COMPARATIVE ANALYSIS OF DIGITAL FILTERS FOR ELECTROGASTROGRAPHY (EGG) SIGNAL PREPROCESSING
Keywords:
Electrogastrography, digital filters, Butterworth filter, Chebyshev Type I filter, FIR filter, signal preprocessing, signal-to-noise ratio (SNR), biomedical signal processing, noise suppression, gastric slow wave preprocessing.Abstract
This paper presents a comparative analysis of three digital bandpass filters—Butterworth, Chebyshev Type I, and FIR Hamming window—for enhancing electrogastrography (EGG) signal preprocessing. EGG signals are severely compromised by physiological interference (~0.25 Hz respiration, ~1 Hz cardiac activity) and motion artifacts in the ultra-low-frequency band (0.015–0.15 Hz). Using synthetic EGG signals, filter performance is quantified using signal-to-noise ratio (SNR) improvement. Results demonstrate that FIR Hamming achieves the highest SNR improvement (+16.88 dB), while Butterworth produces the smoothest temporal output suitable for clinical interpretation, and Chebyshev Type I offers
optimal computational efficiency. These findings provide practical guidance for filter selection based on clinical diagnostic and machine learning requirements.
References
[1] K. L. Koch, "Handbook of Electrogastrography," 2nd ed. Oxford: Oxford University Press, 2013.
[2] P. Oczka, J. Bauer, and P. Pfusterer, "Electrogastrography as a diagnostic and therapeutic tool in gastrointestinal disorders," Frontiers in Medicine, vol. 11, p. 1243, 2024. doi: 10.3389/fmed.2024.1243
[3] N. Paskaranandavadivel, P. Cheng, W. Abeyratne, and N. Kingsley, "Gastrointestinal slow wave analysis using time-frequency analysis techniques," Neurogastroenterology & Motility, vol. 24, no. 10, pp. e489–e498, 2012. doi: 10.1111/nmo.12005
[4] C. Bruno, M. K. Habib, A. Siddiqui, and M. Al Masri, "Machine learning augmentation of gastroenterology: A systematic review," Journal of Clinical Medicine, vol. 13, no. 4, p. 1089, 2024. doi: 10.3390/jcm13041089
[5] R. Komorowski, R. Biederman, and T. Brodzicki, "Adaptive filtering of gastric electrogastrogram signals using ECG-derived respiration," BioMedical Engineering OnLine, vol. 14, no. 1, p. 42, 2015. doi: 0.1186/s12938-015-0038-0
[6] F. Cintra, J. P. Sanchez-Fernandez, and S. Morales, "Wavelet analysis of electrogastric signals for dyspepsia classification," IEEE Transactions on Biomedical Engineering, vol. 62, no. 8, pp. 2054–2065, 2015. doi: 10.1109/TBME.2015.2410698
[7] S. Oppenheim, R. Schafer, and J. Buck, "Discrete-Time Signal Processing," 2nd ed. Upper Saddle River, NJ: Prentice Hall, 1999.
APPENDIX A: IMPLEMENTATION DETAILS
The EGG filtering pipeline can be summarized as follows:
1. Signal Generation
- Gastric (0.05 Hz) + Respiration (0.25 Hz) + Cardiac (1 Hz) + Gaussian noise
- Sampling: 10 Hz, Duration: 600 seconds
2. Filter Design
- Butterworth (4th order): scipy.signal.butter()
- Chebyshev I (4th order, 1 dB ripple): scipy.signal.cheby1()
- FIR Hamming (201 taps): scipy.signal.firwin()
3. Filtering
- Apply zero-phase filtfilt() to avoid phase distortion
- Passband: 0.015–0.15 Hz (normogastria and harmonics)
4. Performance Analysis
- Compute SNR = 10*log10(P_signal / P_noise)
- Extract frequency response at 0.05, 0.25, 1.0 Hz
- Generate time/frequency-domain plots
5. Results Export
- CSV table with SNR metrics
- PNG figures for publication
Completecode: https://colab.research.google.com/drive/1rURgQUnHJtHFby8yk3hCOxOhX-IgIrQ5?usp=sharing
