Multi-Compartment OCT-derived Radiomics Features to predict Anti-VEGF Treatment Durability for Diabetic Macular Edema

Abstract

Purpose : There are currently no established predictors for therapeutic durability of anti-VEGF therapy for diabetic macular edema (DME) and retinal vein occlusion (RVO) secondary to retinal vascular diseases. The main objective of this study was to identify the (i) texture- based radiomics features within individual fluid and retinal tissue compartments of the baseline optical coherence tomography (OCT) images associated with anti-VEGF treatment response and (ii) the specific spatial (fluid and/or retinal tissue) compartments that contribute the most pertinent features for predicting therapeutic response.

Methods : A total of 962 texture-based radiomic features were extracted from each of the fluid and retinal tissue compartments from the baseline OCT images, obtained from the PERMEATE clinical trial. PERMEATE is a prospective open-label IRB- approved study for the treatment of naïve eyes with DME and RVO using intravitreal aflibercept injection (IAI) to evaluate imaging biomarkers and treatment response. The two top-performing features were selected from the consensus of different feature selection methods and were evaluated in conjunction with four different machine learning classifiers: Linear Discriminant Analysis (LDA), Quadratic Discriminant Analysis (QDA), Random Forest (RF), and Support Vector Machine (SVM) in a 3-fold cross-validated approach to distinguish the eyes tolerating extended interval dosing (Non-rebounders) and those requiring more frequent dosing (Rebounders).

Results : In this study, we found that the Laws textural features from within the intraretinal fluid (IRF) subcompartment were most strongly associated with the likelihood of a patient being a rebounder or a non-rebounder. A combination of fluid and retinal tissue features yielded a cross-validated area under the receiver operating characteristic curve (AUC) of 0.78±0.08 in distinguishing rebounders from non-rebounders via 3-fold cross validation on the PERMEATE study.

Conclusions : Our study revealed that the texture-based radiomics features pertaining to the IRF subcompartment were most discriminating between rebounders and non-rebounders to anti-VEGF therapy for macular edema in retinal vascular disease. With further validation, OCT-based imaging biomarkers may have potential for facilitating personalized therapeutic guidance.

This is a 2021 Imaging in the Eye Conference abstract.