Deep learning assisted retinal microvasculature assessment and cerebral small vessel disease in Fabry disease.

Purpose: The aim of this study was to assess retinal microvascular parameters (RMPs) in Fabry disease (FD) using deep learning, and analyze the correlation with brain lesions related to cerebral small vessel disease (CSVD).

Methods: In this retrospective case control study, fundus images from 27 FD patients and 27 age- and sex-matched healthy subjects were collected. RMPs, encompassing diameter, density, symmetry, bifurcation, and tortuosity, were quantified. Laboratory examination results, Mainz severity score index (MSSI) scores, and a brain magnetic resonance imaging scan for CSVD scores were extracted and their relationships with RMPs was analyzed.

Results: Utilizing artificial intelligence-assisted analysis, compared with controls, FD patients exhibited reduced diameter (p = 0.001 for central retinal artery equivalent, p = 0.049 for central retinal vein equivalent), density (p < 0.001 for vessel area density, p = 0.001 for length density), fractal dimension (p < 0.001), and heightened arteriolar and venular asymmetry ratios (p = 0.002 and p = 0.037, respectively), venular curvature tortuosity (p = 0.037), and simple tortuosity (p = 0.037) in retinal microvascular networks. Gender-based differences in RMPs were observed among FD patients. Furthermore, RMPs were significantly associated with disease markers such as plasma globotriaosylsphingosine and α-galactosidase A activity, as well as MSSI scores. Notably, there was a significant negative correlation between the arteriolar asymmetry ratio and CSVD-related scores (age-related white matter changes: r =  - 0.683, p = 0.001; Fazekas: r =  - 0.673, p = 0.001; Lacuna: r =  - 0.453, p = 0.045; small vessel diseases: r =  - 0.721, p = 0.012; global cortical atrophy: r =  - 0.582, p = 0.009).

Conclusions: Fabry disease patients demonstrated increased vascular tortuosity and asymmetry, reduced density and diameter, and a simpler fractal dimension in retinal microvasculature. These microvascular characteristics may serve as preliminary indicators for assessing brain lesions and could represent potential novel biomarkers for CSVD, aiding in the monitoring of FD severity and progression.