Swept-source optical coherence tomography angiography vitreo-retinal segmentation in proliferative diabetic retinopathy

Purpose: To identify a new cortical vitreous segmentation protocol for non invasive standardised investigation of Neovascularisation (NV) with detection of regression of NV activity in Proliferative Diabetic Retinopathy (PDR). Design: Retrospective study. Participants: One hundred and eighty-six eyes of 93 consecutive diabetic patients (mean age: 52.6 ± 11.0 years) imaged using Topcon Triton® Swept-Source Optical Coherence Tomography Angiography (SS-OCTA) from June 2015 to January 2017. Methods: Scans were performed through the macula, optic disc and areas of possible NV in mid-peripheral retina using 6 × 6 mm and/or 9 × 9 mm raster-patterns in three segmentation protocols: Vitreo-Retinal (VRS), Outer-Vitreous (OVS) and Core-Vitreous Segmentation (CVS). Any suspicion of PDR was confirmed in all patients by Heidelberg® Widefield-Fundus-Fluorescein-Angiography (WF-FFA) and/or Optos® Ultra-WF-FFA (UWF-FFA). Inter-observer reproducibility of NV diagnosis and agreement between SS-OCTA and UWF-FFA were assessed. Primary outcome was the identification of an effective and reproducible segmentation protocol. Secondary outcome was the identification of NV regression after treatment. Results: Sensitivity-specificity reached, respectively, the value of 100 to 100 in detecting NVD, and 96.6 to 100 in detecting NVE in compared areas. SS-OCTA was able to confirm absence of blood flow within the residual NV plexus when using VRS protocol in 30 eyes in which regression of NV with absence of leakage was documented on FFA. Conclusion: Three segmentation protocols (VRS, OVS and CVS) with different but complementary characteristics, allowed a reproducible and standardised investigation of NVD and NVE. The proposed new SS-OCTA cortical vitreous segmentation protocols may be of value when identifying and assessing NV-activity (VRS, OVS and CVS) or NV-regression (VRS) in PDR and therefore, response to therapy.