Machine learning for Parkinson's disease: a comprehensive review of datasets, algorithms, and challenges

Area of Science:

• Neurology
• Artificial Intelligence
• Data Science

Background:

• Parkinson's disease (PD) significantly impacts elderly health, affecting mobility and cognition.
• Current treatments are limited, necessitating advanced diagnostic tools.
• Machine learning (ML) offers potential for improved PD detection accuracy.

Purpose of the Study:

• To systematically review ML applications in Parkinson's disease diagnosis.
• To identify datasets, algorithms, and evaluation metrics used in PD diagnosis.
• To discuss challenges and future directions in ML-based PD detection.

Main Methods:

• Systematic literature review (SLR) of 133 papers (2021-April 2024).
• Classification of studies into five data categories: acoustic, biomarkers, medical imaging, movement, and multimodal.
• Analysis of ML algorithms, datasets, merits, limitations, and evaluation factors.

Main Results:

• ML effectively utilizes acoustic, biomarker, imaging, movement, and multimodal data for PD diagnosis.
• Various ML algorithms show promise, with performance varying by dataset type.
• Identified common datasets and evaluation metrics across reviewed studies.

Conclusions:

• ML shows significant potential to enhance the accuracy and efficiency of Parkinson's disease diagnosis.
• Further research is needed to address limitations and optimize ML models for clinical application.
• Multimodal data integration and advanced ML techniques are promising future directions.