EMG Tabanlı Zaman Sabiti Sinir Ağları Kullanarak Kas Yorgunluğu Bozukluklarının Tespiti için Yeni Bir Yaklaşım

Abstract

Son yıllarda, Akışkan Zaman Sabiti (AZS) sinir ağları, karmaşık, zamana bağlı verileri doğru bir şekilde modelleme konusundaki olağanüstü yetenekleri nedeniyle büyük ilgi görmüştür. Çeşitli alanlardaki uygulamaları araştırılmış olsa da elektromiyografi tabanlı kas yorgunluğu veya sakatlık tespiti için AZS Sinir Ağlarının kullanılma potansiyeli araştırılmamıştır. Bu araştırmanın birincil amacı, AZS sinir ağının bu alana özgü zorlukları ele almadaki etkinliğini göstermek ve potansiyel avantajlarına dair yeni bilgiler sunmaktır. Bu hedefi gerçekleştirmek için, hasta muayeneleri sırasında elde edilen EMG sinyallerini analiz etmek için bir AZS sinir ağı kullandık. Toplanan sinyallerden, Ortalama Mutlak Değer, Dalga Biçimi Uzunluğu, Sıfır Geçişleri, Eğim İşareti Değişiklikleri ve Merkez Frekansı dahil olmak üzere beş özelliği hesapladık. Bu özellikler, AZS sinir ağı için girdi olarak kullanıldı. Ağın zamansal verilere dayalı olarak değerleri tahmin etme yeteneği, sinir hasarı veya kas işlev bozukluğunun göstergesi olan sinyal değişikliklerini hassas bir şekilde izlemesini sağladı. EMG sinyallerinden kas yorgunluğunu tespit eetmede AZS sinir ağının performansını geleneksel yöntemlerle ve diğer sinir ağı tabanlı tekniklerle karşılaştırdık. Deneysel sonuçlarımız, AZS sinir ağının %99,72'lik yüksek bir doğrulama doğruluğu elde ettiğini ve bunun kas sakatlığını belirlemedeki etkinliğini gösterdiğini ortaya koymaktadır. Bu bulgular, AZS sinir ağlarının geleneksel yaklaşımlardan daha iyi performans gösterme potansiyeline sahip olduğunu ve EMG tabanlı kas yorgunluğu tespitinde başarılı olduğunu göstermektedir.

Keywords

• Gerçek zamanlı yorgunluk analizi
• Sporcu performansı izleme
• Elektromiyografi
• Akışkan Yapay Sinir Ağı.

A novel Approach for Muscle Fatigue Disorders Detection Using EMG Based Time-Constant Neural Networks

Abstract

In recent years, Liquid Time-Constant (LTC) Neural Networks have gained substantial interest due to their exceptional ability to accurately model complex, time-dependent data. Although their applications in various fields have been explored, the potential of utilizing LTC Neural Networks for electromyography-based muscle fatigue or disability detection has not been investigated. This research aims to showcase the effectiveness of LTC Neural Networks in addressing challenges unique to this domain and to offer new insights into its potential advantages. We employed an LTC Neural Network to analyze EMG signals obtained during patient examinations to accomplish this objective. We calculated five features from the collected signals, including Mean Absolute Value (MAV), Waveform Length (WL), Zero Crossings (ZC), Slope Sign Changes (SSC), and Center Frequency (CF). These features were used as input for the LTC Neural Network. The network's ability to predict values based on temporal data enabled it to precisely monitor signal changes indicative of nerve damage or muscle dysfunction. We compared the performance of the LTC Neural Network with traditional methods and other neural network-based techniques in detecting muscle fatigue from EMG signals. Our experimental results reveal that the LTC Neural Network achieved a high validation accuracy of % 99.72, indicating its effectiveness in identifying muscle disability. These findings suggest that LTC Neural Networks have the potential to outperform conventional approaches and provide successful results in the field of EMG-based muscle fatigue detection.

Keywords

• Real-time fatigue analysis
• Athlete performance monitoring
• Electromyography
• Liquid Artificial Neural Network

