Classification of Knee Abnormality Using sEMG Signals with Boosting Ensemble Approaches

Abstract

Knee problems, although increasing in the elderly, are one of the most important orthopedic problems that occur at any age and reduce the person's standard of living by making it difficult to move. In recent years, increasing in the use of surface Electromyography (sEMG) signals from muscles has highlighted the use of these signals in the detection of movement and movement disorders. In this study, sEMG signals, from patients with different knee abnormalities and healthy individuals, the muscles responsible for the bending (flexion) and stretching/extension (extension) movements of the knee (rectus femoris (RF), biceps femoris (FB), semitendinosus (ST), vastus medialis (VM)), recorded during gait, sitting, and standing were evaluated with some statistical-based features. Unlike the literature, the classification processes were alsoperformed for each muscle and each movement, and therefore the effect of the muscles on the classification performance was examined. The ensemble trees methods of Boosted and RUSboosted trees were used in the classification. The results show that the knee problem can be identified by using single muscle sEMG (RF) and single movement, with a performance about 92% for the movement of standing. The highest accuracy rate is obtained as 98.8% with Boosted Trees classifier for sitting by using all muscles sEMG signals.

Keywords

• knee disorder
• knee muscle
• surface EMG
• decision trees
• ensemble classification.

Boosting Ensemble Yaklaşımları ile sEMG Sinyallerini Kullanarak Diz Anormalliğinin Sınıflandırılması

Abstract

Diz sorunları yaşlılarda artmakla birlikte her yaşta ortaya çıkan ve hareket etmeyi zorlaştırarak kişinin yaşam standardını düşüren en önemli ortopedik sorunlardan biridir. Son yıllarda kaslardan alınan yüzey Elektromiyografi (sEMG) sinyallerinin kullanımının artması, bu sinyallerin hareket ve hareket bozukluklarının tespitinde kullanımını ön plana çıkarmıştır. Bu çalışmada, farklı diz anormallikleri olan hastalardan ve sağlıklı bireylerden gelen sEMG sinyalleri, dizin bükülme (fleksiyon) ve germe/ekstansiyon (ekstansiyon) hareketlerinden sorumlu kasların (rektus femoris (RF), biceps femoris (FB), yürüme, oturma ve ayakta durma sırasında kaydedilen semitendinosus (ST), vastus medialis (VM)) bazı istatistiksel temelli özelliklerle değerlendirildi. Literatürden farklı olarak her kas ve her hareket için sınıflandırma işlemleri de yapılmış ve bu nedenle kasların sınıflandırma performansına etkisi incelenmiştir. Sınıflandırmada Boosted ve RUSboosted ağaçlarının topluluk ağaçları yöntemleri kullanılmıştır. Sonuçlar, diz probleminin tek kas sEMG (RF) ve tek hareket kullanılarak, ayakta durma hareketi için yaklaşık %92 performansla tanımlanabileceğini göstermektedir. Tüm kas sEMG sinyalleri kullanılarak oturma için Boosted Trees sınıflandırıcısı ile en yüksek doğruluk oranı %98,8 olarak elde edilmiştir.

Keywords

• diz rahatsızlıkları
• diz kasları
• yüzey emg
• karar ağaçları

References

1. A. Vijayvargiya, N. Dey, R. Kumar, M. R. S. Tavares, “Comparative analysis of machine learning techniques for the classification of knee abnormality”, IEEE 5th International Conference on Computing Communication and Automation (ICCCA), Galgotias University, Greater Noida, UP, India. Oct 30-31, 2020
2. J. C. Huang, S. J. Shanq-JangRuan, W. C. Hsu, Y. T. Liu, C.H. Hsu, “3D-CLDNN: An effective architecture on deep neural network for sEMG-based lower limb abnormal recognition”, IEEE 8th Global Conference on Consumer Electronics (GCCE), 906-907, 2019.
3. A. Gautam, M. Panwar, D. Biswasand A. Acharyya, "MyoNet: A transfer-learning-based LRCN for lower limb movement recognition and knee joint angle prediction for remote monitoring of rehabilitation progress froms EMG", IEEE Journal of Translational Engineering in Healthand Medicine, 8, 1-10, 2020.
4. M. Janidarmian, K. Radecka, Z. Zilic, “Automated diagnosis of knee pathology using sensory data,” in Proc. 4th Int. Conf. Wireless Mobile Commun. Healthcare (Mobihealth), Nov. 2014, pp. 95–98, 2014.
5. R. Uzun, O. Erkaymaz, İ. Şenyer Yapıcı, “Comparison of artificial neural network and regression models to diagnose of knee disorder in different postures using surface Electromyography”, Gazi University Journal of Science, 31(1), 100-110, 2018.
6. O. Sanchez, J. Sotelo, M. Gonzales, G. Hernandez, “Emg dataset in lower limb data set”, UCI Machine Learning Repository, 2014, 2014-02.
7. A. M. Fraser, H. L. Swinney, “Independent coordinates for strange attractors from mutual information”, Phys. Rev. A, 33, 1134-1140, 1986.
8. T. Higuchi, “Approach to an irregular time-series on the basis of the fractal theory”, Physica D, 31, 277–283, 1988.

9. R. Agrawal, T. Imielinskiand A. Swami, “Database Mining: A Performance Perspective”, IEEE Transactions on Knowledge and Data Engineering, 5(6), 914-925, 1993.
10. Z.-H. Zhou, “Ensemble methods: foundations and algorithms”, New York: CRC Press, 2012.
11. Breiman, L., Bagging predictors. Machine Leraning, 24 (2) (1996) 123-140.
12. Efron, B.,Tibshirani, R., An Introduction to the Bootstrap.Chapman and Hall. London. (1993) 430