Abstract
Purpose :
Adaptive optics (AO) retinal imaging has enabled visualization of the photoreceptor mosaic, however quantification of mosaic parameters has generally been limited to discretely sampled regions of interest (ROIs). Here, we automatically estimate pointwise inter-cell distance (ICD) and cone density across retinal montages, with the intent of building a normative database from every retinal locus within a montage.
Methods :
Confocal AO image sequences of the human photoreceptor mosaic were acquired at retinal locations surrounding the fovea and extending ~1.6 mm along each meridian in 20 eyes from 20 normal-sighted controls using a custom built AO scanning light ophthalmoscope. Image sequences were registered, distortions from eye motion were removed, and images were montaged. Montages were resized to a common scale between subjects. ICD over each subject’s montage was determined by extracting a grid of ROIs from the montages, taking the discrete Fourier transform (DFT) for each ROI, and automatically finding the ICD and cone density corresponding to the peak of the DFT modal ring using a custom MATLAB algorithm. ICD and density estimates from grid locations with multiple overlapping images were combined using a weighted average based on estimation quality, where quality was determined by the distinctiveness of the modal ring. Grid locations with quality assessments less than two standard deviations less than the mean quality were excluded from further analysis for each subject. Finally, ICD and density from all subjects at all grid locations were collated and the mean ICD and density was calculated across subjects.
Results :
Mean ICD and density across all 20 subjects was 4.3, 5.4, 7.3, and 9.3 µm and 68,000, 39,000, 21,000, and 13,000 cones/mm2 at 0.2, 0.4, 0.8, and 1.6 mm temporal to fixation, respectively (standard deviation: 0.4, 0.4, 0.5, 0.6 µm, and 13,000, 5,300, 2,700, and 2,000 cones/mm2 respectively). Similar to previous reports, ICD was ~10% higher along superior and inferior meridians relative to the temporal and nasal meridians. Our estimation quality metric routinely excluded grid locations within the central 300 µm horizontally and 260 µm vertically.
Conclusions :
We created a normative database of pointwise ICD and density across individuals using automated analysis of AO montages. This database will be useful for future studies comparing diseased and normal retina at precisely matched retinal locations.
This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.