close

Вход

Забыли?

вход по аккаунту

?

978-3-319-70063-2 14

код для вставкиСкачать
Chapter 14
Application of Discrete Cosine Transform for
Pre-Filtering Signals in Electrogastrography
Dariusz Komorowski1 and Barbara Mika1
Abstract. Electrogastrography (EGG) is the technique of the cutaneous recording
of the myoelectrical activity of the stomach. Due to its noninvasiveness and correlation with the gastric motility it is the attractive complement for imaging stomach’s
diagnostic methods. As the EGG signal is the mixture of the electrical activity of
the stomach and surrounding organs so the raw EGG also contains the noise, the
electrocardiographic (ECG), and the respiration (RESP) signals. The aim of this paper is to present the effective tool for pre-filtering EGG signal. The filtering in the
Cosine Discrete Transform (DCT) domain has been proposed as an efficient tool
for denoising EGG signal. The obtained results are compared with the outcomes
determined by means of the traditional digital (Butterworth, in this case) filtering
method.
Keywords: EGG, Discrete Cosine Transform (DCT), filtering
14.1 Introduction
Electrogastrography is a method for recording the gastric myoelectrical activity using cutaneous electrodes placed on the anterior abdominal wall overlying the stomach [1]. From the pacemaker area located on the grater curvature of the stomach
electrical excitation spreads circumferentially and distally down the stomach walls
[5]. The signal recorded by means of the electrogastrography technique is called
electrogastrogram (EGG). The EGG recording besides the leading rhythm, called
the normogastric rhythm (2.4–3.6 cpm), can also possess the some pathological
rhythms: bradygastric (<2.4 cpm), tachygastirc (3.6–9.0 cpm), and arrhythmia (the
ranges slightly differ, depending on the institution) [9]. Nowadays, the EGG signal
Silesian University of Technology, Faculty of Biomedical Engineering, Department of Biosensors and Processing of Biomedical Signals, Zabrze, Poland, e-mail: dariusz.komorowski@
polsl.pl
© Springer International Publishing AG 2018
M. Gzik et al. (eds.), Innovations in Biomedical Engineering,
Advances in Intelligent Systems and Computing 623,
https://doi.org/10.1007/978-3-319-70063-2_14
125
126
Dariusz Komorowski, Barbara Mika
is a valuable complement to the other diagnostic methods of the stomach diseases
[2, 3, 11] especially those connected with the disturbance of the gastric motility.
The signals, which are available on the stomach surface include not only the EGG
signals but also the electrocardiographic signals, the signals connected with the respiratory movements, and the random noise [2]. The main component of the EGG
signal so called the slow wave has the frequency ∼3 cpm, that is, ∼0.05 Hz and
the amplitude in the range 50-500 μ V [6];thus, the EGG is the weak signal embedded in the much stronger components such as ECG and RESP. In order to use EGG
as the diagnostic tool that interfering signals must be removed. Fig.18.1 shows the
frequency ranges of EGG, RESP, and ECG signals. As these ranges are partially
covered, there are some difficulties in choosing the effective method of processing
the EGG. Respiratory signal is considered to be the strongest disturbing factors influencing both the ECG and the EGG in the overlapped ranges of the frequency.
Typically EGG signal is extracted by means of the analog or the digital band-pass
Fig. 14.1: The frequency ranges of signals available on the stomach surface: EGG magenta, RESP - red , ECG - blue
filters. The range of frequency for EGG signal is from 0.015 Hz to 0.15 Hz (0.99.0 cpm (cycle per minute)). Recently Hang Sik Shin [4] proposed Index-Blocked
Discrete Cosine Transform Filtering Method (IB-DCTFM) to design the ideal frequency range filter in the DCT domain for the biomedical signals. IB-DCTFM removes the unwanted frequency range signal in the time domain by blocking the
specific DCT index in the DCT domain. In this paper the IB-DCTFM algorithm is
suggested as the good performing method for attenuation the RESP signal from the
EGG data.
Fig.14.2 shows the exemplary 4-channel signal recorded from the surface of the
stomach, further called the raw EGG. It is easy to notice that the raw EGG contains
the ECG, RESP, and the random noise. The aim of this paper is to present the effective tool for pre-filtering EGG signal. The filtering in the Cosine Discrete Transform
(DCT) domain has been proposed as an alternative method to the commonly used
the filtering methods. The obtained results are compared with the results determined
by means of the Butterworth digital filtering [8].
