Abstract
Purpose :
To describe a method for rapid optical coherence tomography (OCT) image capture and multidimensional visualization of retinopathy of prematurity (ROP) in preterm infants in BabySTEPS2.
Methods :
We developed retinal OCT imaging protocols for use with an investigational 400kHz, 1060nm swept-source system with a 700g handheld, noncontact probe in preterm infants at risk for ROP. Two imaging protocols were used at the bedside, the day of clinical ROP screening, without sedation or use of an eyelid speculum: Scan 1 (0.7 sec, 1900 A-scans/B-scan, 128 B-scans per 10x10mm volume, with pairs of adjacent A-scans summed producing 950 A-scans/B-scan) for cross-sectional assessment of retinal layers; and Scan 2 (1.4 sec, 640 A-scans/B-scan and 640 B-scans per 10x10mm isodense volume) for en face retinal vascular assessment. For multidimensional visualization, we developed infant-specific software (DOCTRAP v66.2) to optimize automatic segmentation of OCT volumes from Scan 1 and generate total retina thickness maps. We extracted en face retinal vessel maps from maximum pixel intensity at each A-scan position from Scan 2. With auto-montaging software (i2k Retina; DualAlign, Inc.), we created a single posterior pole vessel map per eye and overlaid corresponding thickness maps for comparison to color fundus photos.
Results :
We were able to capture, integrate, and visualize multidimensional data from high-speed investigational bedside OCT imaging of preterm infant retinas. This included: depth-resolved analysis of retinal microanatomy at locations of interest (e.g., at the foveal center and vascular-avascular junction); intensity maps of total retinal layer thicknesses; and retinal vessel maps (Figure 1). These data demonstrated important ROP pathology (e.g., retinal neovascularization and plus disease) and in contrast to color photographs, enabled integrated assessment of the developing preterm retina (e.g., retinal layer thicknesses and presence of macular edema).
Conclusions :
This method for multidimensional data analysis will be used by human graders and analyzed by semi-automated software (i.e., ROPtool) in BabySTEPS2, and may be integrated into future artificial intelligence-based ROP screening methods.
This abstract was presented at the 2022 ARVO Annual Meeting, held in Denver, CO, May 1-4, 2022, and virtually.