Role of artificial intelligence in optical coherence tomography in myopia and pathological myopia

Myopia and pathological myopia (PM) have been recognized as one of the leading causes of visual impairment globally. Optical coherence tomography (OCT) provides high-resolution imaging of retinal and choroidal structural changes and plays an increasing role in the diagnosis and prognostication of PM and myopia-related complications. Recent advances in OCT technology have produced a potential platform for artificial intelligence (AI), particularly deep learning (DL), to enhance diagnostic accuracy and prognostic capabilities. First, AI-assisted detection of myopia based on OCT-derived biomarkers such as retinal curvature, optic nerve morphology, and inner retinal thinning have the potential to detect high myopia. However, precise refractive error estimation or differentiation of lower-grade myopia remains modest. Future integration of OCT angiography may refine the prediction of myopia progression. Second, AI may improve automated segmentation and quantification of the choroid, with DL algorithms consistently delineating choroidal boundaries and quantifying region-specific choroidal thicknesses. Recent algorithms have extended beyond basic segmentation to choroidal sublayer segmentation and calculating choroidal vascularity indices, enhancing structural characterization in myopic eyes. Third, AI methods have advanced the detection of PM-related OCT lesions, reliably identifying critical lesions including myopic traction maculopathy, myopic choroidal neovascularization, and dome-shaped macula. Recent models have also shown the ability to categorize disease severity according to validated clinical frameworks, such as the Atrophy-Traction-Neovascularization and myopic tractional maculopathy staging systems. Despite these advances, current AI methods face challenges including inconsistent OCT protocols, limited longitudinal data, inadequate external validation, and difficulties handling poor-quality scans. Addressing these limitations could facilitate clinical integration, enhancing early diagnosis, prognostication, and possibly, personalized myopia management in the future.