Impaired balance assessment in older adults with muscle weakness caused by stroke

Abstract

Stroke causes a severe impact on movement in daily life. It affects static balance in the human who has impaired movement control. Assessing balance using evaluation criteria or a check-list can be replaced by movement monitoring. For example, one can use the Berg Balance Scale, which is the gold standard in clinical assessment. This tool can also be supplemented by electronic motion detector sensors. To analyze the balance assessment results, the physiotherapist uses statistical methods to interpret the data. This research studies the suitable classification algorithms for evaluating balance control in stroke patients who have muscle weakness. After finetuning, the proposed methodology will improve the algorithm’s accuracy of data prediction for measuring the validity of regaining balance while standing. The dataset consists of three main factors: personal information, a diagnostic result from a physiotherapist, and the balance control performance while standing still on the Nintendo Wii Fit Balance Board. By evaluating various scenarios with different combinations of attributes, the dataset with three attributes has the highest accuracy rate. The clinical assessment is used as ground truth for assessing the prediction on how to regain a patient’s balance control during standing. Among four algorithms: Support Vectors Machine, Decision Tree, Naive Bayesian, and Artificial Neural Network, the most accurate classification model is the Artificial Neural Network with 93% accuracy of prediction.