Título

USING ARTIFICIAL INTELIGENCE TO CORRELATE MIDDLE RETINAL LAYERS STRUCTURAL CHANGES AND DEEP CAPILLARY PLEXUS DROPOUT IN DIABETIC RETINOPA

Objetivo

To compare areas of deep capillary plexus dropout to structural changes in the middle retinal layers in diabetic retinopathy (DR) patients utilizing swept source optical coherence tomography angiography (SSOCT-A).

Método

Patients with diabetes and DR of different severity levels as well as healthy controls were included. Cross sectional and en face structural swept source OCT (ssOCT) and en face 6x6 mm ssOCT-A imaging were performed with the Plex Elite 9000 © (Carl Zeiss Meditec, Inc). Proprietary Carl Zeiss automated software was utilized to segment the superficial vascular complex (SVC) and deep vascular complex (DVC). The number and area of capillary dropout (CDO) zones were calculated by using ImageJ software v 1.51. Analysis of variance was used to compare capillary dropout of different areas and DR severity levels (GraphPad Prism 6.0). The number and area of ssOCT middle retina structural changes were measure on ImageJ and compared to CDO zones.

Resultado

38 eyes of 24 patients and 30 eyes of 15 controls were included in the study. CDO in the DVC colocalized with middle retina layer thinning seen on en face ssOCT. The number and area of middle retina layer thinning increased with DR severity. The area of CDO in the DVC also correlated with DR severity (0.10% ± 0.18, 9.57% ± 8.60, 15.53% ± 8.93, 27.05% ±13.50 in controls, moderate, severe and proliferative DR, respectively; p<0.0001.) There was a significant correlation between area of CDO in the DVC and area of middle retina layer thinning seen on structural em face ssOCT.

Conclusão

En face ssOCTA can detected DVC changes before structural changes area seen in the middle retinal layers. A minimum area of CDO is required for structural changes to develop. The presence of middle retinal layer thinning on ssOCT is a potential new biomarker of disease severity in DR patients.