Automatic segmentation of multitype retinal fluid from optical coherence tomography images using semisupervised deep learning network.

BACKGROUND/AIMS: To develop and validate a deep learning model for automated segmentation of multitype retinal fluid using optical coherence tomography (OCT) images. METHODS: We retrospectively collected a total of 2814 completely anonymised OCT images with subretinal fluid (SRF) and intraretinal fluid (IRF) from 141 patients between July 2018 and June 2020, constituting our in-house retinal OCT dataset. On this dataset, we developed a novel semisupervised retinal fluid segmentation deep network (Ref-Net) to automatically identify SRF and IRF in a coarse-to-refine fashion. We performed quantitative and qualitative analyses on the model's performance while verifying its generalisation ability by using our in-house retinal OCT dataset for training and an unseen Kermany dataset for testing. We also determined the importance of major components in the semisupervised Ref-Net through extensive ablation. The main outcome measures were Dice similarity coefficient (Dice), sensitivity (Sen), specificity (Spe) and mean absolute error (MAE). RESULTS: Our model trained on a handful of labelled OCT images manifested higher performance (Dice: 81.2%, Sen: 87.3%, Spe: 98.8% and MAE: 1.1% for SRF; Dice: 78.0%, Sen: 83.6%, Spe: 99.3% and MAE: 0.5% for IRF) over most cutting-edge segmentation models. It obtained expert-level performance with only 80 labelled OCT images and even exceeded two out of three ophthalmologists with 160 labelled OCT images. Its satisfactory generalisation capability across an unseen dataset was also demonstrated. CONCLUSION: The semisupervised Ref-Net required only la few labelled OCT images to generate outstanding performance in automate segmentation of multitype retinal fluid, which has the potential for providing assistance for clinicians in the management of ocular disease.

Joint optic disk and cup segmentation for glaucoma screening using a region-based deep learning network.

OBJECTIVES: To develop and validate an end-to-end region-based deep convolutional neural network (R-DCNN) to jointly segment the optic disc (OD) and optic cup (OC) in retinal fundus images for precise cup-to-disc ratio (CDR) measurement and glaucoma screening. METHODS: In total, 2440 retinal fundus images were retrospectively obtained from 2033 participants. An R-DCNN was presented for joint OD and OC segmentation, where the OD and OC segmentation problems were formulated into object detection problems. We compared R-DCNN's segmentation performance on our in-house dataset with that of four ophthalmologists while performing quantitative, qualitative and generalization analyses on the publicly available both DRISHIT-GS and RIM-ONE v3 datasets. The Dice similarity coefficient (DC), Jaccard coefficient (JC), overlapping error (E), sensitivity (SE), specificity (SP) and area under the curve (AUC) were measured. RESULTS: On our in-house dataset, the proposed model achieved a 98.51% DC and a 97.07% JC for OD segmentation, and a 97.63% DC and a 95.39% JC for OC segmentation, achieving a performance level comparable to that of the ophthalmologists. On the DRISHTI-GS dataset, our approach achieved 97.23% and 94.17% results in DC and JC results for OD segmentation, respectively, while it achieved a 94.56% DC and an 89.92% JC for OC segmentation. Additionally, on the RIM-ONE v3 dataset, our model generated DC and JC values of 96.89% and 91.32% on the OD segmentation task, respectively, whereas the DC and JC values acquired for OC segmentation were 88.94% and 78.21%, respectively. CONCLUSION: The proposed approach achieved very encouraging performance on the OD and OC segmentation tasks, as well as in glaucoma screening. It has the potential to serve as a useful tool for computer-assisted glaucoma screening.