A modular and adaptable approach for automated morphological feature extraction in meibography images.

This study presents a modular and adaptable approach for the automated extraction of morphological features from meibography images, focusing on Meibomian gland (MG) analysis. The proposed method leverages piecewise linear modeling to derive clinically interpretable metrics that capture key structural characteristics of MGs. The workflow consists of three main stages: (1) semi-automated region of interest (ROI) selection, (2) MG identification and segmentation, and (3) extraction of gland- and image-level metrics. The approach was validated using 616 meibography images from two different imaging systems, demonstrating robustness, adaptability, and high classification accuracy for Meiboscale grading. Key metrics such as the shortening ratio and dropout area proved effective in distinguishing different stages of Meibomian gland dysfunction (MGD). By balancing automation, interpretability, and computational efficiency, this method provides a practical and scalable tool for the objective assessment of MG morphology, with potential applications in clinical practice and large-scale ophthalmic research.