Application of a Multilayer Perceptron Neural Network for Differentiation of the Influence of Ankle Stretch Duration for Hemiplegic Affected Ankle Dorsiflexion Quantified by a Smartphone Functioning as a Wearable and Wireless Gyroscope Platform

Abstract

The amalgamation of wearable and wireless inertial sensor systems and machine learning with accessibility to Cloud computing resources offers the opportunity to substantially advance the efficacy of a therapeutic rehabilitation intervention. The smartphone represents a wearable and wireless inertial sensor system equipped with a gyroscope for the acquisition of clinically perceptible signal data. Machine learning, such as a multilayer perceptron neural network, can provide definitive distinction of an assortment of therapeutic prescriptions. With Cloud computing accessibility a subject and clinical team supervising the therapy endeavor can perform their roles in a remote Internet connected context. For example, a subject with a hemiplegic affected ankle can undergo a series of prescribed ankle stretch durations using a wedge board. The subsequent ability to conduct dorsiflexion of the hemiplegic affected ankle can be quantified through a smartphone functioning as a wearable and wireless gyroscope platform. The gyroscope signal data can be consolidated to a feature set for machine learning classification, and a multilayer perceptron neural network can be applied. Considerable classification accuracy has been attained to distinguish between an assortment of ankle stretch durations through a wedge board for hemiplegic affected ankle dorsiflexion quantified by a smartphone serving as a wearable and wireless gyroscope platform. The implications of the amalgamation of wearable and wireless inertial sensor systems with machine learning and the synergy with Cloud computing resources infer a highly interactive therapeutic intervention experience with significantly augmented clinical situational awareness for improved determination of therapy efficacy.