AI Based Early Detection of MotorDisorder Symptoms in Parkinson’sDisease using Tremor and Gait Patterns

Abstract Movement disorders due to diseases like Parkinson’s negatively affect any motor activity thereby causing symptoms including action tremor and changes in gait. The tracking of these symptoms is crucial for an early detection of the worsening condition and for appropriate changes in the medications. This study explores the potential of digital biomarkers to track motor symptoms in Parkinson’s Disease (PD), using two primary assessment tools: the spiral drawing test using a digitized writing tablet and the gait analysis using Inertial Measurement Units (IMU) sensors. The spiral test is used to measure fine motor movement and properly quantify the intensity of the tremors and motor coordination. At the same time, spatiotemporal gait characteristics and gait symmetry data from IMU sensors characterize movement disorders whilst performing an activity such as walking. For a quantitative assessment and tracking of these biomarkers is done using various machine learning and deep learning classification algorithms. The best performing models, Swin Transformer (90.9\%) and MLP with residual connections (83.9\%) for the digital spiral drawing and IMU-based gait measurements, respectively, show that they form an accurate and trustworthy method for measuring Parkinson’s disease. This gives a strong platform for disease evaluation improvement and optimization of treatment methods.