Fully automated segmentation and quantification of the macular pigment burden detected on SS-OCTA and SD-OCT scans

Abstract

Purpose : To automatically identify, segment, and compare the extent of hyperpigmentation using optical attenuation coefficients (OACs) with swept-source optical coherence tomography (SS-OCTA) and spectral domain (SD) OCT scans, patients were imaged both on SS-OCTA and SD-OCT instruments on the same day and the same algorithm was used to assess the total macular pigment burden.

Methods : Patients with age-related macular degeneration (AMD) and evidence of macular hyperpigmentation were imaged using the 6x6 mm scan patterns on both SS-OCTA (PLEX® Elite 9000; ZEISS, Dublin, CA) and SD-OCT (Cirrus® HD-OCT; ZEISS, Dublin, CA). The data cube of the SD-OCT system consisted of 200 A-scans per B-scans and 200 B-scans along the y-axis, while the data cube from the SS-OCTA system consisted of 500 A-scans per B-scans and 500 B-scans along the y-axis, each B-scan repeated twice at each position. OCT images were converted to depth-resolved OAC representations. The en face sum projection of the slab between inner limiting membrane and Bruch’s membrane was generated for pigment segmentation (Figure 1 A, B). An adaptive threshold algorithm was applied to compensate for the difference in image appearance between SD-OCT and SS-OCTA. A user-friendly software package was developed that automatically produced the binary pigment mask and the area of the total pigment burden (Figure 2). The area measurements of pigment burden on SD-OCT and SS-OCTA were compared with manual segmentation.

Results : A total of 24 eyes from 24 patients were included. The software segmented the hyperpigmentation on both SD-OCT and SS-OCTA scans. The mean area measurements of the total pigment burden from manual segmentation among all the cases was 0.062 ± 0.056mm², while the mean area measurements from the algorithm using the SD-OCT and SS-OCTA scans were 0.096± 0.076mm² and 0.112± 0.081mm² respectively. The correlation between automated and manual segmentations of the total pigment burden was 0.83 for the SS-OCTA and 0.77 for the SD-OCT scans.(Figure1 C)

Conclusions : The automated algorithm was able to effectively and accurately identify, segment, and quantify the macular pigment burden using both SS-OCTA and SD-OCT scans. The SS-OCTA provided higher contrast images of the total pigment using OACs, which resulted in better agreement with the manual segmentations.

This abstract was presented at the 2023 ARVO Annual Meeting, held in New Orleans, LA, April 23-27, 2023.