Abstract
Purpose :
Remote OCT applications require cost-effective acquisition devices. This challenging constraint often leads to technology choices that provide inferior image quality compared to state-of-the-art clinical devices, and thereby result in unreliable image analysis. The purpose of this work is to demonstrate a strategy for robust macula thickness analysis using low-cost OCT that can navigate data quality limitations and improve correlation of thickness values compared with benchmark clinical systems.
Methods :
Our method relies on acquiring multiple OCT scans of the same region and intelligently combining the macula thickness maps generated by individual scans. In this method, inner limiting membrane and retina pigment epithelium are delineated, and the corresponding thickness map and segmentation confidence map (SCM) are generated for each OCT volume. Subsequently, all thickness maps are registered to the thickness map (reference) with highest confidence derived from the SCM. The registration transformation parameters are then used to register and combine the thickness maps into a single map based on the local quality of individual SCMs.
To evaluate this method, we acquired OCT scans over an area of 5.78mm x 5.78mm from each eye, repeated 3 times on a low-cost OCT. 38 patients with retina pathology were enrolled in an IRB-approved study. The OCT macula thickness map was compared to one acquired using CIRRUSTM 5000 HD-OCT (ZEISS, Dublin, CA). Thickness values were compared over an ETDRS grid.
Results :
Figure 1 shows examples of thickness maps acquired using the low-cost OCT (before and after combining) and using CIRRUS 5000 HD-OCT. Bland-Altman analysis was performed to compare the results of single acquisition and triple acquisition of the low-cost system with a single acquisition of CIRRUS 5000 HD-OCT. The results are summarized in Table 1. The combined acquisition provides slightly higher mean difference while improving the correlation with the clinical system in the outer ring.
Conclusions :
Our analysis suggests that combining macula thickness maps from multiple acquisitions provides an overall improved correlation between the low-cost system and clinical system, and thereby improves the diagnostic utility of the low-cost OCT instrument.
This abstract was presented at the 2022 ARVO Annual Meeting, held in Denver, CO, May 1-4, 2022, and virtually.