Towards a Reliable and Rapid Automated Grading System in Facial Palsy Patients: Facial Palsy Surgery Meets Computer...

Area of Science:

• Medical imaging analysis
• Artificial intelligence in healthcare
• Neurosurgery and reconstructive surgery

Background:

• Facial palsy (FP) management requires reliable, cost-effective, and clinician-friendly diagnostic tools.
• Existing automated FP grading systems often suffer from insufficient accuracy and high hardware costs.
• This study aimed to develop an improved automated facial palsy grading system.

Purpose of the Study:

• To develop and evaluate an automated facial palsy grading system.
• To utilize the House and Brackmann scale (HBS) for objective FP assessment.
• To overcome limitations of previous automated grading systems.

Main Methods:

• A neural network was trained and validated using image datasets from 86 facial palsy patients.
• The dataset was collected between June 2017 and May 2021 at the University Hospital Regensburg.
• The algorithm analyzed nine facial poses per patient for grading.

Main Results:

• The automated grading system achieved 100% accuracy in classifying facial palsy severity.
• The direct classification form (100%) outperformed the modular form (99%).
• Early Fusion techniques yielded superior accuracy (100%) compared to Late Fusion (96%) and sequential methods (97%).

Conclusions:

• The developed automated FP grading system demonstrates high accuracy and cost-effectiveness.
• This tool has the potential to expedite the grading process for facial palsy patients.
• The algorithm can significantly facilitate the workflow of facial palsy surgeons.