Deep Learning –Based Automated OCT Segmentation in Clinical Routine

Recently, many ophthalmologists have heard the keywords artificial intelligence, machine learning, deep learning, and automatization at every conference and keynote lecture and seen them in every ophthalmology journal. Many studies have evaluated the use of such algorithms on large retrospective data sets —primarily on color fundus photographs at first, then on optical coherence tomography (OCT) images as well. Most of these have been study data sets with standardized and well-structured imaging protocols and reading center image collections with a predefined protocol, and therefore of good quality . However, how functional will algorithms be in a busy clinical routine? Can we trust a computer to localize retinal fluid, and will we base our treatment decisions on automated volumetric measurements of retinal fluid in the future? Could the computer lead to wrong decisions if we fail to detect “ erroneous” segmentations in a clinical setting?
Source: JAMA Ophthalmology - Category: Opthalmology Source Type: research