127
14 Application of Discrete Cosine Transform. . .
4500
A1
A2
A3
A4
4000
Recorded Signal (u.u.)
3500
3000
2500
2000
1500
1000
00:00:00
00:00:01
00:00:02
00:00:04
00:00:03
00:00:06
00:00:05
time (hh:mm:ss)
00:00:07
00:00:08
00:00:09
00:00:10
Fig. 14.2: Exemplary of the recorded 4-channel signal from the surface of the stomach (raw EGG)
14.2 Methods
The discrete cosine transform (DCT) express a the finite discrete sequence data of a
sum of cosine functions oscillating at different frequencies eq.14.1.
N
y(k) = w(k) ∑ x(n)cos
n=1
π (2n − 1)(k − 1)
,
2N
where
1
w(1) = √
N
∧
w(k) =
2
N
f or
k = 1, ..., N
(14.1)
2≤k≤N
N denotes the length of x and y, which are the same size. The filtering process
is conducted by applying the DCT transform and its inverse version IDCT. The
DCT provide the decomposition of EGG signal, so that the unwanted components
(frequencies) can be identified in the EGG signal and removed, whereas the IDCT
allows to reconstruct the EGG signal already without interferences eq.14.2.
x(n) =
N
∑ w(k)y(k)cos
n=1
π (2n − 1)(k − 1)
,
2N
where
1
w(1) = √
N
∧
w(k) =
2
N
f or
n = 1, ..., N
(14.2)
2≤k≤N
The typical EGG examination takes above two hours and consist of three phases:
preprandial, meal, and postprandial. In order to assess the effectiveness of DCT
method the chosen parameter, such as, the normogastria index, obtained after DTC
filtering and traditional filtering have been compared. Due to the fact, that for the
128
Dariusz Komorowski, Barbara Mika
mentioned parameter the calculation of the dominant frequency (DF) are required,
the power spectrum density (PSD) of EGG signal was determined after applying the
DCT and the traditional filtering method. The EGG signals have been filtered and
then split into 30-minute sections. Next, each fragment has been divided into 60 or
256-second segments and the power spectrum density (PSD) of the each segment
has been calculated and the dominant frequencies have been found [7]. In addition
the overall (average) PSD of the section has been calculated. Based on the overall
PSD, dominant frequencies for each section has been found and the normogastria
rhythm index (NI) defined, as the ratio of the number of segments for which DF is
in the range of 2.4 − 3.6
cpm (the range slightly differs, depending on the institution) [7, 9, 10] to the total number of segments has been calculated. The results
of the methods performance are presented for the exemplary EGG signal from the
postprandial phase after the stomach stimulation with the 400 ml of water.
14.3 Results
The methods was validated by the means of the real human signals. The results are
presented for the EGG signal recorded from the young 24 year old woman with the
body mass index (BIM) 19.9. The short fragments of EGG after the classical, in
this case, the Butterworth filtering and the DCT filtering are presented in the Fig.
14.3 and Fig. 14.4, respectively. Fig. 14.5 presented the examples of overall PSD
[7] of EGG signals without filtering, after using Butterworth filter (4th order, cutoff
frequencies 0.9–9.0 cpm) and the block DCT filter (cutoff frequencies 0.9–9.0 cpm).
250
A1
A2
A3
A4
200
Filtered EGGD Signal (uV)
150
100
50
0
−50
−100
−150
−200
00:00:00
00:00:20
00:00:40
00:01:00
00:01:20
00:01:40
00:02:00
time (hh:mm:ss)
