Abstract
Purpose :
Type 2 Macular Telangiectasia (MacTel) is a disease of the central retina, leading over time to photoreceptor (PR) degeneration and potentially to bilateral loss of central vision. SD-OCT 'en face' area measurement of an Ellipsoid Zone (EZ) break is an established method for following disease progression in MacTel, however it is not an early sign in the natural history of MacTel. Finding biomarkers that provide quantitative measures over a wider range of disease severities may be useful for following disease progression in trials of potential new therapies. Our aim was to assess the feasibility and utility of measuring PR layer thickness in MacTel.
Methods :
Heidelberg Spectralis OCT volume scans of 22 eyes with early MacTel (EZ break absent or <0.16mm2) of 14 participants in the Mactel Natural History Observation Registry Study were included. OCT volume scans of 24 healthy eyes were obtained from publicly available data. A novel five-boundary segmentation algorithm based on Eff-UNet was trained using 5238 B-scans of healthy, MacTel, DME and AMD eyes. Boundaries segmented were the anterior surfaces of the ILM, OPL, EZ and the RPE and the posterior surface of the RPE. The photoreceptor layer (PR) was approximated as the slab defined by the anterior surfaces of the OPL and the RPE. Normative retinal layer thickness maps were created by alignment of all volumes of healthy eyes, with the mean and standard deviation of each layer calculated at each location within the volume. MacTel volumes were then aligned to the normative database and a Z-score was calculated at each location. Abnormal areas were defined as a thinning >1.96*SD of the mean.
Results :
The segmentation algorithm achieved an intersection over union (IoU) of 0.93 across all OCT layers in the MacTel sample. Automated segmentation was able to detect a thinning in the PR in all cases detected visually. The area of PR thinning correlated significantly with the presence of an EZ break (r2 = 0.62, p =0.023) and the increase over time in the area of PR thinning was larger than that in the area of the EZ break.
Conclusions :
A thinning of the PR layer appears to precede the incidence and exceed the extent of an EZ break. Measuring PR thickness may offer a tool for quantifying disease severity prior to formation of an EZ break. Our results are encouraging and warrant further investigations on a larger sample made possible by automated analysis.
This abstract was presented at the 2023 ARVO Annual Meeting, held in New Orleans, LA, April 23-27, 2023.