Evaluation of meibomian gland dysfunction with deep learning model considering different datasets and gland morphology

Nilüfer Yesilirmak, Volkan Okbay, Yener Yesilirmak, Ömer Mustafa Bilgiç, Öykü Diribas, Onat Akça, Gozde Bozdagi Akar

Abstract: Highlights•Development of a deep learning-based framework for automated meiboscore grading, integrating image embeddings with meibomian gland (MG) morphological attributes (area, length, thickness, and tortuosity) to enhance diagnostic accuracy.•First study to combine MG morphology with deep learning for multi-dataset meiboscore evaluation, demonstrating robust performance across datasets despite domain shifts caused by different imaging devices.•AI-driven standardization to improve meibography grading consistency, reducing reliance on expert evaluation and paving the way for scalable, automated diagnostic tools in ophthalmology.

External IDs:dblp:journals/cbm/YesilirmakOYBDAA25