Free papers detail | 11th Asia Pacific Tele Ophthalmology Society Symposium

Title
Quantitative Retinal Vascular Profiling in Pathologic Myopia via Deep Learning-Based Fundus Analysis

Authors
MO WANG, Shiheng Lu, Xiaowei Tong

Presenting
MO WANG

PURPOSE:
To characterize dynamic retinal vascular changes and identify potential staging patterns during pathologic myopia (PM) progression using deep learning-based AI fundus analysis.

METHODS:
This retrospective cross-sectional study enrolled PM patients and controls with high myopia without significant fundus lesions. The EvisionAI system quantified retinal vasculature from color fundus photographs, measuring arteriolar/venular diameters, arteriole-to-venule ratio (AVR), tortuosity, arcade vessel angle (AVA) and distance (DVA), vessel density, fractal dimension (FD), venular length, and peripapillary atrophy (PPA) area. Analyses compared overall, gender-specific, and under-25 subgroups. Univariate and multivariate linear regression assessed vascular parameter correlations with PPA width.

METHODS:

This retrospective cross-sectional study enrolled PM patients and controls with high myopia without significant fundus lesions. The EvisionAI system quantified retinal vasculature from color fundus photographs, measuring arteriolar/venular diameters, arteriole-to-venule ratio (AVR), tortuosity, arcade vessel angle (AVA) and distance (DVA), vessel density, fractal dimension (FD), venular length, and peripapillary atrophy (PPA) area. Analyses compared overall, gender-specific, and under-25 subgroups. Univariate and multivariate linear regression assessed vascular parameter correlations with PPA width.

RESULTS:
Analysis included 671 fundus images from 275 patients (mean age 41.2±18.6 years; 61.2% female). Vascular parameters differed significantly between genders. After adjusting for age and sex, multivariate analysis revealed that the increased PPA width ratio was significantly associated with reductions in AVA (1PD), DVA (1PD), vessel caliber (0.5-1.0 PD), tortuosity (0.5-1.0 PD), vessel density, and FD (all P<0.001). The evolution of vascular morphology followed a biphasic pattern: an initial phase of rapid decline in tortuosity and AVA was followed by a phase of slower change, whereas vessel density and FD exhibited an opposite trend—declining slowly at first, then more rapidly.

RESULTS:

Analysis included 671 fundus images from 275 patients (mean age 41.2±18.6 years; 61.2% female). Vascular parameters differed significantly between genders. After adjusting for age and sex, multivariate analysis revealed that the increased PPA width ratio was significantly associated with reductions in AVA (1PD), DVA (1PD), vessel caliber (0.5-1.0 PD), tortuosity (0.5-1.0 PD), vessel density, and FD (all P<0.001). The evolution of vascular morphology followed a biphasic pattern: an initial phase of rapid decline in tortuosity and AVA was followed by a phase of slower change, whereas vessel density and FD exhibited an opposite trend—declining slowly at first, then more rapidly.

CONCLUSIONS:
This study delineates two distinct stages of retinal vascular alteration in PM progression, suggesting an initial phase dominated by tractional forces followed by a later atrophic phase. These findings advance the understanding of PM pathogenesis from a vascular perspective and highlight the utility of AI-based quantitative fundus analysis.