Vertical Beam Segmentation in Mobile Slit Lamp Images for Peripheral Anterior Chamber Angle Measurement

The accurate segmentation of anterior chamber illumination in slit lamp images is crucial for calculating the anterior chamber angle (ACA), a vital parameter in diagnosing glaucoma and other anterior segment disorders. However, segmentation is complicated by challenges such as specular reflections, non-uniform illumination, and inconsistent image quality. This study evaluates and compares the performance of conventional image processing methods, specifically Otsu’ s multi-thresholding, and a deep learning-based UN et architecture for semantic segmentation. A robust preprocessing pipeline was developed to address artifacts and enhance vertical beam segmentation accuracy for peripheral ACA measurement on mobile slit lamp images. Experimental results demonstrate that Otsu's multi-thresholding achieves more refined vertical beam with a sensitivity of 0.8565, IoU of 0.6260, and DSC of 0.7603, outperforming the UNet-based approach (sensitivity: 0.8161, IoU: 0.5432, DSC: 0.6915). This finding highlights the advantages and limitations of both approaches, with optimized preprocessing playing a key role in improving ACA measurement. This research provides insights into the practical integration of segmentation techniques into clinical workflows, bridging the gap between advanced computational methods and their real-world application in ophthalmology.