Artificial intelligence improves diagnosis and risk prediction for vision-destroying disease
2022
Challenge
Age-related macular degeneration (AMD) is the leading cause of blindness in industrialized countries but often remains undiagnosed in the early stages when interventions would improve clinical outcomes. While early diagnosis of the disease is crucial, traditional diagnostic and assessment methods have limited accuracy, even when performed by specialists who are not available in many regions.
Advance
Using deep learning programs and large, long-term datasets, IRP researchers developed automated artificial intelligence (AI) algorithms to perform diagnosis, classification, and prognostic risk prediction for AMD. The algorithms were able to diagnose AMD and classify its severity using color photographs of the eye’s retina with greater accuracy than retina specialists. Moreover, other algorithms the researchers developed detected important features of the disease that would otherwise require advanced imaging techniques to identify. Finally, the IRP scientists' algorithms can use a patient’s data to predict his or her risk of progressing to advanced disease over the next decade — again, with greater accuracy than specialists.
Impact
AI techniques like deep learning are already beginning to transform clinical care for, and research on, diseases affecting the eye, such as diabetic retinopathy. This study shows that AI algorithms can be a boon to the study and treatment of AMD as well. Importantly, the development of AI algorithms that match or exceed specialist evaluations could enormously increase access to care. In addition, such algorithms can help patients receive a diagnosis more quickly, permitting early intervention that leads to improved clinical outcomes. To hasten their adoption in clinical and research settings, the IRP scientists have made their algorithms publicly available or embedded them in desktop software for research use.
Publications
Agrón E, Domalpally A, Cukras CA, Clemons TE, Chew EY, Keenan TDL; AREDS and AREDS2 Research Groups. (2022). Reticular pseudodrusen: The third macular risk feature for progression to late age-related macular degeneration. Ophthalmology. Oct 1;129(10):1107-1119. Epub 2022 May 31. doi: 10.1016/j.ophtha.2022.05.021.
Peng Y, Keenan TD, Chen Q, Agrón E, Allot A, Wong WT, Chew EY, Lu Z. (2020). Predicting risk of late age-related macular degeneration using deep learning. NPJ Digit Med.Aug 27; 3(1):111. eCollection 2020. doi: 10.1038/s41746-020-00317-z.
Keenan TDL, Chen Q, Peng Y, Domalpally A, Agrón E, Hwang CK, Thavikulwat AT, Lee DH, Li D, Wong WT, Lu Z, Chew EY. (2020). Deep learning automated detection of reticular pseudodrusen from fundus autofluorescence images or color fundus photographs in AREDS2. Ophthalmology.Dec; 127(12):1674-1687. Epub 2020 May 21. doi: 10.1016/j.ophtha.2020.05.036.
Keenan TD, Dharssi S, Peng Y, Chen Q, Agrón E, Wong WT, Lu Z, Chew EY. (2019). A deep learning approach for automated detection of geographic atrophy from color fundus photographs. Ophthalmology.Nov; 126(11):1533-1540. Epub 2019 Jun 11. doi: 10.1016/j.ophtha.2019.06.005.
Peng Y, Dharssi S, Chen Q, Keenan TD, Agrón E, Wong WT, Chew EY, Lu Z. (2019). DeepSeeNet: a deep learning model for automated classification of patient-based age-related macular degeneration severity from color fundus photographs. OphthalmologyApr; 126(4):565-575. Epub 2018 Nov 22. doi: 10.1016/j.ophtha.2018.11.015.
This page was last updated on Monday, August 19, 2024