A smart glove to evaluate Parkinson's disease by flexible piezoelectric and inertial sensors

Abstract

Parkinson's disease (PD), to date, is widespread. It is a neurodegenerative disease that impairs the quality of life of the affected, as it is a slowly but progressively evolving disease. This paper presents a smart glove for evaluating PD patients by monitoring hand tremors and evaluating specific exercises involved by the MDS-UPDRS (Movement Disorder Society - Unified Parkinson Disease Rating Scale), enabling disease evolution assessment. The smart glove consists of a TPU flexible support, integrating two flexible MEMS piezoelectric sensors based on Aluminum Nitride and an inertial sensor to detect finger and arm movements. The smart glove integrates an electronic conditioning section for piezoelectric signals to make them suitable for the following acquisition by a microcontroller section based on nRF52840 SoC, which jointly processes the piezoelectric and inertial signals related to standard patient's hand and arm exercises (i.e., finger tapping, fist opening/closing of the hand, resting hand tremor), assigning them scores according to the MDS-UPDRS. Three embedded Machine Learning (ML) algorithms based on Neural Networks (NN) were deployed to classify piezoelectric and inertial signals. Seven individuals, six of them with diagnosed PD, were involved in developing ML models. Datasets were gathered to train and test the ML algorithms, constituted by signal samples related to three tests involved in the UPDRS scale according to PD severity. The tests demonstrated the proper operation of the proposed smart glove in tracking the movement changes induced by PD; also, the developed embedded ML algorithms showed performance in classifying hand/arm movements, reaching 95.12 %, 98.39 %, and 96.62 % for finger-tapping, hand-fist closure, and resting tremor, respectively.