Professional Documents
Culture Documents
Adaptive Individualized Modeling From Limited Clinical Data For Precise Anemia Management
Adaptive Individualized Modeling From Limited Clinical Data For Precise Anemia Management
fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3106856, IEEE Access
Date of publication xxxx 00, 0000, date of current version xxxx 00, 0000.
Digital Object Identifier 10.1109/ACCESS.2017.DOI
ABSTRACT It is challenging in practice to achieve a steady-state value for external human recombinant
erythropoietin (EPO) dosage to be administrated to maintain Hemoglobin (Hb) level within the desired range
of 11-12 g/dl based on population-based models for anemia management due to inter-and intra-variability of
the patients. On the other hand, Pharmacokinetic (PK) and Pharmacodynamic (PD) characteristics can vary
for the patients over the course of treatment due to aging and other life changes. To address the inter-and
intra-variability in anemia management, the semi-blind robust identification approach is proposed to obtain
individualized patient models using limited number of clinical patient data. Semi-blind robust identification
utilizes the effect of the initial condition during system identification to reduce the identification error. To
reflect the patient’s true dose-response relation as time passes and ensure the suitability of the individualized
model for the controller, the model (In)validation technique is discussed to provide appropriate mathematical
evidence about the suitability of the individualized model for dose prediction and controller design via
testing it on new clinical data of the particular patient. One-step-ahead prediction results are shown for
identified individualized patient models. The individualized patient models provide decision support to the
clinicians about EPO dosage to avoid undershoot or overshoot of Hb level. Minimum mean squared error
(MMSE) is calculated for the predicted values obtained by the models identified using semi-blind robust
identification with and without the model (In)validation against clinically acquired EPO-Hb data.
INDEX TERMS Anemia management, adaptive modeling, drug dosing, individualized patient modelling,
model (In)validation, robust system identification.
VOLUME 4, 2016 1
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3106856, IEEE Access
A. Affan et al.: Adaptive Individualized Modeling from Limited Clinical Data for Precise Anemia Management
tance to make the anemia treatment more targeted and ef- l1 -error bound. However, sometimes both time and frequency
ficient [7]– [14]. Most of the existing techniques however domain data may be available for a single system. In this case,
are population-based [7]– [11] and treat individual patients the mixed H∞ /l1 -robust identification procedure is used [10].
with a one-size-fits-all model [15]. While population-based The behavior of a system (patient) is highly affected by the
models may be useful in the analysis of drug properties at initial conditions at t=0. In anemia management, these could
large, however, they are not well suited to guide the treatment be due to previous drugs used by the patient. The techniques
of individual patients. To address the impact of inter-and such as l1 and H∞ in robust identification assume zero initial
intra-individual variability in dose-response characteristics, conditions for the system, which highly affects the result
individualized patient-specific models are needed instead of of the identification process. To solve this issue, the semi-
traditional population-based models, and assurance about blind identification technique is introduced in [11]– [13],
model suitability, i.e., model (In)validation for controller which accommodates the effect of the initial conditions of
design is required as well. the patient in identification process.
Many researchers have proposed different techniques to As the patients of CKD are treated with EPO for an
develop does-response models for number of different dis- extended time. It is reported in the literature that the Pharma-
eases. Bayesian-based drug delivery using population patient cokinetic (PK) and Pharmacodynamic (PD) characteristics
data is discussed in [16]. Bayesian framework for warfarin vary over the course of treatment due to aging, the addition of
dosing is presented in [17]. In [18], control oriented models new medication, new disease, and change in life habits [23]–
are developed for Type 1 Diabetes to capture the intra-patient [25]. As the patient’s PK and PD changes, the individualized
variability of insulin drug patients. Artificial Intelligence- patient model identified earlier may (In)valid to predict and
based neural network models are discussed in [19]– [21]. not suitable for controller design. Therefore, the individual-
Some researchers attempted to identify individual patient ized patient models are required to be adaptive in order to
models in [22]. In [5] the focus was to predict the value reflect the current patient’s dose-response status. The model
for EPO instead of Hb, which is not desired as predicting (In)validation technique tests the identified model on the
Hb level gives values of EPO but not true for vice versa. new clinically acquired data. The model is adapted if exiting
However, most of these models are based on predetermined model (In)validates [11].
model structure and noise distribution, that are not suit-
In this research work, the individual patient models for
able for anemia management as each patient poses different
CKD patients are obtained using a robust system identifica-
characteristics. The models obtained by the classical system
tion technique and patient-specific time-domain clinical data.
