Gait classification of stroke survivors - An analytical study

Abstract

Gait of post-stroke patients have been analyzed and classified in this paper. Data pertaining to gait of both post-stroke patients and healthy people of the same age group were obtained and analyzed. Xsens motion capture system was used for obtaining this data from 40 people belonging to both categories. Advanced modern machine learning techniques such as Logistic Regression, Multilayer Perceptron (MLP), Support Vector Machine (SVM), and Extreme Gradient Boosting (XGBoost) were employed for classifying the data obtained. Accurate spatiotemporal parameters values for gait were obtained. Amongst the different machine learning techniques, XGBoost gave the most accurate classification result of about 96%. The walking patterns of the patients who had undergone a stroke attack were analyzed. severity of abnormal gait patterns was a key factor that was taken into consideration and points were given accordingly. The data presented in this paper can be used to develop diagnostic tools for gait rehabilitation of stroke survivors, to evaluate and estimate their way of walking in order to understand their progress. It can be used to understand the different types of walking disorders post-stroke and thereafter select the right kind of treatment that needs to be implemented for proper recovery.