Search

Purpose: To evaluate the diagnostic accuracy of machine learning (ML) techniques applied to radiomic features extracted from OCT and OCT angiography (OCTA) images for diabetes mellitus (DM), diabetic retinopathy (DR), and referable DR (R-DR) diagnosis. Design: Cross-sectional analysis of a retinal image dataset from a previous prospective OCTA study (ClinicalTrials.gov NCT03422965). Participants: Patients with type 1 DM and controls included in the progenitor study. Methods: Radiomic features were extracted from fundus retinographies, OCT, and OCTA images in each study eye. Logistic regression, linear discriminant analysis, support vector classifier (SVC)-linear, SVC-radial basis function, and random forest models were created to evaluate their diagnostic accuracy for DM, DR, and R-DR diagnosis in all image types. Main Outcome Measures: Area under the receiver operating characteristic curve (AUC) mean and standard deviation for each ML model and each individual and combined image types. Results: 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 3 x 3 mm superficial capillary plexus OCTA scan (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 DM, DR and R-DR diagnosis. The performance of the models was similar in unilateral and bilateral eyes image datasets. Conclusions: Radiomics extracted from OCT and OCTA images allow identification of patients with DM, DR, and R-DR 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 opti, The authors made the following disclosures: A.V.: Grants e Spanish research grant (grant no.: PID2019-104551RB-I00). J.Z.V.: Grant e Fundació La Marató de TV3, La Marató 2015, Diabetis i Obesitat (grant no.: 201633.10), Instituto de Salud Carlos III (grant nos.: PI18/00518, PI21/01384). E.R.: Grant e Spanish research grant (grant no.: PID2019-104551RB-I00)., Peer Reviewed, Article signat per 20 autors/es: Laura Carrera-Escalé, MSc,1,2; Anass Benali, MSc,1,2; Ann-Christin Rathert, MSc,1,2; Ruben Martín-Pinardel, MSc,1,2,3; Carolina Bernal-Morales, MD, PhD,4; Anibal Alé-Chilet, MD,4; Marina Barraso, MD, PhD,4; Sara Marín-Martinez, MD,4; Silvia Feu-Basilio, MD,4; Josep Rosinés-Fonoll, MD, 4; Teresa Hernandez, OD,3,4; Irene Vilá, OD,3,4; Rafael Castro-Dominguez, OD,4; Cristian Oliva, OD,3,4; Irene Vinagre, MD, PhD,3,5,6; Emilio Ortega, MD, PhD,3,5,6; Marga Gimenez, MD, PhD,3,5,6; Alfredo Vellido, PhD,1,2; Enrique Romero, PhD,1,2; Javier Zarranz-Ventura, MD, PhD3,4,5,7 // 1 Intelligent Data Science and Artificial Intelligence (IDEAI) Research Center; 2 Department of Computer Science, Facultat d’Informàtica de Barcelona (FIB), Universitat Politècnica de Catalunya (UPC), Barcelona, Spain; 3 August Pi i Sunyer Biomedical Research Institute (IDIBAPS), Barcelona, Spain; 4 Institut Clínic d'Oftalmología (ICOF), Hospital Clínic de Barcelona, Barcelona, Spain; 5 Diabetes Unit, Hospital Clínic de Barcelona, Spain; 6 Institut Clínic de Malalties Digestives i Metaboliques (ICMDM), Hospital Clínic de Barcelona, Spain; 7 School of Medicine, Universitat de Barcelona, Spain., Postprint (published version)

In current ophthalmology, images of the vascular system in the human retina are used as exploratory proxies for pathologies affecting different organs. In this brief paper, we use multi-modal retinal images for assisting diagnostic decision making in diabetes mellitus and diabetic retinopathy. We report the use of matrix factorization-based source extraction techniques to pre-process the images as a data quality control step prior to their classification. Through this quality control, we unveil some relevant sources of bias in the data. After correcting for them, promising pathology classification results are still obtained, which merit further exploration., This research is partially funded by research grant PID2019-104551RB-I00, Peer Reviewed, Article signat per 20 autors/es: Anass Benali 1,2; Laura Carrera 1,2; Ann Christin 1,2; Ruben Martín 1,2,6; Anibal Alé 3; Marina Barraso 3; Carolina Bernal 3; Sara Marín 3; Silvia Feu 3; Josep Rosinés 3; Teresa Hernandez 3,6; Irene Vilá 3,6; Cristian Oliva 3,6; Irene Vinagre 4,5,6; Emilio Ortega 4,5,6; Marga Gimenez 4,5,6; Enric Esmatjes 4,5,6; Javier Zarranz-Ventura 3,4,6; Enrique Romero 1,2 and Alfredo Vellido 1,2 // 1 Intelligent Data Science and Artificial Intelligence (IDEAI-UPC) Research Center; 2 Computer Science Department, Facultat d’Informàtica de Barcelona (FIB), Universitat Politècnica de Catalunya (UPC BarcelonaTech); 3 Institut Clínic d’Oftalmología (ICOF), Hospital Clínic de Barcelona, Barcelona, Spain; 4 Diabetes Unit, Hospital Clínic de Barcelona, Barcelona, Spain; 5 Institut Clínic de Malalties Digestives i Metaboliques (ICMDM), Hospital Clínic de Barcelona, Barcelona, Spain; 6 August Pi i Sunyer Biomedical Research Institute (IDIBAPS), Barcelona, Spain, Postprint (author's final draft)