A computer vision-based framework for objective evaluation of sunken upper eyelid

Area of Science:

• Ophthalmology
• Computer Vision
• Medical Imaging

Background:

• Sunken upper eyelid correction is crucial for periorbital reconstruction.
• Current evaluation methods rely on expensive equipment or subjective assessments.
• Objective and accessible evaluation tools are needed.

Purpose of the Study:

• To develop a computer vision (CV) framework for objective assessment of sunken upper eyelid morphology.
• To provide a reliable method for evaluating surgical outcomes in periorbital reconstruction.

Main Methods:

• A two-stage CV framework was developed.
• Facial landmarks were detected to isolate the periocular region, followed by normalization and segmentation.
• Key features including Variance of Gray Value (VGV), Structural Similarity Index (SSIM), and Degree of Eyelid Wrinkles (DEW) were extracted.
• A Support Vector Machine (SVM) model integrated these features to score overall morphology.

Main Results:

• Significant differences in VGV, SSIM, and DEW were observed between normal and patient groups.
• The proposed method demonstrated measurable improvements in postoperative surgical outcomes.
• The SVM model output, L2 distance to the separating hyperplane (D(f)), effectively scored morphological features.

Conclusions:

• The CV framework offers an objective and accessible method for assessing sunken upper eyelid morphology.
• This approach can aid in evaluating the effectiveness of surgical interventions in periorbital reconstruction.
• The method shows potential for improving clinical decision-making and patient care.