Publication

Parkinson’s Disease (PD) is a neurodegenerative disorder characterized by the degeneration of dopamine neurons, leading to various motor symptoms, including excessive muscle tone. Understanding the mechanisms of postural control is essential, yet previous studies involving postural control models for PD patients have not incorporated brain dopamine information. This paper aims to establish a relationship between the state of dopamine and muscle tone parameters using a machine learning model, with a specific focus on muscle tone critical for maintaining an upright stance. In this study, we analyzed data from eight PD patients who underwent Dopamine Transporter Single-Photon Emission Computed Tomography (DAT-SPECT) and motor function evaluations, particularly during a quiet stance. A neural network model with fully connected layers was utilized to input DAT-SPECT images and estimate the magnitude squared of the muscle tone control parameter vector. The optimization and validation of the model were conducted using a double-cross validation approach, focusing on architecture optimization. The learning outcomes were visualized using Gradient-weighted Class Activation Mapping. The optimized model consisted of three layers with two nodes, achieving an average mean squared error of 2.30. Visualizations prominently highlighted areas around the striatum. This study underscores the potential of using DAT-SPECT images to estimate neural controller model parameters in PD patients.