Author: Aaron Y. Lee MD MSCI

Affiliation(s): University of Washington, Seattle WA

Purpose: To apply deep learning artificial intelligence technology to optical coherence tomography and retinal imaging.

Methods: Traditionally difficult computational problems were identified in retinal imaging such as classification and segmentation. State-of-the-art deep learning and artificial intelligence techniques were used to advance the field in classification and segmentation tasks.

Results: Deep learning was able to classify macular degeneration accurately with an AUROC of greater than 0.97. In addition, deep learning was able to segment intraretinal fluid on macular OCT B scans with Dice coefficient of 0.91. Finally, deep learning was able to transform regular standard structural OCT into OCTA.

Conclusions: Artificial intelligence algorithms hold great promise in transforming the field of ophthalmology.

Financial Disclosure: Topcon (honorarium), Carl Zeiss Meditec (Grant support), Novartis (Grant Support), NVIDIA (Grant support), Microsoft (Grant Support)