UPFP-SG: A New Benchmark for Unilateral Peripheral Facial Paralysis Severity Grading.

Area of Science:

• Medical Imaging and Computer Vision
• Neurology
• Biomedical Engineering

Background:

• Unilateral facial palsy significantly impairs daily life and quality of life.
• Current diagnosis relies on subjective clinical judgment, lacking objective quantification.
• Limited public datasets and inability to analyze distinct facial nerve branches hinder automated detection.

Purpose of the Study:

• To introduce a novel benchmark, UPFG-SG (Unilateral Peripheral Facial Paralysis Severity Grading), for objective assessment.
• To develop an automated method for grading facial palsy severity across different facial nerve branches.
• To address the limitations of subjective diagnosis and data scarcity in facial paralysis research.

Main Methods:

• Establishment of the UPFG-SG dataset with an enhanced subjective evaluation system for peripheral facial nerve branches.
• Development of a novel deep learning method integrating diverse facial features for palsy severity rating.
• Implementation of an enhanced regression module to improve the accuracy of severity evaluation.

Main Results:

• The proposed method effectively captures subtle facial expression changes and fine local details.
• Experimental results on the UPFG-SG dataset show superior performance compared to existing deep learning methods.
• The developed benchmark and method provide a more objective and accurate approach to facial palsy grading.

Conclusions:

• The UPFG-SG benchmark and associated automated grading method offer a significant advancement in assessing unilateral peripheral facial palsy.
• This work provides a valuable resource for researchers and clinicians, facilitating more objective and reliable diagnosis.
• The proposed approach has the potential to improve patient care and outcomes for individuals with facial paralysis.