SSMD-UNet: semi-supervised multi-task decoders network for diabetic retinopathy segmentation
Abstract Diabetic retinopathy (DR) is a diabetes complication that can cause vision loss among patients due to damage to blood vessels in the retina. Early retinal screening can avoid the severe consequences of DR and enable timely treatment. Nowadays, researchers are trying to develop automated dee...
Main Authors: | , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
Nature Portfolio
2023-06-01
|
Series: | Scientific Reports |
Online Access: | https://doi.org/10.1038/s41598-023-36311-0 |
_version_ | 1797806750667112448 |
---|---|
author | Zahid Ullah Muhammad Usman Siddique Latif Asifullah Khan Jeonghwan Gwak |
author_facet | Zahid Ullah Muhammad Usman Siddique Latif Asifullah Khan Jeonghwan Gwak |
author_sort | Zahid Ullah |
collection | DOAJ |
description | Abstract Diabetic retinopathy (DR) is a diabetes complication that can cause vision loss among patients due to damage to blood vessels in the retina. Early retinal screening can avoid the severe consequences of DR and enable timely treatment. Nowadays, researchers are trying to develop automated deep learning-based DR segmentation tools using retinal fundus images to help Ophthalmologists with DR screening and early diagnosis. However, recent studies are unable to design accurate models due to the unavailability of larger training data with consistent and fine-grained annotations. To address this problem, we propose a semi-supervised multitask learning approach that exploits widely available unlabelled data (i.e., Kaggle-EyePACS) to improve DR segmentation performance. The proposed model consists of novel multi-decoder architecture and involves both unsupervised and supervised learning phases. The model is trained for the unsupervised auxiliary task to effectively learn from additional unlabelled data and improve the performance of the primary task of DR segmentation. The proposed technique is rigorously evaluated on two publicly available datasets (i.e., FGADR and IDRiD) and results show that the proposed technique not only outperforms existing state-of-the-art techniques but also exhibits improved generalisation and robustness for cross-data evaluation. |
first_indexed | 2024-03-13T06:12:02Z |
format | Article |
id | doaj.art-ff536066ff9640f4b88eebcb53e8f7ca |
institution | Directory Open Access Journal |
issn | 2045-2322 |
language | English |
last_indexed | 2024-03-13T06:12:02Z |
publishDate | 2023-06-01 |
publisher | Nature Portfolio |
record_format | Article |
series | Scientific Reports |
spelling | doaj.art-ff536066ff9640f4b88eebcb53e8f7ca2023-06-11T11:13:29ZengNature PortfolioScientific Reports2045-23222023-06-0113111610.1038/s41598-023-36311-0SSMD-UNet: semi-supervised multi-task decoders network for diabetic retinopathy segmentationZahid Ullah0Muhammad Usman1Siddique Latif2Asifullah Khan3Jeonghwan Gwak4Department of Software, Korea National University of TransportationDepartment of Computer Science and Engineering, Seoul National UniversityFaculty of Health and Computing, University of Southern QueenslandPattern Recognition Lab, DCIS, PIEAS, NiloreDepartment of Software, Korea National University of TransportationAbstract Diabetic retinopathy (DR) is a diabetes complication that can cause vision loss among patients due to damage to blood vessels in the retina. Early retinal screening can avoid the severe consequences of DR and enable timely treatment. Nowadays, researchers are trying to develop automated deep learning-based DR segmentation tools using retinal fundus images to help Ophthalmologists with DR screening and early diagnosis. However, recent studies are unable to design accurate models due to the unavailability of larger training data with consistent and fine-grained annotations. To address this problem, we propose a semi-supervised multitask learning approach that exploits widely available unlabelled data (i.e., Kaggle-EyePACS) to improve DR segmentation performance. The proposed model consists of novel multi-decoder architecture and involves both unsupervised and supervised learning phases. The model is trained for the unsupervised auxiliary task to effectively learn from additional unlabelled data and improve the performance of the primary task of DR segmentation. The proposed technique is rigorously evaluated on two publicly available datasets (i.e., FGADR and IDRiD) and results show that the proposed technique not only outperforms existing state-of-the-art techniques but also exhibits improved generalisation and robustness for cross-data evaluation.https://doi.org/10.1038/s41598-023-36311-0 |
spellingShingle | Zahid Ullah Muhammad Usman Siddique Latif Asifullah Khan Jeonghwan Gwak SSMD-UNet: semi-supervised multi-task decoders network for diabetic retinopathy segmentation Scientific Reports |
title | SSMD-UNet: semi-supervised multi-task decoders network for diabetic retinopathy segmentation |
title_full | SSMD-UNet: semi-supervised multi-task decoders network for diabetic retinopathy segmentation |
title_fullStr | SSMD-UNet: semi-supervised multi-task decoders network for diabetic retinopathy segmentation |
title_full_unstemmed | SSMD-UNet: semi-supervised multi-task decoders network for diabetic retinopathy segmentation |
title_short | SSMD-UNet: semi-supervised multi-task decoders network for diabetic retinopathy segmentation |
title_sort | ssmd unet semi supervised multi task decoders network for diabetic retinopathy segmentation |
url | https://doi.org/10.1038/s41598-023-36311-0 |
work_keys_str_mv | AT zahidullah ssmdunetsemisupervisedmultitaskdecodersnetworkfordiabeticretinopathysegmentation AT muhammadusman ssmdunetsemisupervisedmultitaskdecodersnetworkfordiabeticretinopathysegmentation AT siddiquelatif ssmdunetsemisupervisedmultitaskdecodersnetworkfordiabeticretinopathysegmentation AT asifullahkhan ssmdunetsemisupervisedmultitaskdecodersnetworkfordiabeticretinopathysegmentation AT jeonghwangwak ssmdunetsemisupervisedmultitaskdecodersnetworkfordiabeticretinopathysegmentation |