Multimodal SLO-OCT ophthalmic imaging dataset

Multimodal Ophthalmic Imaging Dataset The code for our paper is at code.zip, code.z01, ..., code.z07. The dataset is at datazet.zip, dataset.z01, ..., dataset.z205. The labels are at the labels folder. Modalities 2D SLO images (Scanning Laser Ophthalmoscopy) 3D OCT volumes: Macular-centered OCT (OCTmacular) Disc-centered OCT (OCTdisc) Diseases Covered Macular Edema (ME) Diabetic Retinopathy (DR) Glaucoma (GL) Dataset Composition Modality Combinations # Samples SLO (only) 1822, OCTmacular (only) 2092, OCTdisc (only) 2310, SLO + OCTmacular (paired)1346, SLO + OCTdisc (paired)1170, Total8740. Collected from 5235 patients (2019-2022) using: Optos Panoramic200 (SLO) CIRRUS HD-OCT 500 (OCT) Key Features Real-world diversity: Includes both modality-complete and modality-incomplete samples Reflects clinical scenarios with personalized exams Supports multiple diseases (ME, DR, GL) unlike single-disease datasets Labeling Sources: Electronic medical records Ophthalmologist reviews Classification: ME/DR: Binary (present/absent) GL: Ternary (glaucoma/non-glaucoma/suspicious) Splits Stratified 60:20:20 train/val/test split Addresses class imbalance