Deep learning model shows high accuracy for detecting corneal perforation on ASOCT images.

Photo by Mohammed Mahjoub Kikha / Unsplash

medRxiv Published April 23, 2026 Medically reviewed April 24, 2026 Study authors: Rhode, L.; Reddy, K. N.; Ibukun, F.; Kuyyadiyil, S.; Jain, E.; Parmar, G. S.; Chellappa, R.; Shekhaw… DOI ↗ By Dr. Lars van Dijk, PhD · Surgical, Procedural & Diagnostic

Key Takeaway

Consider this deep learning model as a potential decision support tool for corneal perforation detection, pending external validation.

This is a diagnostic accuracy study evaluating four convolutional neural network models (ResNet architecture) trained on anterior segment optical coherence tomography (ASOCT) images to detect corneal perforation in microbial keratitis. The study included 150 patients with microbiologically confirmed keratitis, using contralateral healthy eyes as controls.

The best-performing model (Model 1) achieved an AUC of 0.965 (95% CI, 0.911-0.995) for detecting corneal perforation. It demonstrated a sensitivity of 84.0% (95% CI, 70.0%-97.1%) and a specificity of 97.8% (95% CI, 96.1%-99.3%). These results suggest the model can accurately classify the presence or absence of corneal perforation on ASOCT imaging.

The authors note that external validation of the model is not reported, limiting generalizability. The study does not report follow-up duration, safety data, or adverse events. Practice relevance is framed as potential support for automated ASOCT analysis as a clinical decision aid, not as a replacement for clinical judgment.

Findings are based on a single study with 150 patients. The model's performance requires further validation before clinical implementation. Causation between deep learning and improved patient outcomes cannot be inferred from this diagnostic accuracy study.

Study Details

Sample sizen = 150

EvidenceLevel 5

PublishedApr 2026

View Original Abstract ↓

Purpose: To develop and evaluate deep learning models for automated detection of corneal perforation in microbial keratitis using anterior segment optical coherence tomography (ASOCT) images. Methods: We enrolled 150 patients with microbiologically confirmed keratitis. Contralateral healthy eyes served as controls. Four convolutional neural network models using ResNet architecture were trained and evaluated using ASOCT images to classify the presence or absence of corneal perforation at the eye level. Ground truth labels for perforation were established following consensus grading by two masked ophthalmologist graders. Models differed in inclusion of healthy controls and masking of non-corneal anterior segment anatomy. Results: The best-performing model (Model 1), which included healthy controls and randomly applied masking of the inferior image portion during training, achieved an AUC of 0.965 (95% CI, 0.911-0.995), sensitivity of 84.0% (95% CI, 70.0%-97.1%), and specificity of 97.8% (95% CI, 96.1%-99.3%) for detection of corneal perforation. Models including healthy controls outperformed those without, and lens masking improved discrimination. Conclusions: Deep learning models achieved high diagnostic accuracy for detecting corneal perforation on ASOCT imaging in eyes with microbial keratitis. These findings support the potential role of automated ASOCT analysis as a clinical decision support tool for identifying this vision-threatening complication.