Abstract
Purpose :
To study atrophy growth rate, its dependence on area, time and retinal regions, using an automated segmentation method in retinal SD-OCT scans.
Methods :
589 SD-OCT volumes of 97 patients (99 eyes) with nonexudative atrophic AMD and at least 1 year follow-up were selected and segmented using an automated atrophy detection algorithm. Total en face atrophic area was calculated for each scan using the segmentation output.
The square root atrophy areas were used to compute the growth rate [mm/year] by fitting a linear model to each patient and averaging across all subjects and patient subgroups: those with small (<5mm2) and large (>5mm2) baseline areas as well as those with foveal and extrafoveal atrophy at the baseline. A time-normalized growth map was constructed for each patient and averaged across the whole dataset to obtain progression growth rate with respect to retinal location.
Results :
The mean follow-up time was 43 months and the mean area at the baseline was 4.6±4.2 mm2. The average growth rate was 0.31±0.21mm/year. Despite square root transformation, the growth rate was significantly higher for patients with smaller baseline atrophies and amounted to 0.35±0.22 mm/year, than for larger ones (0.26±0.17mm/year, p=0.042). Baseline atrophy area and the square root area growth rate were correlated (R2=0.083, p=0.005). The growth rate decreased both as a function of progressing atrophy area and time since the baseline (Figure 1). Out of 99 eyes, 37 (37%) had foveal involvement at the baseline. The growth was slower for them (0.22±0.16 mm/year) than for the eyes with extrafoveal atrophy (0.38±0.22 mm/year, p=0.0002).
The topological analysis of the growth rate revealed that the atrophy progresses faster in a ring around the fovea, towards the nasal and inferior area relative to the fovea (Figure 2).
Conclusions :
The automated atrophy segmentation method enabled analysis of atrophy progression of a large patient dataset. It has shown that growth rate varies significantly across retinal regions, depending on the lesion area and time elapsed since the baseline.
This is a 2021 ARVO Annual Meeting abstract.