identification techniques do not yield good results as they
For model identification, the semi-blind system identification
assume that predefined model structure and model order are
technique is used [11]. The system model is a combination
close to the actual system (patient) and one mathematical
of a nominal model and uncertainty bound [15]. The identi-
model works for all dynamics regardless of complexity and
fication technique develops a nominal model and uncertainty
disturbances in the system (patient). The modeling stage
bound using patient-specific dose-response (EPO-Hb) data.
should include the effect of all disturbances and uncertainties
For the efficient controller, an appropriate mathematical ev-
which are being introduced during operations. In contrast
idence is required about the model suitability for controller
to classical identification techniques, robust system identi-
design. To address this requirement, the identified models
fication considers system (patient) uncertainties, unmodeled
are (In)validated using new patient-specific clinical data. We
dynamics, and model complexity, i.e., there is no assumption
summarize our contributions as follows:
on the model order, uncertainties, and noise affecting data.
The robust system identification procedure does not re- • We present a semi-blind robust system identification
quire large measurement data set or information of mea- technique to identify individualized patient models for
surement noise nor the information on the structure of the CKD patients using limited number of time-domain
system (patient) model to be identified is required. In this clinical data.
technique, the information on the maximum gain of the • We present the model (In)validation framework to show
model, K, the stability margin of the model response r, the suitability of models for prediction and controller
and a bound on the noise is required. Depending on the design by testing the models on the new clinical dataset.
nature of a posteriori information, the robust identification
techniques may lead to different identification methods. The The remainder of this paper is organized as follows; First,
selection of a robust system identification method depends we discuss the semi-blind robust system identification in
on the data type as time-domain data and frequency domain section-II. Section III discusses the model (In)validation and
data pose different characteristics and different techniques section-IV provides a complete algorithm of the semi-blind
are applied to obtain a model for the data type [6], [9]. robust identification with the model (In)validation. Section-V
For frequency domain data, the H∞ -identification technique presents the patient model results obtained by the one-step-
is implemented, which calculates the uncertainty bounds ahead prediction along with the error analysis followed by
in terms of the H∞ norm. For time-domain data, the l1 - the conclusion.
identification technique is implemented, which calculates an
2 VOLUME 4, 2016
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3106856, IEEE Access
A. Affan et al.: Adaptive Individualized Modeling from Limited Clinical Data for Precise Anemia Management
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3106856, IEEE Access
A. Affan et al.: Adaptive Individualized Modeling from Limited Clinical Data for Precise Anemia Management
for controller. To adapt the model to patient’s current status, the solution of the non-convex problem, the following convex
the next section discusses the model (In)validation technique relaxation is considered as discussed in [11]. Fig. 3 shows
which provides mathematical evidence about the suitability the alternative setup for the model (In) validation, where
.
of model for controller design by testing the model on data measurement noise is also affected by ∆, η = (1 + ∆) η̄.
unseen by identification process. Equation (7) can be modified as follows:
−
y = (I + ∆) Tg u + ΓM
III. MODEL (IN)VALIDATION g u + η̄ (8)
The actual plant consists of a nominal system model and Equation (8) is satisfied if a triple (u− , η̄, ∆) exists and
uncertainty. Once the individualized system (patient) model k∆k∞ < 1. Further details of convex relaxation of the model
has been identified by using a semi-blind robust identification (in)validation process, please see [11].
technique, then the model needs to be validated (or invali-
dated) by utilizing additional patient-specific data that is not
seen by the identification process before it is used in the con-
troller design. The theory offers the tools to design the model
which is robust and stable in theory but no information is
available regarding the stability and robustness of the model
in practice. This issue arises due to the system uncertainties,
unmodeled dynamics and model complexity in the system.
The model (In)validation techniques provide evidence about
the usability of the model under these uncertainties for
controller design to some extent by testing the model on a
FIGURE 3. The convex relaxed (In)validation framework for semi-blind robust
new experimental data set. By assuming multiplicative and identification.
additive noise, the problem can be stated as follows:
Problem 4: Given M experimental data points (yi , ui ), the
nominal model G (z) ∈ S, descriptions of admissible noise IV. ALGORITHM
N , uncertainty ∆ and initial conditions x0 , determine if there Section II and III provide theoretical insight on semi-blind
exists at least one triple (η, ∆, x0 ) that can reproduce the robust identification, and model (In)validation. Algorithm-
available experimental data by the following equation [11]: 1 shows the semi-blind robust system identification frame-
work for individualized predictive dose-response modeling.
y = (I + ∆) Tg u+ + Tgic x0 + η,
(6)
The algorithm starts with the patient ID/tag, the number of
where u+ is the input after t = 0, Tg maps the input to the clinical EPO-Hb data points to be used for identification, Nt ,
output, whereas Tgic maps the initial conditions to the output. stability bound, r, and the initial reduced model order, O.
