Research on Establishment Objective Evaluation System of Facial Paralysis Based on Facial Pattern Characteristics.

Background: Facial paralysis severely impacts patients' quality of life, yet current assessment methods remain subjective, inconsistent, and inefficient. Conventional tools like FACE-gram rely on manual facial landmark identification, which limits accuracy and reproducibility in clinical evaluations.

Methods: The authors developed a machine learning-based system that enhances the Dlib framework to enable automatic and precise detection of key facial landmarks, including eyebrows, eyes, nose, and lips. The system integrates TensorFlow for iris detection and applies algorithms such as coordinate system transformation and absolute distance calculation to convert pixel-level data into precise physical measurements, ensuring objective evaluations.

Results: The authors' system demonstrated significant improvements in accuracy and efficiency over conventional methods by automating facial landmark detection. Through providing standardized and reproducible assessments, the system establishes a foundation for advancing consistent diagnostic approaches. It also facilitates monitoring during treatment and long-term follow-up, enabling clinicians to comprehensively evaluate and manage facial paralysis across all stages of care.

Conclusions: By automating precise facial landmark detection and objective assessment, the authors' machine learning-based system addresses key limitations in current assessment tools. This innovation not only promises to standardize evaluation methods but also holds the potential to transform the clinical management of facial paralysis, ultimately improving outcomes and quality of care for affected patients.

Level of evidence: Level IV.