Radiomics-Based Assessment of Optical Coherence Tomography (OCT) and OCT Angiography (OCTA) Images for Diabetes Mellitus and Diabetic Retinopathy Diagnosis
Laura Carrera-Escalé,Anass Benali,Ann-Christin Rathert,Ruben Martín-Pinardel,Carolina Bernal-Morales,Anibal Alé-Chilet,Marina Barraso,Sara Marín-Martinez,Silvia Feu-Basilio,Josep Rosinés-Fonoll,Teresa Hernandez,Irene Vilá,Rafael Castro,Cristian Oliva,Irene Vinagre,Emilio Ortega,Marga Gimenez,Alfredo Vellido,Enrique Romero,Javier Zarranz-Ventura
DOI: https://doi.org/10.2139/ssrn.4147834
2022-01-01
SSRN Electronic Journal
Abstract:Background: To evaluate the diagnostic accuracy of machine learning (ML) techniques applied to radiomic features extracted from optical coherence tomography (OCT) and OCT angiography (OCTA) images for diabetes mellitus (DM), diabetic retinopathy (DR) and referable DR (R-DR) diagnosis, in a dataset from a previous prospective OCTA study (ClinicalTrials.gov NCT03422965). Methods: Radiomic features were extracted from fundus retinographies (FR), OCT and OCTA images in each study eye. Logistic regression (LR), linear discriminant analysis (LDA), support vector classifier (SVC)-linear, SVC-rbf and random forest (RF) models were created to evaluate their diagnostic accuracy for DM, DR and R-DR diagnosis in all images type. Models´ performance was described by their Area-under-Curve (AUC) mean and standard deviation (SD). Findings: A dataset of 726 eyes (439 individuals) were included. For DM diagnosis, the greatest AUC was observed for OCT (0.82, 0.03). For DR detection the greatest AUC was observed for OCTA (0.77, 0.03), especially in the 3x3 mm superficial capillary plexus OCTA scan (AUC 0.76, 0.04). For R-DR diagnosis, the greatest AUC was observed for OCTA (0.87, 0.12) and the deep capillary plexus OCTA scan (0.86, 0.08). The addition of clinical variables (age, sex, etc.) improved most models AUC for both DM and DR diagnosis. Interpretation: Radiomics extracted from OCT and OCTA images allow identification of DM, DR and R-DR patients using standard ML classifiers. OCT was the best test for DM diagnosis, OCTA for DR and R-DR diagnosis and the addition of clinical variables improved most models. This pioneer study demonstrates that radiomics-based ML techniques applied to OCT and OCTA images may be an option for screening in the community.Trial Registration: The retinal images dataset was collected in a prospective OCTA trial (ClinicalTrials.gov NCT03422965). Funding: JZV acknowledge funding from Fundació La Marató de TV3, La Marató 2015, Diabetis i Obesitat (grant number 201633.10) and Instituto de Salud Carlos III through the projects PI18/00518 and PI21/01384 co-funded by European Union. AV and ER acknowledge funding from Spanish research grant PID2019-104551RB-I00.Declaration of Interest: JZV is a grant holder for Novartis Pharmaceuticals, Bayer and Allergan, and a consultant for Abbvie, Alcon, Alimera Sciences, Bausch and Lomb, Bayer, Brill Pharma, DORC, Novartis Pharmaceuticals, Preceyes, Roche, Topcon, and Zeiss. MISSING OTHER AUTHORSEthical Approval: This study was approved by the Institutional Review Board (IRB) (HCB/2021/0350) and adhered to the Declaration of Helsinki.