Classification of Sleep Apnea Syndrome From EEG Signals Using Spectrogram-Based Entropy and MLPNN Model

Abstract

In this study, we focus on the classification of sleep apnea syndrome from EEG signals by using the spectrogram-based entropy and multilayer perceptron neural network (MLPNN) classifier model. For this aim, EEG signals with different apnea-hypopnea index (AHI) taken from Polysomnography (PSG) recordings are divided into 30 sec windows, the windowed EEG signals are decomposing into frequency sub-bands by using short time Fourier transform (STFT), and then these frequency sub-bands are normalized into the range of [0, 1]. Next, Shannon entropy values of spectrograms obtained from the normalized frequency sub-bands are used as input to the MLPNN model for the classification of sleep apnea syndrome. Finally, although high correct classification ratios were achieved in the implemented classification experiments, the highest success ratio was succeeded in the classification of severe sleep apnea syndrome.

Keywords

• EEG signals
• sleep apnea
• Spectrogram
• entropy
• MLPNN.

Spektrogram-Tabanlı Entropi ve MLPNN Modeli Kullanılarak EEG İşaretlerinden Uyku Apne Sendromu Sınıflandırması

Abstract

Bu çalışmada, spektrogram-tabanlı entropi ve çok katmanlı algılayıcı sinir ağı (MLPNN) sınıflandırıcı modelini kullanarak EEG işaretlerinden uyku apne sendromunun sınıflandırılmasına odaklanılmaktadır. Bu amaç için, Polisomnografi (PSG) kayıtlarından alınan farklı apne-hipoapne indeksine (AHI) sahip EEG işaretleri 30 saniyelik pencerelere bölünmekte, pencereli EEG işaretleri kısa zamanlı Fourier dönüşümü (STFT) kullanılarak frekans alt bantlarına ayrıştırılmakta ve bu altbantlar [0, 1] aralığına normalize edilmektedir. Daha sonra, normalize edilen frekans altbantlarından elde edilen spektrogramların Shannon entropi değerleri uyku apne sendromu sınıflandırılması için MLPNN modeline giriş olarak kullanılmaktadır. Sonuç olarak, gerçekleştirilen sınıflandırma deneylerinde yüksek doğru sınıflandırma oranları elde edilmiş olsa da en yüksek başarı oranına şiddetli uyku apne sendromunun sınıflandırılmasında ulaşılmıştır.

Keywords

• EEG işareti
• Uyku apnesi
• Spektrogram
• Entropi
• MLPNN.

References

1. Abdulla, S., Diykh, M., Siuly, S. and Ali, M., (2023). An intelligent model involving multi-channels spectrum patterns based features for automatic sleep stage classification, International Journal of Medical Informatics, 171: 105001.
2. Acir, N., Guzelis, C., (2004). Automatic Recognition of Sleep Spindles in EEG by using Artificial Neural Networks, Expert Systems with Applications 27: 451–458.
3. Aksahin, M.F., Erdamar, A., Isık, A., Karaduman, A., (2017). Identification of Sleep Apnea using EEG, ECG and Respiratory Signals, IEEE.
4. Álvarez-Estévez D., Moret-Bonillo V., (2009). Fuzzy Reasoning used to Detect Apneic Events in the Sleep Apnea-Hypopnea Syndrome, Expert Systems with Applications 36: 7778–7785.
5. Aydin H., Ozgen F., Yetkin S, Sutcigil L., (2005). Sleep and Sleep-disordered Breathing, GATA Printing House (in Turkish).
6. Balakrishnan, G., Burli, D., Burk J.R., Lucas E. A., and Behbehani K., (2005). Comparison of a Sleep Quality Index Between Normal and Obstructive Sleep Apnea Patients, Engineering in Medicine and Biology, pp 1154–1157, Shanghai, China.
7. Civaner, O. F. and Kamsak, M., (2018). Classification of Pediatric Snoring Episodes using Deep Convolutional Neural Networks. In 2018 26th Signal Processing and Communications Applications Conference (SIU) (pp. 1-4), IEEE.
8. Dogan, B., (2016). Detection of Sleep Apnea using Respiratory Sounds. MsC Thesis, Istanbul Technical University Istanbul, Turkey.

9. Duman F., Erdamar A., Eroglu O., Telatar Z., Yetkin S., (2009). Efficient Sleep Spindle Detection Algorithm with Decision Tree, Expert System with Application 36: 9980-9985.
10. Erdamar A (2007). Model Development for Prediction of Sleep Apnea and Tongue Muscle Stimulation. Ph.D. Thesis, Hacettepe University, Ankara, Turkey.
11. Hafezi, M., Montazeri, N., Saha, S., Zhu, K., Gavrilovic, B., Yadollahi, A., & Taati, B. (2020). Sleep Apnea Severity Estimation from Tracheal Movements using a Deep Learning Model, IEEE Access, 8, 22641–22649.
12. Han. Y., Yarovoy, A. and Fioranelli, F. (2022). An approach for sleep apnea detection based on radar spectrogram envelopes, Proceedings of the 18th European Radar Conference, 5–7 April 2022, London, UK.
13. Karamustafaoglu, G., (2014). Automatic Diagnosis of Sleep Apnea by Processing Polysomnography Signals. MsC Thesis, Istanbul University, Istanbul, Turkey.
14. Koseoğlu M., Uyanık H., (2023). Effect of Spectrogram Parameters and Noise Types on the Performance of Spectro-Temporal Peaks based Audio Search Method, Gazi University Journal of Science, 36(2): 624-643.
15. Szilagyi L., Benyo Z., Szilagyi S. M., (2002). A New Method for Epileptic Waveform Recognition using Wavelet Decomposition and Artificial Neural Networks, Proceeding of the Second Joint EMBS/BMES Conference, 3, 2025–2026.
16. Ubeyli, E. D., (2008). Analysis of EEG Signals by Combining Eigenvector Methods and Multi-Class Support Vector Machines, Computers in Biology and Medicine, Volume 38, No 1, 14–22.
17. Ucar M.K., Bozkurt M.R., Polat K., Bilgin C., (2014). The Effect of Digital Filtering on Sleep Stage Classification using EEG Signals, ELECO 2014 Electrical - Electronics - Computer and Biomedical Engineering Symposium, Bursa, Turkey.
18. Umut, I., (2011). Developing Digital Signal Processing Software and Using Electroencephalography Records with This Software to Distinguish Individuals with Obstructive Sleep Apnea from Individuals without Apnea. Ph.D. Thesis. Trakya University, Edirne, Turkey.
19. Yoruk, A., (2019). Spectral Analysis of EEG Data in Sleep Apnea. MsC Thesis. Kutahya Dumlupınar University, Kutahya, Turkey.