Purchase this article with an account.
Bryan Ming-Tak Wong, Richard Cheng, Wendy Hatch, Efrem Mandelcorn, Edward Margolin, Peng Yan, Anna Theresa Santiago, Wendy Lou, Christopher Hudson; Validation of optical coherence tomography retinal segmentation algorithm in neuro-degenerative disease. Invest. Ophthalmol. Vis. Sci. 2017;58(8):1307.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
This study compared the automated retinal segmentation software of spectral domain optical coherence tomography (SD-OCT) with manually corrected segmentation in order to guide and validate its use in a prospective clinical study of neuro-degenerative diseases (NDD).
The sample comprised 30 NDD patients, including individuals with vascular cognitive impairment (n=13), frontotemporal dementia (n=6), Parkinson's disease (n=6) and Alzheimer's disease (n=5). Retinal SD-OCT scans were acquired for both eyes and then segmented using the Heidelberg Spectralis (Heidelberg, Germany) software (version 188.8.131.52). All SD-OCT scans had a quality score of 20 or better. For all the B-scans enclosed by a 3.6mm circle centered on the foveola of one randomly selected eye of each patient, one of two trained observers manually corrected erroneous internal limiting membrane, retinal nerve fiber layer (RNFL), outer plexiform layer and Bruch's membrane lines. Mean volume and mean thickness measurements for four retinal layers (total retina, RNFL, all inner retinal layers and all outer retinal layers) were then obtained. Intra-class correlation coefficients (ICCs) and Bland-Altman analyses were conducted on the data.
The ICCs between the automated software and a trained observer were excellent (>0.98) for retinal thickness and volume of all 4 retinal layers. Mean differences in volume between software and observers were 0.003mm3, 0.001mm3, 0.006mm3, and -0.003mm3, respectively, for total retina, RNFL, inner retinal layers, and outer retinal layers, while mean differences in thickness were -0.004μm, 0.492μm, 1.855μm, and -1.859μm.
There was excellent agreement between the software and trained observers in identifying the retinal layer segmentation lines. These findings provide a foundation for future non-invasive analyses of retinal morphology in patient populations with neurodegenerative diseases.
This is an abstract that was submitted for the 2017 ARVO Annual Meeting, held in Baltimore, MD, May 7-11, 2017.
This PDF is available to Subscribers Only