Research on Classification of Scoliosis Based on Pulse Signal under Dynamic Pressure

Abstract

Scoliosis is a disease with a high incidence and a serious health risk for patients. Existing methods of scoliosis detection mainly include manual physical detection and X-ray examination. However, there is misdiagnosis in manual detection and radiation in X-ray examination. This paper presents a non-invasive scoliosis detection method based on wrist pulse signal. Acquisition of pulse signals from scoliosis patients before and after treatment using a dynamic pressure pulse signal acquisition device developed by the Integrated Microsystem Laboratory of Peking University. After preprocessing, useful signals segmentation is performed and pulse signals under light, moderate and heavy pressure are extracted. Then, features are extracted from the time domain and frequency domain. Feature dimension reduction is also performed. Finally, the results of classification are obtained with SVM and XGBoost classifiers. In our research, the accuracy of SVM is 0.7856, F1-score is 0.8238, the accuracy of XGBoost is 0.8046, and F1-score is 0.8540. For scoliosis and non-scoliosis classification tasks, the experimental results show that the classification performance of XGBoost is better than that of SVM. The results showed that the method can effectively classify scoliosis.