Predicting Choroidal Vascular Abnormalities in High Myopia with Longitudinal OCTA Analysis and Data-Driven Modeling

Abstract

High myopia is prone to induce choroidal vascular lesions, which severely affect visual acuity. The aim of this research is to achieve effective prediction of choroidal vascular lesions in high myopia through longitudinal optical coherence tomography angiography (OCTA) analysis and data-driven modeling. During the research process, a large amount of longitudinal OCTA image data of high-myopia patients was collected. Firstly, image pre-processing algorithms were utilized to perform operations such as denoising and enhancement on the original images to improve the image quality. Subsequently, the Convolutional Neural Network (CNN) in deep-learning algorithms was adopted to extract and analyze image attributes, and to excavate key attributes such as the morphology and structure of choroidal blood vessels. On this basis, combined with time-series analysis methods, a data-driven model was constructed. The experimental results demonstrate that this model can accurately predict the occurrence and development of choroidal vascular lesions in high myopia, providing strong support for early clinical intervention.