Abstract
Purpose :
To evaluate the performance of NOA in 2 different aspects, precision of fluid volume quantification on consecutive in-clinic self-imaging as well as agreement with reference office OCT scans acquired in the same visit
Methods :
The NOA is a deep learning algorithm that analyzes scans of the Notal Vision Home OCT (NVHO). The data is comprised of 242 NVHO volume scans acquired from 47 patients and 88 eyes with nAMD. Repeatability of fluid volume was calculated for all eyes with at least two repeats via coefficient of variance (CV). Retinal fluid agreement with manually annotated Cirrus scans was evaluated on a subset of the data and was estimated globally and spatially. Global agreement estimation was calculated with Pearson correlation coefficient (PCC) and Lin's concordance correlation coefficient (CCC). Spatial agreement was calculated with 2D correlation of the fluid thickness maps and on gridded versions of the original maps. All analysis was performed on eyes with clinically meaningful retinal fluid volume greater than 3 nanoliter
Results :
NOA repeatability performance was evaluated on a set of scans from 26 eyes resulting in a CV of 4.1%. NOA agreement with Cirrus was evaluated on a subset of 23 eyes with an expert annotation of retinal fluid in the Cirrus scans. The PCC and CCC of retinal fluid volume between Cirrus and the NVHO with NOA were 0.984 and 0.975 respectively. For spatial agreement, the median (IQR) 2D correlation was 0.869 (0.769-0.940) and the median (IQR) 2D correlation of the maps with grids with densities of 20x20,10x10 was 0.841 (0.720-0.941) and 0.915 (0.757-0.963), respectively
Conclusions :
The results validate the performance of the home OCT system comprised of a device and an AI-based algorithm. The low CV allows for accurate retinal fluid quantification with small variation in clinically meaningful amounts of retinal fluid. The high PCC and CCC of fluid volume and spatial correlation confirm that the same pathologies were accurately identified by NOA and the reference office OCT. Overall, the results present further evidence for the ability of the Notal Vision Home OCT to remotely monitor nAMD disease activity
This abstract was presented at the 2022 ARVO Annual Meeting, held in Denver, CO, May 1-4, 2022, and virtually.