Comparative analysis of quantitative fundus results of the left and right eyes and screening for coronary atherosclerotic disease: A binocular study

Background Fundus imaging of the retina, characterized by its non-invasive nature and high information content, has garnered widespread attention as a potential biological window to reflect systemic vascular health and has gradually become an important tool in cardiovascular disease research. At present, non-invasive fundus photography and deep learning techniques for the quantitative measurement of retinal microvascular parameters are rapidly developing. The aim of this study was to assess the consistency and differences in retinal vascular characteristics between the two eyes using quantitative analysis of fundus retinal vessels. Methods This study was based on the database of the Cardiovascular Center of Beijing Tongren Hospital, Capital Medical University. For patients with coronary atherosclerotic disease, color fundus photographs (CFPs) were captured using a non-mydriatic 45-degree fundus camera. A vascular segmentation model was constructed using a deep learning-based semantic segmentation network, ResNet 101-UNet, for fully automated quantitative measurement of retinal vascular parameters on CFPs. Data obtained included fractal dimension (FD), vessel density (VD), mean branch angle (MBA), mean blood vessel diameter (MBVD), mean arterial diameter (MAD), mean venous diameter (MVD), and arteriovenous ratio (AVR). Retinal parameters of both eyes in the aforementioned population were compared based on coronary angiography results and fundus quantitative results to explore the symmetry characteristics of bilateral fundus data. Results A total of 862 CFPs from both eyes of 431 patients were included in this study. In terms of quantitative results of fundus retinal vessel morphology, there were no statistically significant differences between the two eyes in the overall population. Based on coronary angiography results, the population was divided into four groups according to Gensini score (0, 1-30, 31-60, and >60), into CAD and Non-CAD groups according to the presence of coronary atherosclerotic heart disease (CAD), and into left coronary artery group (left main artery, left anterior descending artery, left circumflex artery) and right coronary artery group according to the location of the most severely affected vessel in coronary artery disease, making a total of three different groupings. Results showed that there were no statistically significant differences between the two eyes in all three groupings (P > 0.05). The Pearson correlation coefficient (0.387-0.833) and intraclass correlation coefficient (ICC) (0.384-0.824) for all variables between the two eyes demonstrated good consistency and correlation. In the linear regression analysis of interocular differences with age and gender, regarding gender (defined as 0 for females and 1 for males), ΔVD (β=0.354, 95% CI: 0.055-0.652, P=0.02) was significantly correlated with gender. After adjusting for past medical history, personal history, and some glycolipid-related variables, we found that ΔMBA (β=-3.294, 95% CI: -6.285, -0.302, P=0.03) was significantly correlated with gender. Finally, we conducted a generalized linear mixed model analysis for different fundus quantitative results in different groups, adjusting for age, gender, past medical history, personal history, and some glycolipid-related variables. Results indicated that the P-values for multiple fundus variables in different groups were all greater than 0.05, and there was insufficient statistical evidence to support significant interocular differences in these fundus indicators within the same group across different subgroups. Conclusion Our study results are derived from an artificial intelligence-based fully automated retinal vascular parameter measurement method. By exploring the interocular differences in the aforementioned fundus quantitative results among patients with coronary atherosclerotic disease, we concluded that the differences in the fundus indicators provided by CFP between the two eyes are not statistically significant overall. Such studies may potentially use indicators from either eye to represent the overall condition of both eyes for research purposes.