OCT Hyperreflective Retinal Foci as Sign of Microglial Activation in Diabetic Retinopathy: an AI Automatic Quantification Approach

Abstract

Purpose : To validate the performance of an artificial-intelligence (AI) based quantification software of hyperreflective retinal foci (HRF), specifically defined as biomarker of microglia cells activation, in eyes affected by diabetic macular edema (DME)

Methods : After adequately training of the deep learning algorithm using normal and diabetic OCT images, a specific tool was developed for the automatic detection and quantification of HRF (defined as hyperreflective intraretinal isolated elements, with size ≤30µm, medium reflectivity, similar to RNFL, absence of back shadowing and location both in the inner and outer retina) in the central 3 mm on the linear high resolution SD-OCT scan passing through the fovea. Eyes affected by DME were then enrolled from four different Italian reference centers for diabetic retinopathy and maculopathy. Each scan of each eye was analyzed by the AI automatic quantification software and by centralized clinical evaluation, manually performed by one blinded medical retinal expert. ICC was calculated between the clinical evaluation and the automated tool and Bland-Altmann plot was developed.

Results : Three hundred and three eyes were enrolled among the four centers. The mean central subfield thickness was 386.5 ± 130.2 micron. The mean number of HRF detected by the software was 71.9 ± 22.8 vs 71.9 ± 22.7 detected by clinical evaluation. The ICC between the clinical evaluation and the automated tool was excellent (0.97). In the Bland-Altmann plot almost all measured differences were into the range (2 SD), with a mean difference between the clinical and automatic count of 0.03 ± 5.277. No significant trend was evident. No error in foveal identification and automatic retinal layers segmentation was identified in any of the linear scans analyzed.

Conclusions : HRF are a diffusely used biomarker of severity of the disease and response to treatment in DME, with particular regard to its neuroinflammatory component. A fully validated AI based tool, able to detect and quantify HRF in a reliable and repeatable way, offers the clinicians an objective way of planning and following DME eyes.

This abstract was presented at the 2023 ARVO Annual Meeting, held in New Orleans, LA, April 23-27, 2023.