The above problem has a term Tgic x0 , where x0 is the initial As this is not a fixed model order, the algorithm will select
condition and unknown. During the identification, this term the appropriate model order based on model (In)validation
at step 4. This value is to start the algorithm. Steps 1-3 in
Algorithm-1 produce models with model orders equal to the
number of data points used in the identification step. Based
on these parameters, the parametric and non-parametric por-
tions of the models are identified using a semi-blind robust
identification technique, and a reduced-order model is gener-
ated at steps 4-5 as shown. The next step is to determine the
number of experimental (clinical patient) data points, which
have not previously been used in identification for model
(In)validation. The algorithm uses the same number of data
points for the model (In)validation process as the numbers
FIGURE 2. The (In)validation framework for semi-blind robust identification. of coefficients. Then model (In)validation is performed as
explained in Section-III.
can be replaced with some term representing the effect of If the conditions for model (In)validation are satisfied, then
these initial conditions such as u− ∈ U− , Therefore, (6) can the algorithm displays the validated individualized patient
be modified as follows: model transfer function for the patient. Alternatively, if the
− (In)validation conditions are not satisfied, i.e., the model is
y = (I + ∆) Tg u+ + ΓM
g u + η, (7)
invalidated, then the reduced model order is increased for
where U− and ΓM g denote the past admissible inputs and Han- the same r. If the reduced-order model is greater than 6,
kel matrix, respectively. Figure 2 shows the framework for then r is increased. However, if r > 2 then the number
the (In)validation for the models obtained by the semi-blind of data points Nt used for model identification is increased
robust identification technique. To avoid the issues related to and the algorithm returns to step 1. In the last case, the
4 VOLUME 4, 2016
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3106856, IEEE Access
A. Affan et al.: Adaptive Individualized Modeling from Limited Clinical Data for Precise Anemia Management
25: exit
The model prediction results of the above model for patient-
1 are shown in Fig. 4. In the following figures of semi-
V. SIMULATION RESULTS blind robust identification simulation results, the red line
For the simulation purposes, the clinical data of 50 patients with square markers shows the actual clinical Hb values of
has been collected from the University of Louisville, Kid- the patient, the solid blue line shows the model prediction
ney Disease Program. This study is IRB approved by the results of the full order model obtained using semi-blind
Human Subjects Protection Program Office at University of robust identification technique, the green line with diamond
Louisville. Due to the space limitations results for only four markers shows the prediction results of the reduced-order
patients with the most challenging cases from the available model, the magenta vertical dashed line shows the number
database are shown in the paper. These patients represent of data points, Nt , used in the identification process for the
the good and bad responder to the medication. Each patient first time and the cyan vertical dashed line represents the
received three EPO dosages per week and the Hb level points where the model is (In)validated and then updated
was tested once a week. To match the input-output data (i.e., the model could not be validated with the recent pa-
dimension, the average of three EPO dosages is used as v, tient data through the model (In)validation algorithm and
while input u for the identification is the difference between therefore model identification algorithm is rerun as shown
v values. One-step-ahead prediction is used to show the in Algorithm-1.) Finally, the vertical blue bars in Figures
predicting capabilities of the identified models. The perfor- 4 through 11 show the weekly EPO dosages. It is impor-
VOLUME 4, 2016 5
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3106856, IEEE Access
A. Affan et al.: Adaptive Individualized Modeling from Limited Clinical Data for Precise Anemia Management
FIGURE 5. Prediction results for patient-1 model obtained by semi-blind Patent-13 is a challenging patient model because the values
robust identification without model (In)validation. of Hb are varying 8g/dl to 14g/dl and EPO dosages are
6 VOLUME 4, 2016
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3106856, IEEE Access
A. Affan et al.: Adaptive Individualized Modeling from Limited Clinical Data for Precise Anemia Management
FIGURE 7. Prediction results for patient-13 model obtained by semi-blind FIGURE 8. Prediction results for patient-19 model obtained by semi-blind
robust identification without model (In)validation. robust identification with model (In)validation.