00:02:20
00:02:40
00:03:00
00:03:20
00:03:40
Fig. 14.3: The 4-channel EGG signal after the Butterworth filtering
00:04:00
129
14 Application of Discrete Cosine Transform. . .
250
A1
A2
A3
A4
200
Filtered EGGD Signal (uV)
150
100
50
0
−50
−100
−150
−200
00:00:00
00:00:20
00:00:40
00:01:00
00:01:20
00:01:40
00:02:00
time (hh:mm:ss)
00:02:20
00:02:40
00:03:00
00:03:20
00:03:40
00:04:00
Fig. 14.4: The 4-channel EGG signal after the DCT filtering
Overall PSD(PER4) of EGG Signal, Channel (1)
3.018
−70
Magnitude (dB)
−80
−90
−100
−110
−120
0
1
2
3
4
5
6
7
8
9
Frequency (cpm)
10
11
12
13
14
15
Overall PSD(PER4) of EGG Signal, Channel (1)
Overall PSD(PER4) of EGG Signal, Channel (1)
3.020
−70
−70
−80
−80
Magnitude (dB)
Magnitude (dB)
3.018
−90
−90
−100
−100
−110
−110
−120
0
1
2
3
4
5
6
7
8
9
Frequency (cpm)
10
11
12
13
14
15
−120
0
1
2
3
4
5
6
7
8
9
Frequency (cpm)
10
11
12
13
14
15
Fig. 14.5: The example of the overall PSD of the EGG signal (top-left) for original EGG, after the Butterworth filtering (bottom-left), and after the DTC filtering
(bottom-right)
130
Dariusz Komorowski, Barbara Mika
In order to verify the method performance the normogastria index (NI) [7] have
been calculated for the EGG signals processed by the traditional and the DCT methods. Fig. 14.6 and Fig. 14.7 present the distributions of the dominant frequency for
the 4-channel EGG signal processing by both the Butterworth and the DCT filtering.
In the Table 33.1 and 26.2 the values of NI index for all phases of the recorded EGG
signal, that is, preprandial, meal and postprandial obtained by the each of two kinds
of filtering are summarised. In addition, the postprandial phased has been divided
into the two 30-minute parts: postprandial 1, and postprandial 2, which corresponds
to the phase of the digestive process after the stomach stimulation with water (in
this case).
Magnitude (dB)
−50
Magnitude (dB)
−50
Magnitude (dB)
−50
Magnitude (dB)
Max Power Spectrum, channel (1)
−50
0
normo = 0.898
0
1
2
4
3
5
9
8
7
6
Frequency (cpm)
Max Power Spectrum, channel (2)
10
11
12
14
13
15
0
normo = 0.816
0
1
2
4
3
5
9
8
7
6
Frequency (cpm)
Max Power Spectrum, channel (3)
10
11
12
14
13
15
0
normo = 0.827
0
1
2
4
3
5
9
8
7
6
Frequency (cpm)
Max Power Spectrum, channel (4)
10
11
12
14
13
15
0
normo = 0.969
0
1
2
4
3
5
6
8
7
Frequency (cpm)
