A Quantitative Evaluation Method for Parkinson's Disease

Objective, quantifiable, and easy to operate evaluation methods are crucial for assisting in the diagnosis of Parkinson's disease. In the widely used Unified Parkinson's Disease Rating Scale (UPDRS) III, item 3.12 (postural stability) is used to evaluate the patient's body balance. But in actual clinical scenarios, doctors' judgments are subjective and have high internal variability. We propose a method based on monocular vision to objectively evaluate patients' body balance. Firstly, we use a combination of deep algorithms to obtain joint point sequences of patients and doctors in the video. Then, we use these sequences to further extract three features to evaluate the patient's balance, which are the patient's backward steps, the patient's body tilt, and the patient doctor's body distance. We collected and tested 23 Parkinson's disease patients. The experimental results indicate that our proposed method provides valuable and quantifiable patient posture and motion data in clinical settings, and to some extent reduces the subjective judgment of doctors and increases the accuracy of diagnosis.