also varying a lot. It can indicate that the patient is a poor of models obtained by semi-blind robust identification with-
responder to the medication and finding an optimal dosage out a model (In)validation. It can be seen by the prediction
is a difficult task for this patient. The prediction results of results and error table that semi-blind robust identification
patient-13 in Fig. 6 show that it was difficult to find an with the model (In)validation performed better by incorporat-
appropriate model which satisfies all the conditions even at ing the fluctuations in the patient characteristics which is not
earlier time steps, the model is updated three times within possible without a model (In)validation. The mathematical
the first 10 weeks. The interesting part is between the week expressions for patient-19 models obtained by semi-blind
17 and 23, where the EPO dosage is zero. This phenomenon robust identification with model (In)validation are as follows:
is challenging for the identification algorithm as the model
0.9z 5 −0.5z 4 −0.02z 3 +0.04z 2 −0.1z+0.01
identified is for non-zero EPO dosages, therefore, it will
z 5 −2.1z 4 +1.6z 3 −0.7z 2 +0.2z−0.02
n=5
0.5z5 +0.2z4 −0.6z3 +0.3z2 −0.1z+0.004 6≤ n≤10
be challenging for the model to be validated for zero EPO
G19 (z)= z 5 −2.1z 4 +1.6z 3 −0.6z 2 +0.1z−0.005 (13)
dosages. However, the semi-blind robust identification tech- 0.6z 4 −0.3z 3 +0.4z 2 −0.6z+1.3
11≤ n≤ 12
z 4 −1.3z 3 +0.5z 2 −0.3z+0.04
nique updated the model to capture the dynamics of this por- 0.7z5 −0.01z4 +0.5z3 +0.9z2 −0.7z+0.6
13≤ n≤50
z 5 −1.2z 4 +0.5z 3 −0.04z 2 −0.5z+0.2
tion as soon as the identification algorithm sensed this trend
in the data. The model (In)validation results for the patient- The patient-19 is a very interesting case as the EPO dosage
13 are shown in TABLE 2. All values of k∆k∞ are less than has frequent changes and Hb values are ranging between
one which satisfies the conditions of model (In)validation. 8g/dl and 14g/dl as shown in Fig. 8. This patient does not
To compare the performance of models shown in (11), the qualify as a poor responder to medication because the change
mathematical model of patient 13 using identification process between EPO dosages is low even though Hb values are
without model (In)validation is shown in (12) and prediction varying a lot. Therefore, it becomes an interesting scenario
results are shown in Fig. 7. where the patient response is not ordinary. However, the iden-
0.7z 5 − 0.5z 4 + 0.1z 3 − 0.1z 2 − 0.03z + 0.003 tification process with the model (In)validation can identify
G13 (z) = (12)
z 5 − 2z 4 + 1.3z 3 − 0.5z 2 + 0.2z − 0.02 the model and timely update the model to represent the true
The prediction error in models identified by semi-blind dose-response characteristics of the patient. TABLE 3 shows
robust identification with model (In)validation is smaller than the model (In)validation results for the patient-19. The values
of k∆k∞ are less than one which satisfies the condition
VOLUME 4, 2016 7
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3106856, IEEE Access
A. Affan et al.: Adaptive Individualized Modeling from Limited Clinical Data for Precise Anemia Management
FIGURE 9. Prediction results for patient-19 model obtained by semi-blind FIGURE 10. Prediction results for patient-21 model obtained by semi-blind
robust identification without model (In)validation. robust identification with model (In)validation.
for the model (In)validation. For the sake of comparison, 0.9z5 −0.3z4 −0.1z3 +0.002z2 −0.2z+0.1
the model of patient-19 without model (In)validation is as
z 5 −1.3z 4 +0.2z 3 +0.1z 2 +0.03z+0.03 5≤ n≤28
follows: −0.3z4 +0.3z3 +3z2 +0.02z−2
29≤ n≤38
G21 (z)= z 4 −0.3z 3 −1.2z 2 −0.03z+0.4 (15)
0.6z 4 +0.1z 3 −1.5z 2 +1.4z+2.4
z 4 +0.2z 3 −1.2z 2 −0.2z+0.3 39≤ n≤ 41
1.1z 5 − 0.9z 4 0.2z 3
− 0.08z 2
− 0.1z − 0.0004
+
−1.7z 3 +4.9z 2 −5.6z+3
G19 (z) = (14)
42≤ n≤51
z 5 − 2z 4 + 1.5z 3 − 0.6z 2 + 0.2z + 0.00014 z 3 −1.9z 2 +1.2z−0.24
8 VOLUME 4, 2016
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3106856, IEEE Access
A. Affan et al.: Adaptive Individualized Modeling from Limited Clinical Data for Precise Anemia Management
TABLE 4. Model (In)Validation Results of Patient-21 in the response of the patient due to the progress of the
disease, life changes of the patient and at one point the
Data Range (Weekly) k∆k∞ model may not any more accurately represent the specific
5-28 0.87 patient and thus it will need to be updated, i.e., adapted
29-38 0.92
39-41 0.94
accordingly to those changes. In this paper, we investigate
42-51 0.34 the model (In)validation technique as a model adaptation
strategy for the patient-specific individualized models. It
is shown by the simulations using time-domain patient-
specific clinical data that the model (In)validation algorithm
low error within an appropriate time frame for each patient.
