Deep Learning for Multi‐Disease OCT Classification in Clinical Retina Practice Using Attention Mechanisms and Contrastive Learning
- 1Department of Biomedical Engineering Meybod University Meybod Iran
- 2Department of Biomedical Engineering Amirkabir University of Technology (AUT) Tehran Iran
- 3Department of Electrical Engineering Faculty of Engineering Vali‐e‐Asr University of Rafsanjan Rafsanjan Iran
Abstract
Artificial intelligence (AI) is increasingly used in ophthalmic imaging to support early detection of vision‐threatening retinal diseases such as age‐related macular degeneration, diabetic retinopathy (DR) and choroidal neovascularisation (CNV). However, optical coherence tomography (OCT)‐based deep learning models may remain sensitive to image degradation, struggle to resolve subtle inter‐class morphological differences and show reduced reliability beyond controlled benchmark settings. We propose AESNet‐SEC, a lightweight attention‐enhanced SqueezeNet framework integrating multi‐scale dilated convolutions, spatial attention, squeeze‐and‐excitation recalibration and contrastive representation learning. The framework was evaluated on the Retinal OCT‐C8 dataset (24,000 images; eight classes) and a prospectively collected clinical dataset from Zeyaei Hospital, Iran (9732 OCT B‐scans; four classes). AESNet‐SEC achieved 98.56% accuracy on OCT‐C8 and 96.33%–96.98% across four repeated clinical evaluations. Ablation and ROC analyses showed the strongest overall discrimination for the complete configuration, while controlled Gaussian‐noise testing revealed greater degradation for DR and drusen. Grad‐CAM visualisations highlighted anatomically plausible retinal regions, with independent retinal‐specialist review providing qualitative support for the observed activation patterns. Overall, AESNet‐SEC combines strong multi‐class discrimination, computational efficiency, controlled robustness and visual interpretability, while broader patient‐disjoint, multi‐centre validation remains necessary.
Related
- Diffusion-augmented contrastive learning for diabetic retinopathy severity classification
- A multi-class fundus disease classification framework based on self-supervised contrastive learning and dual-branch network
- Multi-view attention-based deep learning for benign–malignant classification of pulmonary nodules
- Cross-attention fusion of biparametric MRI sequences with demographic and clinical variables for patient-level classification of clinically significant prostate cancer
- Peptides in Aesthetic Dermatology: Classification, Fibroblast-Centered Mechanisms, Delivery Technologies, and Clinical Evidence.
BibTeX
@article{rezaee2026deep,
title = {{Deep Learning for Multi‐Disease OCT Classification in Clinical Retina Practice Using Attention Mechanisms and Contrastive Learning}},
author = {Khosro Rezaee and Safoura Farsi and Hossein Ghayoumi Zadeh},
journal = {Healthcare Technology Letters},
year = {2026},
month = oct,
doi = {10.1049/htl2.70106},
url = {https://doi.org/10.1049/htl2.70106},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Deep Learning for Multi‐Disease OCT Classification in Clinical Retina Practice Using Attention Mechanisms and Contrastive Learning
AU - Khosro Rezaee
AU - Safoura Farsi
AU - Hossein Ghayoumi Zadeh
PY - 2026
DA - 2026/10/25
JO - Healthcare Technology Letters
DO - 10.1049/htl2.70106
UR - https://doi.org/10.1049/htl2.70106
AB - Artificial intelligence (AI) is increasingly used in ophthalmic imaging to support early detection of vision‐threatening retinal diseases such as age‐related macular degeneration, diabetic retinopathy (DR) and choroidal neovascularisation (CNV). However, optical coherence tomography (OCT)‐based deep learning models may remain sensitive to image degradation, struggle to resolve subtle inter‐class morphological differences and show reduced reliability beyond controlled benchmark settings. We propose AESNet‐SEC, a lightweight attention‐enhanced SqueezeNet framework integrating multi‐scale dilated convolutions, spatial attention, squeeze‐and‐excitation recalibration and contrastive representation learning. The framework was evaluated on the Retinal OCT‐C8 dataset (24,000 images; eight classes) and a prospectively collected clinical dataset from Zeyaei Hospital, Iran (9732 OCT B‐scans; four classes). AESNet‐SEC achieved 98.56% accuracy on OCT‐C8 and 96.33%–96.98% across four repeated clinical evaluations. Ablation and ROC analyses showed the strongest overall discrimination for the complete configuration, while controlled Gaussian‐noise testing revealed greater degradation for DR and drusen. Grad‐CAM visualisations highlighted anatomically plausible retinal regions, with independent retinal‐specialist review providing qualitative support for the observed activation patterns. Overall, AESNet‐SEC combines strong multi‐class discrimination, computational efficiency, controlled robustness and visual interpretability, while broader patient‐disjoint, multi‐centre validation remains necessary.
SN - 2053-3713
ER -