IoT-Enabled Diagnostic Model Using Deep Learning for Early Detection of Parkinson’s Disease from Gait and Brain Imaging
Nellore Kapileswar, Judy Simon
Abstract
Early diagnosis is crucial in managing Parkinson's disease, as timely intervention can significantly reduce symptom progression and enhance patient outcomes. Parkinson's disease, also known as PD, is a common neurological disorder marked by motor and mental disorders, usually emerging approximately from the age of 50, with symptoms that includes gait related issues and speech impairments. Medication can be used to treat symptoms, but a cure is still difficult. Personal visits for medical care and surveillance are difficult for patients with Parkinson's disease (PWP) since they have speech and mobility issues. Early discovery of PD allows for treatment, allowing people to live normal lives. The necessity for early, accurate, and distant PD detection is highlighted by the world's aging population. Hence, in this research a framework is proposed to detect the PD in the early stages by using the real time gait analysis along with the neuroimaging information. In this research work, the wearable smart insoles and sensors are used to gather the information about the gait structure and the correlation between the structural and functional neuro images are found by the proposed system. To perform this, the hybrid combination is proposed with Graph Neural Networks (GNNs) that helps in extracting the features of medical images and Bidirectional Long Short-Term Memory (BiLSTM) that helps in developing the gait data model. This proposed dual framework is robust in detecting the early stage of Parkinson’s disease and the same is justified with an experimental result that ensures 94% accuracy with significant error rates. The proposed system demonstrates its superior performance over the other conventional methods for the Parkinson’s disease detection.