Classification of fallers in Parkinson's disease through machine learning based feature analysis.

Area of Science:

• Neurology
• Biomedical Engineering
• Data Science

Background:

• Parkinson's disease (PD) presents heterogeneous motor and non-motor symptoms, complicating the classification of individuals prone to falls.
• Identifying reliable markers for fall risk in PD is crucial for effective intervention and patient management.

Purpose of the Study:

• To develop and validate a machine learning model for classifying fallers in Parkinson's disease.
• To identify key clinical and gait-derived markers associated with faller status in PD patients.

Main Methods:

• A machine learning model was developed integrating clinical assessments and GAITRite gait metrics from 396 participants (298 training, 98 validation).
• Fall history classified participants into PD fallers, PD non-fallers, and healthy controls.
• Feature selection utilized statistical and importance-based approaches, with seven algorithms trained; the Extra Trees classifier showed optimal performance.

Main Results:

• The Extra Trees classifier achieved high accuracy (88% internal, 89% external validation).
• Key domains predicting faller status included fear of falling (FoF), balance/gait parameters (stride length, velocity, 360° rotation), and autonomic dysfunction.
• The model demonstrated feasibility for externally validated, multidomain faller classification in PD.

Conclusions:

• Machine learning integration of clinical and gait data offers a robust method for classifying fallers in Parkinson's disease.
• Fear of falling, balance/gait metrics, and autonomic dysfunction are significant predictors of falls in PD.
• This approach supports the development of objective, data-driven tools for fall risk assessment in PD management.