Building a Machine-Learning Framework to Remotely Assess Parkinson's Disease Using Smartphones

Area of Science:

• Neurology
• Biomedical Engineering
• Data Science

Background:

• Parkinson's disease (PD) assessment traditionally relies on infrequent clinical evaluations.
• Remote patient monitoring using smartphones offers continuous, objective behavioral data collection.
• Smartphone data can capture subtle motor and vocal changes associated with PD.

Purpose of the Study:

• To develop and validate a machine-learning framework for automated Parkinson's disease assessment using smartphone sensor data.
• To leverage cross-modality features from smartphone data for objective PD evaluation.
• To explore the potential of remote monitoring for tracking disease fluctuations and severity.

Main Methods:

• A machine-learning framework employing a two-step feature selection and an elastic-net regularized model was developed.
• Smartphone sensor data (balance, dexterity, gait, tremor, voice) were collected from Parkinson's disease patients and healthy controls.
• Behavioral features were extracted and analyzed to create PD-specific behavioral atlases.

Main Results:

• The framework successfully discriminated between individuals with Parkinson's disease and healthy controls.
• The system demonstrated the ability to estimate the disease severity in individuals with PD.
• Analysis of 437 behavioral features from 72 subjects over six months showed promising results.

Conclusions:

• The proposed machine-learning framework shows significant potential for analyzing remotely collected smartphone sensor data in Parkinson's disease.
• This approach offers a scalable and objective method for Parkinson's disease monitoring and assessment.
• Remote smartphone monitoring can complement traditional clinical assessments, providing valuable insights into disease progression.