Abstract

Parkinson's Disease (PD), a progressive neurodegenerative disorder, impacts approximately 1% of the elderly population worldwide. The severity assessment is crucial to plan medication and treatment effectively. Observing and assessing whether a tremor is associated with Parkinson's disease is a key area of both research and clinical practice. Normally, doctors, especially neurologists, use their observations and assessments to diagnose Parkinson's disease. These assessments can be subjective and might lead to mistakes, especially in the early stages. In this work, a novel framework is proposed to assess and evaluate PD severity from hand tremor videos with good accuracy. A custom 1D convolutional neural network with LSTM architecture is devised for temporal motion analysis. The results demonstrated over 73% accuracy with the 1DCNN-LSTM algorithm to differentiate PD severity levels. This tool offered objective, quantifiable assessments of disease severity, ideally suited for early detection, especially for remotely or underserved populations.