Diagnosing Parkinson Disease Through Facial Expression Recognition: Video Analysis

Area of Science:

• Neurology
• Artificial Intelligence
• Digital Health

Background:

• Increasing prevalence of neurological diseases like Parkinson disease (PD) necessitates innovative diagnostic approaches.
• Digital health technologies offer new avenues for patient consultation, disease tracking, and rehabilitation.
• Facial expression recognition presents a novel method for diagnosing neurological conditions.

Purpose of the Study:

• To investigate the efficacy of artificial intelligence (AI)-driven facial expression recognition for diagnosing Parkinson disease (PD).
• To develop and validate computational models for objective PD assessment using facial cues.

Main Methods:

• Collected facial expression videos from individuals with PD and healthy controls.
• Extracted facial expression features (amplitude, muscle group shaking) using key point analysis from Face++.
• Applied traditional machine learning (SVM) and deep learning (LSTM) algorithms for PD diagnosis.

Main Results:

• AI models demonstrated significant capability in facial expression recognition for PD diagnosis.
• A long short-term model (LSTM) achieved 86% precision and 75% F1 score.
• A support vector machine (SVM) model reached a 99% F1 score using specific facial features.

Conclusions:

• AI-based facial expression analysis is a viable tool for digital diagnosis of Parkinson disease.
• Validated diagnostic models can aid clinicians in real-time disease monitoring and remote assessment.
• This approach offers potential for more accessible and efficient PD diagnosis and management.