leads to efficient adaptive models that capture the varying
This is especially important for efficient controller design
model dynamics. Performance of these identified models
and hence for finding an individualized drug-dose regimen
is compared with models identified by semi-blind robust
for the patients. The MMSE values for the identified models
identification without the model (In)validation by calculating
using model (In)validation are considerably smaller than of
MMSE. The MMSE results show that the models obtained
models identified without a model (In)validation and accept-
via model (In)validation performed very well, and these
able. This shows that the semi-blind robust identification
models are shown to be stable and robust based on the
performs better with the model (In)validation to identify the
model (In)validation results. The future aim is to design the
individualized models by using a considerably lower number
controller based on the individualized models identified by
of clinical patient-specific data points and by updating the
the semi-blind robust identification technique.
models as well using time-domain clinical patient-specific
data.
ACKNOWLEDGMENT
The authors would like to thank Drs. Michael E. Brier and
Adam E. Gaweda at Department of Medicine of the Univer-
sity of Louisville for their valuable discussions and providing
the clinical data. Funding from NSF under grant 1722825 is
gratefully acknowledged.
REFERENCES
[1] A.S. Levey, and J. Coresh, "Chronic kidney disease". The lancet, vol.379,
no. 9811, pp.165-180, 2012.
[2] E. Moore and R. Bellomo, "Erythropoietin (EPO) in acute kidney injury,"
Annals of Intensive Care, vol. 1, no. 1, pp. 1-10, 2011.
[3] M. Panjeta, I. Tahirovic, J. Karamehic, E. Sofic, O. Ridic and J. Coric,
"The Relation of Erythropoietin Towards Hemoglobin and Hematocrit in
Varying Degrees of Renal Insufficiency," Materia socio-medica, vol. 27,
no. 3, pp. 144-148, 2015.
FIGURE 12. Error Comparison between semi-blind robust identification with [4] F. Valderrábano, "Erythropoietin in chronic renal failure," Kidney Interna-
and without model (In)validation. tional, vol. 50, no.4, pp. 1373-1391, 1996.
[5] J. D. Martín-Guerrero, G. Camps-Valls, E. Soria-Olivas, A. J. Serrano-
López, J. J. Pérez-Ruixo and N. V. Jiménez-Torres, "Dosage individualiza-
tion of erythropoietin using a profile-dependent support vector regression,"
VI. CONCLUSION IEEE Trasaction on Biomedical Engineering, vol. 50, no.10, pp. 1136-
1142, 2003.
Patients with CKD suffer from the deficiency of red blood [6] R. S. Sánchez-Peña and M. Sznaier, "Robust Systems Theory and Appli-
cells due to the damaged kidneys and require a robust and cations", New York: Wiley, 1998.
stable drug management system to maintain the Hb level [7] L. Ljung, "System identification-theory for the user, 2nd ed., NJ: prentice-
hall, Upper Saddle River, 1999.
between 11-12 g/dL in response to the EPO dosage. In [8] M. C. Mazzaro, P. A. Parrilo and R. S. Sánchez Peña, "Robust Identifi-
this research work, the individualized patient models have cation: An approach to select the class of candidate models," Internation
developed using a semi-blind robust identification technique Journal of Control, vol. 74, no. 12, pp. 1210-1218, 2001.
[9] J. Schoukens, Y. Rolain and R. Pintelon, "On the use of parametric and
considering non-zero initial conditions. The measurement non-parametric noise-models in time- and frequency domain system iden-
data of 50 patients from the University of Louisville, Kidney tification," in 49th IEEE Conference on Decision and Control, Atlanta,
Disease Program has been used to verify the performance of GA, USA, 2010, pp. 316-321.
