Gait analysis and evaluation for hemiplegic elderly people

Objective: Stroke is defined by the World Health Organization as a clinical syndrome consisting of rapidly developing clinical signs of focal (or global in case of coma) disturbance of cerebral function lasting more than 24 hours or leading to death with no apparent cause other than a vascular origin (Mazzoni et al., 2009). Specifically in medical texts, hemiplegia is defined as a one-sided pattern of muscle overactivation and reduction in motor activity, leading to increased muscle tightness, and reflexes, weakness and loss of selective motor control (Barnes Fairhurst, 2012). In this study, we aimed to predict stages (Brunnstrom Classification) of hemiplegic patients by analyzing their walking data. Method: The triaxial accelerometer signals used in the study were separated to the sixth level using the Daubechies5 (DB5) wavelet transform in the MATLAB program. The features of the sixth level separated approach signal was selected to form new signals. Then, various classification algorithms were used in WEKA and MATLAB programs to analyze these data. In WEKA, Iterative Classifier Optimizier, AdaBoost, Bagging, Classification via Regression (CVR), Logit Boost, OneR, J48, Random Forest, Random SubSpace, Multi Class Classifier and RepTree algorithms were used. In MATLAB, Lineer Discriminant, Complex Tree, Subspace Discriminant and RUSBoosted Trees algorithms were used. Results: In this study, the stages of hemiplegic patients were estimated by using discrete wavelet transform method and machine learning algorithms on gait signals. LogitBoost and RUSBoosted Trees algorithms are found to be the best classifiers to predict Brunnstrom stages of hemiplegic patients. Result: It has been observed that wavelet transform method and machine learning methods can be used in stage determination which is important in the treatment of hemiplegic patients. LogitBoost and RUSBoosted Trees algorithms produced the best results. Key words: Gait analysis, hemiplegia, discrete wavelet transform, machine learning, classification