor in a

mouse model of Parkinson′s disease: In Vivo Imaging with [18F]Fallypride, Movement

Disorders, 25 (2010), pp. 2777–2784

[20] Slaney TR, Mabrouk OS, Porter-Stransky KA, Aragona BJ, Kennedy RT,

“Chemical Gradients within Brain Extracellular Space Measured using Low Flow

Push-Pull Perfusion Sampling in Vivo”, ACS Chem Neurosci. 2013;4:321–329

[21] R. LeMoyne, C. Coroian, T. Mastroianni, “Quantification of Parkinson’s disease

characteristics using wireless accelerometers”, ICME International Conference on

Complex Medical Engineering, 2009. CME.

[22] R. LeMoyne, T Mastroianni and W Grundfest “Wireless accelerometer

configuration for monitoring Parkinson's disease hand tremor”, Scientific Research,

vol. 2, pp. 62-67, 2013

[23] Patel, S., Park, H., Bonato, P., Chan, L. and Rodgers, M. (2012) “A review of

wearable sensors and systems with application in rehabilitation”, Journal of

NeuroEngineering and Rehabilitation, 9, 1-17. doi:10.1186/1743-0003-9-21

109
[24] Jankovic J., “Parkinson's disease: clinical features and diagnosis”, J Neurol

Neurosurg Psychiatry 2008; 79: 368–376.

[25] Davie CA, “A review of Parkinson’s disease”, Br Med Bull, 2008;86:109–27

[26] G M. Earhart, “Dance as Therapy for Individuals with Parkinson Disease”, Eur J

Phys Rehabil Med. 2009 June; 45(2): 231–238.

[27] de Dreu, MJ, van der Wilk, AS, Poppe, E, Kwakkel, G, van Wegen, EE.

“Rehabilitation, exercise therapy and music in patients with Parkinson's disease: a

meta-analysis of the effects of music-based movement therapy on walking ability,

balance and quality of life”, Parkinsonism Related Disorders, 2012; 18 (1): 114–19.

[28] Duncan RP, Earhart GM, “Randomized controlled trial of community-based dancing to

modify disease progression in Parkinson disease”. Neurorehabil Neural Repair,

2012;26:132-43.

[29] Scheife RT, Schumock GT, Burstein A, et al. “Impact of Parkinson’s disease and its

pharmacologic treatment on quality of life and economic outcomes”, Am J Health Syst

Pharm 2000;57:953–62.

[30] B. Feodoroff, A. Raabe-Oetker, T. Stuckenschneider, “Effects of active-assisted

training on gait and tremor of Parkinson’s disease patients”, German Parkinson

Congress, Wurzburg, March 2013.

[31] J. S. Richman and J. R.Moorman, “Physiological time-series analysis using

approximate entropy and sample entropy,” Am. J. Physiol. Heart Circ. Physiol., vol.

278, no. 6, pp. H2039–H2049, Jun. 2000.

[32] J. S. Richman, D. E. Lake, and J. R. Moorman, “Sample entropy,” Methods

Enzymol., vol. 384, pp. 172–184, 2004.

110
[33] S. M. Pincus, “Assessing serial irregularity and its implications for health,” Ann. NY

Acad. Sci., vol. 954, pp. 245–267, Dec. 2001.

[34] N. Stergiou and L. M. Decker, “Human movement variability, nonlinear dynamics,

and pathology: Is there a connection?,” Human Movem.Sci., vol. 30, no. 5, pp. 869–

888, Oct. 2011.

[35] L. A. Lipsitz and A. L. Goldberger, “Loss of ‘complexity’ and aging. Potential

applications of fractals and chaos theory to senescence,” JAMA, vol. 267, no. 13, pp.

1806–1809, Apr. 1, 1992.

[36] D. E. Vaillancourt and K. M. Newell, “Changing complexity in human behavior and

physiology through aging and disease,” Neurobiol. Aging, vol. 23, no. 1, pp. 1–11,

Jan.–Feb. 2002.

[37] D. E. Vaillancourt, A. B. Slifkin, and K. M. Newell, “Regularity of force tremor in

Parkinson’s disease,” Clin. Neurophysiol., vol. 112, no. 9, pp. 1594–1603, Sep. 2001.

111