[10] T. Inanc, M. Sznaier, P. A. Parrilo and R. S. Sánchez Peña, "Robust identifi-
the approach. Due to the space limitations results for only cation with mixed parametric/nonparametric models and time/frequency-
four patients were shown in the paper. The ones shown in domain experiments: theory and an application," IEEE Transactions on
the paper show the results for the most challenging cases Control Systems Technology, vol. 9, no.4, pp. 608-617, 2001.
[11] M. Wenjing, "Semi-blind Robust Identification and Model (In)Validation,"
of the available data, especially for patients 13, 19, and 21. Ph.D. thesis, Pennsylvania State University, Pennsylvania, USA, 2006.
To design the controller based on the identified models, the [Online]. Available: https://etda.libraries.psu.edu/catalog/8026
model (In)validation framework is also described and applied [12] Y. Muhittin, "Robust Systems Theory Applications to Macroeconomic
Stabilization Problems," Ph.D. thesis, Pennsylvania State
to the identified models. This is because as time progresses University, Pennsylvania, USA, 2005. [Online]. Available:
during the treatment of a patient there will be some changes https://etda.libraries.psu.edu/catalog/6870
VOLUME 4, 2016 9
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3106856, IEEE Access
A. Affan et al.: Adaptive Individualized Modeling from Limited Clinical Data for Precise Anemia Management
[13] E. Akabua, T. Inanc, A. Gaweda, M. E. Brier, S. Kim and J. M. Zurada, JACEK M. ZURADA (LF’15) received the Ph.D.
"Individualized model discovery: The case of anemia patients," computer degree in electrical engineering from Gdansk Uni-
methods and programs in biomedicine, vol. 118, no. 1, pp. 23-33, 2014. versity of Technology, Poland, in 1975. He is a
professor of electrical and computer engineering at
[14] H. Mirinejad, A. E. Gaweda, M. E. Brier, J. M. Zurada and T. Inanc , the University of Louisville, USA. He authored or
"Individualized drug dosing using RBF-Galerkin method: Case of anemia coauthored several books and over 380 papers in
management in chronic kidney disease," Computer Methods and Programs computational intelligence, neural networks, ma-
Biomedicine, vol. 148, pp. 45-53, 2017.
chine learning, logic rule extraction, and bioin-
formatics, and delivered more than 100 presenta-
[15] S. Tøffner-Clausen, J. M. Grimble and A. M. Johnson, "System Identifica-
tion and Robust Control", Springer, 1996.
tions throughout the world. Prof. Zurada has been
awarded numerous distinctions, including the 2013 Joe Desch Innovation
[16] R. Bellazzi, "Drug Delivery Optimization through Bayesian Networks: An Award, the 2015 Distinguished Service Award, and five honorary professor-
Application to Erythropoietin Therapy in Uremic Anemia," Computers ships. In 2014, he served as the IEEE V-President and Technical Activities
and Biomedical Research, vol. 26, no. 3, pp. 274-293, 1996. (TAB Chair). He also chaired the IEEE TAB Periodicals Committee, and
the TAB Periodicals Review and Advisory Committee. He was the Editor-
[17] S. M. Öztaner, T. Taşkaya Temizel, S. R. Erdem and M. Özer, "A in-Chief of the IEEE Transactions on Neural Networks from 1997 to 2003
Bayesian Estimation Framework for Pharmacogenomics Driven Warfarin and an Associate Editor of the IEEE Transactions on Circuits and Systems,
Dosing: A Comparative Study," in IEEE Journal of Biomedical and Health Neural Networks, the Proceedings of the IEEE, Neurocomputing, and sev-
Informatics, vol. 19, no. 5, pp. 1724-1733, Sept. 2015. eral other journals. From 2004 to 2005, he was the President of the IEEE
Computational Intelligence Society. He has been a Board Member of IEEE,
[18] M. Moscoso-Vasquez, P. Colmegna, N. Rosales, F. Garelli and R. Sanchez- IEEE CIS, and IJCNN. He has been elected as a Member of Polish Academy
Pena, "Control-Oriented Model With Intra-Patient Variations for an Arti- of Sciences since 2005.
ficial Pancreas," in IEEE Journal of Biomedical and Health Informatics,
vol. 24, no. 9, pp. 2681-2689, Sept. 2020.
10 VOLUME 4, 2016
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/