Machine Learning-Based Differential Diagnosis of Parkinson's Disease Using Kinematic Feature Extraction and Selection

TL;DR

This work tackles the differential diagnosis of PD, PSP, MSA, and healthy controls using a lightweight, sensor-based ML framework grounded in finger-tapping kinematics. It introduces 18 kinematic features (including two novel Thumb-to-Index relative features) and 41 statistical features per signal, totaling 738 features, which are pruned via One-way ANOVA (p < 0.005) and Sequential Forward Floating Selection before training an SVM with LOOCV. The approach achieves 66.67% per data-point accuracy and 88.89% per-subject accuracy, with particularly strong performance for MSA and HC, demonstrating potential for rapid clinical deployment though PD/PSP differentiation remains challenging and requires more data. Overall, the combination of relative kinematic features, hierarchical feature engineering, and robust feature selection offers a practical, low-cost diagnostic tool for movement disorders with real-world clinical impact.

Abstract

Parkinson's disease (PD), the second most common neurodegenerative disorder, is characterized by dopaminergic neuron loss and the accumulation of abnormal synuclein. PD presents both motor and non-motor symptoms that progressively impair daily functioning. The severity of these symptoms is typically assessed using the MDS-UPDRS rating scale, which is subjective and dependent on the physician's experience. Additionally, PD shares symptoms with other neurodegenerative diseases, such as progressive supranuclear palsy (PSP) and multiple system atrophy (MSA), complicating accurate diagnosis. To address these diagnostic challenges, we propose a machine learning-based system for differential diagnosis of PD, PSP, MSA, and healthy controls (HC). This system utilizes a kinematic feature-based hierarchical feature extraction and selection approach. Initially, 18 kinematic features are extracted, including two newly proposed features: Thumb-to-index vector velocity and acceleration, which provide insights into motor control patterns. In addition, 41 statistical features were extracted here from each kinematic feature, including some new approaches such as Average Absolute Change, Rhythm, Amplitude, Frequency, Standard Deviation of Frequency, and Slope. Feature selection is performed using One-way ANOVA to rank features, followed by Sequential Forward Floating Selection (SFFS) to identify the most relevant ones, aiming to reduce the computational complexity. The final feature set is used for classification, achieving a classification accuracy of 66.67% for each dataset and 88.89% for each patient, with particularly high performance for the MSA and HC groups using the SVM algorithm. This system shows potential as a rapid and accurate diagnostic tool in clinical practice, though further data collection and refinement are needed to enhance its reliability.

Figures (7)

• Figure 1: Sensor Control Unit coordinate system.
• Figure 2: Example of the Dataset Signal for Individual Disease.
• Figure 3: Flowchart of the proposed ML-based method for differential diagnosis.
• Figure 4: Kinematic 18 feature for HC labels.
• Figure 5: Kinematic 18 feature for PD labels.