Deep learning model to identify angle dysgenesis on ASOCT images

Anterior segment spectral domain–optical coherence tomography (ASOCT) is increasingly used in glaucoma patients, primarily to investigate the anterior chamber angle and visualize the trabecular meshwork (TM) and Schlemm's Canal (SC) in‐vivo. The anatomical variability of the drainage angle, TM and SC morphology makes it difficult to distinguish normal from abnormal. Therefore, interpreting ASOCT images requires expertise and a deep understanding of the complexity involved in the developmental anomalies of the outflow pathways. While the ASOCT scans can successfully capture the gross changes, they may fail to identify angle dysgenesis in cases where altered extracellular matrix anomalies are subtle enough to get unnoticed by the human eye of an expert. Deep learning (DL) has demonstrated convincing results and potential in classifying ophthalmic diseases. The authors developed DL models, which can identify obscure features of angle dysgenesis in ASOCT scans (ADoA) and validated its performance in eyes with primary open-angle glaucoma. The models performed well on the internal as well as external validation datasets. The model performance on the multiple independent datasets highlights the translational relevance of image-based diagnostic alternatives in clinical settings, as it has the potential to be deployed for screening patients and their family members for the detection of angle dysgenesis in vivo.