Artificial Intelligence Assisted Optical Coherence Tomography for Differential Diagnosis and Management of Irregular Corneas

Oregon Health & Science University, Portland OR

Investigators

Linked publications & trials

Abstract

PROJECT SUMMARY To see well, the cornea must maintain near-perfect clarity and shape. Several disease processes can distort the corneal shape and degrade vision. These include ectasia/keratoconus (stromal thinning and bulging), edema (corneal swelling due to endothelium dysfunction), primary epithelial deformation (contact lens-related warpage, dry eye, epithelial basement membrane dystrophy), stromal opacities (scar, ulcer, stromal dystrophy), and other stromal shape changes (corneal surgery). Some of these conditions can appear similar on standard corneal topography yet require very different treatments. Anterior eye optical coherence tomography (OCT) has the potential to be an all-in-one device that can measure corneal shape, thickness, and reflectance deviation with high resolution. Therefore, this project aims to use OCT, a 3-dimensional imaging technology with micrometer-level resolution, to differentiate between different types of corneal shape irregularities. The unique ability of OCT to map epithelial thickness, anterior & posterior topography, and the reflectance of various corneal layers will be used to develop new metrics for staging and monitoring the progression of ectasia (keratoconus) and endothelial dysfunction (Fuchs’ endothelial corneal dystrophy). Our overall goal is to improve the early detection, differential diagnosis, staging, monitoring, and treatment of irregular corneas by advancing the interpretation of anterior eye OCT with mathematical analyses and artificial intelligence (AI). The specific aims are: (1) Develop an OCT-based AI system to classify corneal irregularities. AI is needed to help the clinician detect pathologic features and synthesize the rich quantitative information that OCT provides on corneal shape, thickness, and reflectance patterns. AI-aided diagnostic classification can help the clinician make management decisions. (2) Develop OCT metrics to assess disease progression and treatment efficacy. Chronic progressive corneal diseases such as ectasia and endothelial dysfunction require precise monitoring of progression to determine whether and when treatment is needed. An assessment of treatment efficacy is needed for further management decisions. (3) Develop an OCT algorithm to predict corneal power change after endothelial keratoplasty. This is needed for the proper selection of intraocular lens power in combined endothelial keratoplasty and cataract surgery to treat both Fuchs’ endothelial corneal dystrophy and cataract, which often co-exist in older patients.