Professional Documents
Culture Documents
Bursikova Stanislava 12110 2
Bursikova Stanislava 12110 2
Bursikova Stanislava 12110 2
f ( ) d t
(1.1)
The equation (1.1) defines Hilbert transform. P is Cauchy principal value. Thanks to this it is possible to compute integral with points of singularity. Definition 1.1 [5]: Let [ , ] be a real interval and let f be a complex-valued function defined on [ , ] . If f is unbounded near an interior point of [ , ] , the integral of f over [ , ] does not always exist. However, the two limits (1.2) and (1.3)
lim
0
f ( x)dx
(1.2) (1.3)
lim
0
f ( x)dx
still may exist, and if they do their sum is called the improper integral of f over [ , ] and is denoted by the ordinary integration symbol (1.4).
f ( x)dx
(1.4)
Even if these two limits (1.2) and (1.3) do not exist, it may happen that the symmetric limit (1.5) exists.
lim+ f ( x)dx + f ( x)dx 0 +
(1.5)
If it does, it is called the principal value integral of f from to and is denoted by the symbol (1.6)
P f ( x)dx
a b
(1.6)
one day in a week in periodical intervals to make so-called big blood glucose profile. Treatment for diabetes mellitus is based on holding blood glucose level in normal range or on prevention of its big variation [2]. Blood glucose level depends on many factors: on food intake, on physical activity (it implicates decrease of blood glucose level), on brain-fag (with increased epinephrinaemia blood glucose level goes up) etc. After eating blood glucose level increases. Healthy human body responds by increased production of insulin. Diabetics control blood glucose variation by diet, by medicines or syringe application of insulin [2]. This paper targets the blood glucose profile of type 1 diabetics. Type 1 diabetes (dependent on insulin) originates by Beta-cells damage (cells of pancreas, where insulin is created) by auto-immune process [3].
There is shown an example of continuous glucose monitoring system in figure 2.1: 1) Measuring electrode patient has it placed in hypodermis for 3-5 days. Measuring electrode (sensor) periodically sends the information about generated amperage and voltage to the monitor. 2) Transmitter it is connected by short wire with measuring electrode and it is attached to skin by a special plaster. Transmitter broadcasts by radio waves every 5 minutes information about glucose concentration to the monitor. 3) Monitor it records measured values and shows its 5 minutes averages on the display. This number of data already enables applying of HHT analysis. Using of this method is justified by clear and explicit complicated mechanism of blood glucose interaction and regulation in human body. So this data profile surely doesnt fall into the stationary neither linear processes class.
Fig. 3. 1: Seven days blood glucose profile of type 1 diabetic. Curve disconnection is caused by sensor changing.
HHT method was applied on the first part of this data (see figure 3.2):
18
16
14
12
10
100
200
300
400
500
600
700
800
900
1000
t [5 min]
Fig. 3.2: First part of glucose data HHT Method will be applied on this data
There are IMF components of original data in figure 3.3. Certain patterns can be seen in this figure. Periodically repeated oscillations are already evident in the first IMF component. This oscillations are changed to alone waves in the following components. This phenomenon would be interpreted as an application of insulin in these intervals together with food intake. But if we look in figure again and more detailed we can uncover that this oscillations are repeated in about 17 hours intervals. It isnt periodical application of insulin flow from this. So this phenomenon is caused by some else mechanism. Unfortunately I obtained this data without any more information about patients life and thats why it is very difficult to deduce what could cause these oscillations.
Now it is possible to apply Hilbert Transform on these IMF components. The result can be seen in figure 3.4. There is a frequency spectrum of IMF 2 and IMF 3 components from the previous figure. There are negative frequencies, which have no physical meaning and are caused by very quick changes of amplitude. This problem can be well treated by using of data normalization. It is a signal transform to the modality with constant amplitude of 1. Process is as following: 1) Make absolute value of signal. 2) Count up an envelope defined by local maxima. 3) Divide absolute signal by this envelope. If the information about amplitude is important for us, we have to include properties of envelope to the result.
4
IMF 2
f(t)
2 0 -2 -4 2
100
200
300
400
500
600
700
800
900
1000
IMF 3
1
f(t)
0 -1
-2
100
200
300
400
t [5 min]
500
600
700
800
900
1000
1 0 -1 4 2 0 -2 2 0 -2 0 2 0 -2 0 2 0 -2 0 2 0 -2 0 2 0 -2 0 100 200 300 400 500 600 700 800 900 1000 100 200 300 400 500 600 700 800 900 1000 100 200 300 400 500 600 700 800 900 1000 100 200 300 400 500 600 700 800 900 1000 100 200 300 400 500 600 700 800 900 1000 0 100 200 300 400 500 t [5 min] 600 700 800 900 1000 0 100 200 300 400 500 600 700 800 900 1000
t [5 min]
IMF 2
f(t)
0 -1 -2 0 2 100 200 300 400 500 600 700 800 900 1000
IMF 3
1.5
f(t)
0.5
100
200
300
400
500
600
700
800
900
1000
t [5 min]
4. Conclusion:
Experiments with data of blood glucose profile data decomposed by EMD and further analyzed by Hilbert Transform show a chance for usage of this method for better understanding of this complicated process. Range of made experiments isnt enough for deducing clinical conclusions so far but from the technical point of view HHT is applicable for glucose profile analysis . For following research would be eligible to cooperate with diabetics, which use continuous blood glucose monitoring system to enable better data interpretation.
5. References:
[1] N.E. Huang, Z. Shen, S.R. Long, M.L. Wu, H.H. Shih, Q. Zheng, N.C. Yen, C.C. Tung and H.H. Liu: The empirical mode decomposition and Hilbert spectrum for nonlinear and nonstationary time series analysis, Proc. Roy. Soc. London A, Vol. 454, pp. 903995, 1998.
https://www.zdravcentra.cz/index.php?act=bq-31&aid=1&oid=123&OFF=30&DIR=&ORD=, 15.1.2007
[2]
http://prodia.milackove.net/clanek/co-je-vlastne-cukrovka, 15.1.2007 http://www.minimed.cz/senzory.php, 15.1.2007 M. Johansson: The Hilbert Transform, Master thesis, Vxjo University, Schweden