Parkinson's disease severity clustering based on tapping activity on mobile device.

Area of Science:

• Neurology
• Digital Health
• Biomedical Data Science

Background:

• Parkinson's disease (PD) severity assessment relies on clinical observation, which can be subjective.
• Mobile health studies offer large-scale data collection for objective disease monitoring.
• The mPower dataset provides extensive data on PD indicators from a large participant pool.

Purpose of the Study:

• To investigate the correlation between smartphone finger-tapping task performance and Parkinson's Disease Rating Scale (MDS-UPDRS) I-II and PDQ-8 scores.
• To explore the potential of using quantitative finger-tapping features for objective PD severity stratification.
• To assess the feasibility of digital health tools in aiding PD management.

Main Methods:

• Utilized the mPower dataset, selecting 1851 participants with both tapping activity and MDS-UPDRS I-II data.
• Extracted nine features from smartphone finger-tapping tasks for analysis.
• Applied K-means unsupervised clustering, validated by elbow method, Silhouette score, and Davies-Bouldin index, to determine three distinct clusters.

Main Results:

• Statistical analysis demonstrated that finger-tapping features successfully segregated participants into three distinct severity groups.
• These identified groups exhibited significant differences in MDS-UPDRS I-II and PDQ-8 scores, correlating with PD severity.
• The findings suggest tapping metrics can reflect varying degrees of Parkinson's disease progression.

Conclusions:

• Smartphone-based finger-tapping analysis offers a quantitative method for assessing Parkinson's disease severity.
• This digital approach could supplement traditional clinical assessments, reducing subjectivity and improving accessibility.
• Developing objective, non-invasive tools is crucial for efficient and scalable Parkinson's disease management.