A Comprehensive Review of Deep Learning Techniques for Identifying Parkinson's Disease from Gait Analysis

Abstract

A group of neurological disorders that manifests in a wide range of motor and cognitive symptoms, Parkinson's Disease (PD) ranks second among those affecting older adults. Many people mistakenly believe that PD symptoms are caused by other conditions, such as essential tremor or natural aging. There may not be a cure for PD, but there are numerous medications that can help with symptoms. Since gait impairment is a prominent and early sign of PD in a clinical context, doctors typically use visual observations to analyse gait abnormalities as one of many manifestations to determine the severity of PD. Nevertheless, therapists' reliance on their own knowledge and judgment makes gait examination a difficult and subjective process. Extracting useful gait features from various input signals for gait analysis has recently been used Deep Learning (DL) models. Intermediate Medical Unit (IMU) patients with movement problems may benefit from a quick and clinically relevant evaluation of gait abnormalities performed automatically utilizing DL algorithms. Using gait analysis, this study provides a comprehensive overview of various DL frameworks that have been created for the purpose of PD prediction and categorization. At the outset, In the first part of the study, various PD prediction and classification systems that have been developed by various researchers and that rely on DL algorithms via gait analysis are reviewed briefly. To find a better way to identify and categorize the PD, we next undertake a comparative analysis to learn about the shortcomings of those algorithms.