9
10
11
12
13
14
15
Fig. 14.6: The distribution of the dominant frequencies of 4-channel EGG signal
processing by the means of Butterworth filtering
Table 14.1: The NI values calculated for the exemplary of 4-channel (A1, A2, A3,
A4) EGG signal after the classical filtering
Signal
All
Preprandial
Meal
channnel A1
channnel A2
channnel A3
channnel A4
0,8980
0,8163
0,8265
0,9694
1,0000
0,7297
1,0000
1,0000
0,5000
1,0000
0,3333
1,0000
NI
Post prandial1
0,8158
0,8158
0,7632
0,9211
Post prandial2
1,0000
0,9412
0,7647
1,0000
131
14 Application of Discrete Cosine Transform. . .
Magnitude (dB)
−50
Magnitude (dB)
Max Power Spectrum, channel (1)
−50
0
normo = 0.908
0
1
2
4
3
5
9
8
7
6
Frequency (cpm)
Max Power Spectrum, channel (2)
10
11
12
14
13
15
0
normo = 0.786
Magnitude (dB)
−50
Magnitude (dB)
0
−50
1
2
4
3
5
9
8
7
6
Frequency (cpm)
Max Power Spectrum, channel (3)
10
11
12
14
13
15
0
normo = 0.786
0
1
2
4
3
5
9
8
7
6
Frequency (cpm)
Max Power Spectrum, channel (4)
10
11
12
14
13
15
0
normo = 0.939
0
1
2
4
3
5
6
8
7
Frequency (cpm)
9
10
11
12
13
14
15
Fig. 14.7: The distribution of the dominant frequencies of 4-channel EGG signal
processing by the means of the DCT filtering
Table 14.2: The NI values calculated for the exemplary of 4-channel (A1, A2, A3,
A4) EGG signal after the DCT filtering
Signal
All
Preprandial
Meal
channnel A1
channnel A2
channnel A3
channnel A4
0,9082
0,7857
0,7857
0,9388
1,0000
0,7568
1,0000
1,0000
0,6667
0,5000
0,3333
1,0000
NI
Post prandial1
0,8158
0,7895
0,7632
0,8421
Post prandial2
1,0000
0,9412
0,5294
1,0000
14.4 Discussion and Conclusions
In this paper the DCT filtering has been proposed as an alternative method for preprocessing the raw EGG data. The obtained results confirm the good performance of
the proposed method. The pure EGG data has been extracted from their cutaneous
raw recording. The outcomes of the Butterworth and the DCT filtering obtained
for the human real EGG data are comparable (Table 33.1, 26.2). The DCT method
slightly influence the parameters of EGG data such as the dominant frequencies and
the normogastria index. The use of the DCT filtering allows more precise shaping
the amplitude-frequency characteristic of the EGG signal processing track, which
may affect the values of the parameters of the EGG examination. The benefit of the
such preprocessing the EGG signal can be found in the case, when the DCT filtering
is used for the noise cancellation in the highly interfered the EGG signals. The block
DCT filtering introduce an additional noise connected with the Gibbs phenomenon.
Our future work will be concentrated on extending the research by comparing the
proposed method with the other known methods (e.g. adaptive filtering) and examining the influence of the level of the noise to the efficiency of filtering the EGG
signal by means of the proposed method.
132
Dariusz Komorowski, Barbara Mika
References
1. Alvarez, W.C.: The Electrogastrogram and What It Shows. Journal of the American Medical
Association, volume 78, 1922
2. Calder, S., O’Grady, G., Cheng, L. & Du, P., 2016.: A theoretical analysis of electrogastrography (EGG) signatures associated with gastric dysrhythmias. IEEE Trans. Biomed. Eng., In
Press
3. Chen, J.D.Z, McCallum, R.W.: Electrogastrograpthy: measurement, analysis and prospective
applications. Medical & Biomedical Engineering & Computing, volume 29, pp.339-350, 1991
4. Hang Sik Shin, Chungkeun Lee, Myoungho Lee., Ideal Filtering Approach on DCT Domain
for Biomedical Signals: Index Blocked DCT Filtering Method (IB-DCTFM), Springer Science
+ Business Media, LLC, DOI 10.1007/s10916-009-9289-2, 2009.
5. Koch, K.L., Stern, R.M.: Handbook of Electrogastrography. Oxford University Press, 2004
6. Levanon, Daniela., Chen, J.Z.: Electrogastrography: Its Role in Managing Gastric Disorders
Journal of Pediatric Gastroenterology & Nutrition 27(4):431-443, October 1998
7. Medtronic A/S., Polygram NetTM Reference Manual. Skovlunde (Denmark), 2002.
8. Oppenheim, Alan V., Ronald W. Schafer, and John R. Buck., Discrete-Time Signal Processing.
2nd Ed. Upper Saddle River, NJ: Prentice Hall,(1999).
9. Parkman H.P., Hasler W.L., Barnett J.L., Eaker E.Y.: Electrogastrography: a Document Prepared by The Gastric Section of The American Motility Society Clinical GI Motility Testing
Task Force. Neurogastroenterology and Motility, Vol. 15, 2003, pp. 89-102
10. Riezzo G., Russo F., Indrio F., Electrogastrography in Adults and Children: The Strength,
Pitfalls, and Clinical Significance of the Cutaneous Recording of the Gastric Electrical Activity. BioMed Research International, vol. 2013, Article ID 282757, 14 pages,
doi:10.1155/2013/282757,(2013).
11. Tomczyk A., Jonderko J., Multichannel electrogastrography as a non-invasive tool for evaluation of the gastric myoelectrical activity a study on reproducibility of electrogastrographic
parameters before and after a meal stimulation. Ann. Acad. Med. Siles., 61, 5, 2007.
12. Yin, J., Chen J.D.Z.: Electrogastrography: Methodology, Validation and Applications, Journal
of Neurogastroenterology and Motility. 2013;19(1):5-17. doi:10.5056/jnm.2013.19.1.5
Документ
Категория
Без категории
Просмотров
0
Размер файла
381 Кб
Теги
978, 70063, 319
1/--страниц
Пожаловаться на содержимое документа