Facial paralysis revisited
Reviewed by Badr Eldin Mostafa

Facial paralysis grading is a confusing issue. Lots of classifications have been proposed but all are subjective. The authors attempted to automate the Sunnybrook Facial Grading System (SFGS) by training a convoluted neural network (CNN) deep learning system. They used 116 patients with unilateral facial paralysis and nine healthy volunteers. A separate model was trained for the 13 elements of the SFGS. The integration of both models and facial landmarks increased the reliability of the system to match human observers. This makes the automated model reliable in the evaluation of unilateral facial paralysis and can be integrated in e-health systems and implemented by clinicians, untrained personnel and other healthcare providers.

Optimization of the automated Sunnybrook Facial grading system- Improving the reliability of a deep learning network with facial landmarks.
Harkel TCT, Bielevelt F, Marres HAM, et al.
EUR ANN OTORHINOLARYNGOL HEAD NECK DIS
2025;142(1):5–11.