References

1. [1] González-Muñoz, A., Perez-Montilla, J. J., Cuevas-Cervera, M., Aguilar-García, M., Aguilar-Nuñez, D., Hamed-Hamed, D., ... & Navarro-Ledesma, S. , " Effects of Photobiomodulation in Sports Performance: A Literature Review," Applied Sciences, vol. 5, no. 13, p. 314, 2023. doi:10.3390/app13053147
2. [2] Cavuoto, L., & Megahed, F. , "Understanding fatigue and the implications for worker safety.," In ASSE Professional Development Conference and Exposition, OnePetro 2016, 26-29 June 2016, Atlanta, Georgia, USA [Online]. Available: https://foundation.assp.org/docs/BPCav_1217z.pdf
3. [3] R. M. Al-Mulla, F. Sepulveda and M. Colley, "A review of non-invasive techniques to detect and predict localised muscle fatigue," Sensors (Basel), vol. 11, no. 4, p. 3545–3594, 2011. doi:10.3390/s110403545
4. [4] P. Konrad, The ABC of EMG, Arizona: Noraxon INC. USA, 2006.
5. [5] Abdulhamit Subasi, M Kemal Kiymik, "Muscle fatigue detection in EMG using time-frequency methods, ICA and neural networks," Journal of Medical Systems, vol. 34(4), pp. 777-85, 2010 Aug. doi:10.1007/s10916-009-9292-7
6. [6] Hasani, R., " Interpretable recurrent neural networks in continuous-time control environments," Doctoral dissertation, Wien, 2020. doi:10.34726/hss.2020.78942
7. [7] Hasani, R., Lechner, M., Amini, A., Rus, D., & Grosu, R. , "Liquid time-constant networks," in Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 35, No. 9, pp. 7657-7666, 2021, May. doi:10.1609/aaai.v35i9.16936
8. [8] Rohit Hooda, Vedant Joshi, Manan Shah, "A comprehensive review of approaches to detect fatigue using machine learning techniques," Chronic Diseases and Translational Medicine, no. 10.1016: j.cdtm.2021.07.002, 25 August 2021. doi:10.1016/j.cdtm.2021.07.002

9. [9] Calderon-Cordova, C., Ramírez, C., Barros, V., Quezada-Sarmiento, P. A., & Barba-Guamán, L., "EMG signal patterns recognition based on feedforward Artificial Neural Network applied to robotic prosthesis myoelectric control," IEEE 2016 Future Technologies Conference (FTC) , pp. 868-875, 2016, December. doi:10.1109/FTC.2016.7821705
10. [10] Choudhary, M., Lokhande, M., Borse, R., & Bhute, A., "A machine learning approach to aid paralysis patients using EMG signals," Academic Press. Advanced Data Mining Tools and Methods for Social Computing, pp. 107-125, 2022). doi:10.1016/B978-0-32-385708-6.00013-8
11. [11] Xia, P., Hu, J., & Peng, Y. , "EMG‐based estimation of limb movement using deep learning with recurrent convolutional neural networks," Artificial organs, vol. 5, no. 42, pp. E67-E77, (2018). doi:10.1111/aor.13004
12. [12] Azhiri, R. B., Esmaeili, M., & Nourani, M. , "Real-time emg signal classification via recurrent neural networks," IEEE International Conference on Bioinformatics and Biomedicine (BIBM), pp. 2628-2635, 2021, December. doi:10.1109/BIBM52615.2021.9669872
13. [13] E. Somuncu and N. Aydın Atasoy, "Realization of character recognition application on text images by convolutional neural network," Journal of the Faculty of Engineering and Architecture of Gazi University, vol. 37, no. 1, pp. 17-28, 2021. doi:10.17341/gazimmfd.866552 [14] A. Taşdelen and B. Şen, "A hybrid CNN LSTM model for pre miRNA classifcation," Scientific Repots, vol. 11, no. 14125, 2021. doi:10.1038/s41598-021-93656-0
14. [15] Ahmed Ebied; Ahmed M. Awadallah; Mohamed A. Abbass; Yasser El-Sharkawy, "Upper Limb Muscle Fatigue Analysis Using Multi-channel Surface EMG," Novel Intelligent and Leading Emerging Sciences 2nd Conference (NILES) of Giza, Egypt, p. doi:10.1109/NILES50944.2020.9257909, 2020.
15. [16] M. Bidollahkhani, F. Atasoy and A. Hamdan, "LTC-SE: Expanding the Potential of Liquid Time-Constant Neural Networks for Scalable AI and Embedded Systems," 18 4 2023. [Online]. Available: https://arxiv.org/abs/2304.08691. [Accessed 25 7 2023].
16. [17] Al-Mulla, M. R., Sepulveda, F., & Colley, M. , "A review of non-invasive techniques to detect and predict localised muscle fatigue," Sensors, vol. 4, no. 11, pp. 3545-3594, 2011.
17. [18] Subasi, A., & Kiymik, M. K. , "Muscle fatigue detection in EMG using time–frequency methods, ICA and neural networks," Journal of medical systems, vol. 34, pp. 777-785, 2010. doi:10.1007/s10916-009-9292-7
18. [19] Batzianoulis, I., El-Khoury, S., Pirondini, E., Coscia, M., Micera, S., & Billard, A., " EMG-based decoding of grasp gestures in reaching-to-grasping motions," Robotics and Autonomous Systems, vol. 91, pp. 59-70, 2017. doi:10.1016/j.robot.2016.12.014