Deep Learning for Detecting Corneal Pathologies From Anterior Segment Optical Coherence Tomography

Background: The corneal disease accounts for a mass of cases of visual impairment and blindness. Developing an automatic corneal pathologies detection system is critical in early management and vision preservation. Methods: A deep learning model, CorNet, was developed to identify 9 categories of corneal pathologies from anterior segment optical coherence tomography (AS-OCT) images. To develop the CorNet, 48644 AS-OCT images were collected and labeled by professional ophthalmologists. We evaluated the performance of CorNet by comparing the clinical readings of 6 ophthalmologists with different seniority who analyzed two independent testing datasets from different hospitals. Findings: By classifying 9 corneal pathologies, our CorNet achieved the area under receiver operating characteristics curve (AUC) of 0.972 (95% CI, 0.961-0.980) on the QDEH testing set (1,203 images) and the AUC of 0.972 ( 0.953-0.983 ) on the SDEH testing set (483 images). The average accuracy, sensitivity, and specificity values of CorNet were 0.947, 0.917, 0.956 on the QDEH testing set and 0.949, 0.932, 0.955 on the SDEH testing set, respectively. Compared to ophthalmologists, the statistical analysis revealed that CorNet was significantly higher than residents in sensitivity and achieved comparable performance in accuracy and specificity between attending ophthalmologists and residents. Interpretation: We constructed a large-scale AS-OCT dataset and developed a deep learning model for corneal pathologies detection. The CorNet achieved high accuracy and good generalization in multicenter clinical validation. Thus, it is a promising strategy for timely screening and treatment of corneal pathologies. Funding Information: Qingdao Science and Technology Demonstration and Guidance Project, under Fund No. 20-3-4-45-nsh. Declaration of Interests: All authors certify that they have no affiliations with or involvement in any organization or entity with any financial interest (such as honoraria; educational grants; participation in speakers' bureaus; membership, employment, consultancies, stock ownership, or other equity interest; and expert testimony or patent-licensing arrangements), or non-financial interest (such as personal or professional relationships, affiliations, knowledge or beliefs) in the subject matter or materials discussed in this manuscript. Ethics Approval Statement: This study adhered to the tenets of the Declaration of Helsinki, whereas approval was issued by the institutional review board of Qingdao Eye Hospital of Shandong First Medical University (No. QYLS 2020-07).