Analysis of infrared images for meibomian glands assessment in dry eye syndrome classification

Meibomian Gland Dysfunction (MGD) is one of the causes of evaporative Dry Eye Syndrome. Traditional clinical evaluation is subjective and affected by high inter-operator variability. Consequently, modern ophthalmology utilizes automated infrared meibography to objectively assess glandular morphology and quantify tissue atrophy. This thesis presents a rigorous verification of the computational scoring logic underlying a diagnostic meibography module. The first phase of the study focused on the systematic verification of the software. This involved a comprehensive review of requirements and rigorous anomaly tracking to ensure the structural reliability of the diagnostic platform. Subsequently, the software’s automated scores were compared against an expertly annotated clinical ground truth. To ensure methodological rigor, the test sample was filtered using stringent criteria, selecting only lower eyelid acquisitions with optimal eversion. The computational investigation addressed a spatial anomaly within the region of interest. A targeted refinement of the segmentation mask was performed, confirming that spatial adjustment is mandatory for objective parameter extraction. Beyond spatial correction, statistical analysis evaluated the significance of discrepancies in severity grading. The comparative analysis highlighted a systematic deviation in the classification logic. Specifically, the algorithm exhibited a computational tendency to converge towards an intermediate severity score. These findings indicate that the primary limitation resides within the edge-detection mechanisms, establishing an objective baseline for future developmental phases.