Video Assessment to Detect Amyotrophic Lateral Sclerosis

Area of Science:

• Neurology
• Computer Science
• Biomedical Engineering

Background:

• Facial muscle weakness is an early symptom of amyotrophic lateral sclerosis (ALS).
• Current diagnostic methods relying on visual assessment of facial expressions can be subjective due to individual variations.
• Objective, quantitative methods are needed for early and accurate ALS detection.

Purpose of the Study:

• To investigate the efficacy of computerized facial expression analysis in differentiating ALS patients from healthy individuals.
• To identify specific facial action units and expressions most indicative of ALS-related facial weakness.

Main Methods:

• Utilized the Toronto NeuroFace Dataset containing videos of nine facial expression tasks from ALS patients and healthy controls.
• Analyzed facial action units extracted from these videos to build classification models.
• Evaluated the performance of the models in distinguishing between the two groups.

Main Results:

• The study achieved a peak classification accuracy of 0.91.
• The 'pretending to smile with tight lips' facial expression yielded the highest accuracy.
• Specific action units related to lip and mouth movements were key indicators.

Conclusions:

• Computerized analysis of facial expressions shows significant potential for objective ALS detection.
• This approach can help overcome the subjectivity inherent in traditional diagnostic methods.
• Further research can refine this technique for clinical application in identifying facial weakness in ALS.