Deep learning-based quantitative assessment inflammatory response of hyperreflective foci in diabetic macular edema

To investigate the associations between inflammatory cytokines (IL-6, ICAM-1, and VEGF), and hyperreflective foci (HRF) on spectral-domain optical coherence tomography (SD-OCT) in diabetic macular edema (DME).

Thirty-nine eyes from 29 patients with DME and 19 eyes from 19 control subjects were studied. Aqueous humor samples were measured using ELISA before anti-vascular endothelial growth factor therapy or cataract surgery. HRF was measured by automatic segmentation system developed in previous research biometry.

The levels of IL-6 ICAM-1, and VEGF were signifcantly higher than in the control subjects (P = 0.046, P = 0.015, P = 0.035, respectively). Spearman correlation analysis showed that the contents of IL-6, ICAM-1 and VEGF in aqueous humor of DME patients were positively correlated with the number of HRF in the whole retina level (r = 0.633, P = 0.002; r = 0.736, P = 0.003; r = 0.622, P = 0.004, respectively). Also, there was a correlation between the levels of IL-6, ICAM-1 in the aqueous humor and the volume of HRF (r = 0.601, P = 0.003; r = 0.719, P = 0.004), while no significant correlation between VEGF content in aqueous humor and HRF volume (r = 0.090, P = 0.716). All cytokines did not significantly correlate with distribution of HRF.

DME exhibit higher inflammatory cytokines levels in the aqueous humor than control group, indicating severe inflammation. The number and volume of HRF are closely related to the content of intraocular inflammatory cytokines. Thus, HRF characteristics to some extent refect intraocular infammatory factors and may guide treatment alternatives.

Assessement biomarkers of retinal inflammation and treatment using Deep Learning and OCT contributes to management and prediction of the treatment effect of DME.