AiCarePWP: Deep learning-based novel research for Freezing of Gait forecasting in Parkinson
Hemant Ghayvat,Muhammad Awais,Rebakah Geddam,Muhammad Ahmed Khan,Lewis Nkenyereye,Giancarlo Fortino,Kapal Dev
DOI: https://doi.org/10.1016/j.cmpb.2024.108254
Abstract:Background and objectives: Episodes of Freezing of Gait (FoG) are among the most debilitating motor symptoms of Parkinson's Disease (PD), leading to falls and significantly impacting patients' quality of life. Accurate assessment of FoG by neurologists provides crucial insights into patients' conditions and disease symptoms. This proposed strategy involves utilizing a Weighted Fuzzy Logic Controller, Kalman Filter, and Kaiser-Meyer-Olkin test to detect the gait parameters while walking, resting, and standing phases. Parameters such as neuromodulation format, intensity, duration, frequency, and velocity are computed to pre-empt freezing episodes, thus aiding their prevention. Method: The AiCarePWP is a wearable electronics device designed to identify instances when a patient is on the brink of experiencing a freezing episode and subsequently deliver a brief electrical impulse to the patient's shank muscles to stimulate movement. The AiCarePWP wearable device aims to identify impending freezing episodes in PD patients and deliver brief electrical impulses to stimulate movement. The study validates this innovative approach using plantar insoles with a 3D accelerometer and electrical stimulator, analysing data from the inertial measuring unit and plantar-pressure foot data to detect and predict FoG. Results: Using a Convolutional Neural Network-based model, the study evaluated 47 gait features for their ability to differentiate resting, standing, and walking conditions. Variable selection was based on sensitivity, specificity, and overall accuracy, followed by Principal Component Analysis and Varimax rotation to extract and interpret factors. Factors with eigenvalues exceeding 1.0 were retained, and 37 features were retained. Conclusion: This study validates CNN's effectiveness in detecting FoG during various activities. It introduces a novel cueing method using electrical stimulation, which improves gait function and reduces FoG incidence in PD patients. Trustworthy wearable devices, based on Artificial Intelligence of Things (AIoT) and Artificial Intelligence of Medical Things (AIoMT), have been developed to support such interventions.