Facia-fix: mobile application for bell's palsy diagnosis and assessment using computer vision and deep learning

Area of Science:

• Medical Technology
• Artificial Intelligence
• Computer Vision

Background:

• Facial paralysis (FP) diagnosis is challenging due to limitations of traditional methods, including time, patient comfort, and clinician availability.
• Early and accurate FP detection is vital for timely intervention, preventing complications and improving patient outcomes.
• Advanced diagnostic tools for FP are not universally accessible to healthcare providers.

Purpose of the Study:

• To introduce Facia-Fix, a mobile application designed for diagnosing Bell's palsy using computer vision and deep learning.
• To provide real-time, objective, and non-invasive assessment of facial paralysis severity.
• To enhance the diagnostic and follow-up experience for clinicians and patients.

Main Methods:

• Developed Facia-Fix, a mobile app integrating computer vision and deep learning for facial landmark analysis.
• Trained classification algorithms on the YouTube FP (YFP) dataset, labeled using the House-Brackmann (HB) method.
• Evaluated various deep learning models including MobileNet, CNN, MLP, VGG16, and Vision Transformer for FP severity classification.

Main Results:

• The MobileNet model, utilizing transfer learning, demonstrated superior performance with high accuracy (0.9812), precision (0.9753), recall (0.9727), and F1 Score (0.974).
• Facia-Fix provides accurate, objective, non-invasive, and real-time quantitative assessment of FP severity.
• Preliminary results indicate Facia-Fix's potential to significantly improve FP diagnosis and patient follow-up.

Conclusions:

• Facia-Fix offers a novel approach to diagnosing facial paralysis using advanced deep learning.
• The application provides a reliable and accessible tool for objective, real-time assessment of FP severity.
• This technology has the potential to revolutionize FP diagnostics, benefiting both healthcare providers and patients.