Multi-task deep learning for glaucoma detection from color fundus images
Abstract Glaucoma is an eye condition that leads to loss of vision and blindness if not diagnosed in time. Diagnosis requires human experts to estimate in a limited time subtle changes in the shape of the optic disc from retinal fundus images. Deep learning methods have been satisfactory in classify...
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Nature Portfolio
2022-07-01
|
| Series: | Scientific Reports |
| Online Access: | https://doi.org/10.1038/s41598-022-16262-8 |
| _version_ | 1818509901195902976 |
|---|---|
| author | Lucas Pascal Oscar J. Perdomo Xavier Bost Benoit Huet Sebastian Otálora Maria A. Zuluaga |
| author_facet | Lucas Pascal Oscar J. Perdomo Xavier Bost Benoit Huet Sebastian Otálora Maria A. Zuluaga |
| author_sort | Lucas Pascal |
| collection | DOAJ |
| description | Abstract Glaucoma is an eye condition that leads to loss of vision and blindness if not diagnosed in time. Diagnosis requires human experts to estimate in a limited time subtle changes in the shape of the optic disc from retinal fundus images. Deep learning methods have been satisfactory in classifying and segmenting diseases in retinal fundus images, assisting in analyzing the increasing amount of images. Model training requires extensive annotations to achieve successful generalization, which can be highly problematic given the costly expert annotations. This work aims at designing and training a novel multi-task deep learning model that leverages the similarities of related eye-fundus tasks and measurements used in glaucoma diagnosis. The model simultaneously learns different segmentation and classification tasks, thus benefiting from their similarity. The evaluation of the method in a retinal fundus glaucoma challenge dataset, including 1200 retinal fundus images from different cameras and medical centers, obtained a $$96.76 \pm 0.96$$ 96.76 ± 0.96 AUC performance compared to an $$93.56 \pm 1.48$$ 93.56 ± 1.48 obtained by the same backbone network trained to detect glaucoma. Our approach outperforms other multi-task learning models, and its performance pairs with trained experts using $$~\sim 3.5$$ ∼ 3.5 times fewer parameters than training each task separately. The data and the code for reproducing our results are publicly available. |
| first_indexed | 2024-12-10T22:51:50Z |
| format | Article |
| id | doaj.art-9b07bc8175674173bcaef4bdf4346910 |
| institution | Directory Open Access Journal |
| issn | 2045-2322 |
| language | English |
| last_indexed | 2024-12-10T22:51:50Z |
| publishDate | 2022-07-01 |
| publisher | Nature Portfolio |
| record_format | Article |
| series | Scientific Reports |
| spelling | doaj.art-9b07bc8175674173bcaef4bdf43469102022-12-22T01:30:25ZengNature PortfolioScientific Reports2045-23222022-07-0112111010.1038/s41598-022-16262-8Multi-task deep learning for glaucoma detection from color fundus imagesLucas Pascal0Oscar J. Perdomo1Xavier Bost2Benoit Huet3Sebastian Otálora4Maria A. Zuluaga5Data Science Department, EURECOMSchool of Medicine and Health Sciences, Universidad del RosarioOrkisMedian TechnologiesSupport Center for Advanced Neuroimaging, University Institute of Diagnostic and Interventional NeuroradiologyData Science Department, EURECOMAbstract Glaucoma is an eye condition that leads to loss of vision and blindness if not diagnosed in time. Diagnosis requires human experts to estimate in a limited time subtle changes in the shape of the optic disc from retinal fundus images. Deep learning methods have been satisfactory in classifying and segmenting diseases in retinal fundus images, assisting in analyzing the increasing amount of images. Model training requires extensive annotations to achieve successful generalization, which can be highly problematic given the costly expert annotations. This work aims at designing and training a novel multi-task deep learning model that leverages the similarities of related eye-fundus tasks and measurements used in glaucoma diagnosis. The model simultaneously learns different segmentation and classification tasks, thus benefiting from their similarity. The evaluation of the method in a retinal fundus glaucoma challenge dataset, including 1200 retinal fundus images from different cameras and medical centers, obtained a $$96.76 \pm 0.96$$ 96.76 ± 0.96 AUC performance compared to an $$93.56 \pm 1.48$$ 93.56 ± 1.48 obtained by the same backbone network trained to detect glaucoma. Our approach outperforms other multi-task learning models, and its performance pairs with trained experts using $$~\sim 3.5$$ ∼ 3.5 times fewer parameters than training each task separately. The data and the code for reproducing our results are publicly available.https://doi.org/10.1038/s41598-022-16262-8 |
| spellingShingle | Lucas Pascal Oscar J. Perdomo Xavier Bost Benoit Huet Sebastian Otálora Maria A. Zuluaga Multi-task deep learning for glaucoma detection from color fundus images Scientific Reports |
| title | Multi-task deep learning for glaucoma detection from color fundus images |
| title_full | Multi-task deep learning for glaucoma detection from color fundus images |
| title_fullStr | Multi-task deep learning for glaucoma detection from color fundus images |
| title_full_unstemmed | Multi-task deep learning for glaucoma detection from color fundus images |
| title_short | Multi-task deep learning for glaucoma detection from color fundus images |
| title_sort | multi task deep learning for glaucoma detection from color fundus images |
| url | https://doi.org/10.1038/s41598-022-16262-8 |
| work_keys_str_mv | AT lucaspascal multitaskdeeplearningforglaucomadetectionfromcolorfundusimages AT oscarjperdomo multitaskdeeplearningforglaucomadetectionfromcolorfundusimages AT xavierbost multitaskdeeplearningforglaucomadetectionfromcolorfundusimages AT benoithuet multitaskdeeplearningforglaucomadetectionfromcolorfundusimages AT sebastianotalora multitaskdeeplearningforglaucomadetectionfromcolorfundusimages AT mariaazuluaga multitaskdeeplearningforglaucomadetectionfromcolorfundusimages |