Abstract
Purpose :
To determine agreement on quantitative retinal and choroidal measurements obtained with Duke Reading Center’s automatic segmentation software, DRCVisualizer and native Spectralis Heidelberg Eye Explorer (HEYEX) OCT software in eyes with neovascular age related macular degeneration (nvAMD) under treatment in a clinical trial.
Methods :
Eighty eyes of eighty participants from a phase 2 nvAMD clinical trial dataset with good quality images were randomly selected. Images were acquired with a 20x20 degree 97-line High Resolution scan protocol. For each scan, the standard nine-field ETDRS grid was centered at the fovea, and the HEYEX software (Version 1.10.4.0) automatically segmented the internal limiting membrane and Bruch’s membrane on each B-scan image. Any segmentation errors were manually corrected in HEYEX and all scans were then imported into DRCVisualizer (Version 2.1.1). We compared the following automatically computed parameters generated by both DRCVisualizer and HEYEX: Central Subfield Thickness (CST), Central 1mm Volume, and Total ETDRS Grid Volume. The following parameters were manually annotated on both software by a trained grader: Center point thickness, foveal neurosensory retinal thickness, foveal subretinal hyperreflective material thickness, foveal subretinal fluid thickness, subfoveal choroidal thickness, and Horizontal and Vertical extent of pigment epithelium detachment. Intraclass correlation coefficients (ICC) were used to determine agreement between measurements obtained.
Results :
There was excellent agreement between HEYEX and DRCVisualizer for all automatically computed thickness and volume parameters as well as for all manually annotated measurements. ICC were excellent (Table 1) for all parameters (range 0.99-1.00). Measured values were very similar using the two software and percent differences between the measured values for all 10 parameters ranged from 0.048 to 1.513% (Table 2).
Conclusions :
Semiautomatic retinal thickness and volume measurements and manually annotated anatomic features generated by the Duke Reading Center’s DRCVisualizer software have excellent agreement with and are very similar to the values generated by the Spectralis HEYEX software in eyes with nvAMD in a clinical trial setting.
This abstract was presented at the 2022 ARVO Annual Meeting, held in Denver, CO, May 1-4, 2022, and virtually.