Abstract
Purpose: :
To test the repeatability and reproducibility of cone mosaic imaging in normal subjects using a commercial Adaptive Optics Flood Illuminated Camera in the clinical setting.
Methods: :
Using the rtx1 Adaptive Optics Flood Illuminated Camera (Imagine Eyes, Orsay, France), images were collected in 3 normal volunteers (ages 27 to 36) at 3 separate time points with and without cycloplegia. Axial lengths were obtained to correct for differences in image magnification. In each imaging session, a series of 25 retinal images were obtained, which spanned a 12 degree by 12 degree field of the central retina. Retinal montages were first created manually with Adobe Photoshop and subsequently in an automated fashion with i2k Retina software. To assess the accuracy of an automated cone counting algorithm, sample retinal areas were selected 2 and 4 degrees temporal to the foveal center. Manual counts were performed by two independent observers and compared to the automatic cone counting algorithm written in Matlab (courtesy of Dr. Joseph Carroll). Additionally, the effect of manual versus automated montaging was evaluated using similar methods.
Results: :
12x12 degree montages of the cone mosaic were obtained in both emmetropes and myopes in under 30 minutes. The rtx1 was unable to resolve cones within a 2 degree radius from the foveal center and this area was excluded from analysis. With automated counting, at 2 degrees from the foveal center the average cone spacing between subjects ranged from 25,910 to 31,208 cones/mm2, while at 4 degrees from the foveal center the average cone spacing ranged from 14,561 to 21,124 cones/mm2. For the three subjects, automated counting correlated well with manual counting. For example, manual counting from the image 4 degrees from the foveal center yielded cone counts ranging from 14,759 to 20,151 cones/mm2. Manual counting from the image 2 degrees from the foveal center yielded cone counts ranging from 24,810 to 30,900 cones/mm2. Cycloplegia did not appear to affect the cone counts; however, subjects found it easier to fixate with cycloplegia. Automated montaging introduces artifacts in the final image compared to manual montaging, but the effect on cone counts was minimal.
Conclusions: :
We demonstrate that imaging cone photoreceptors with an Adaptive Optics Flood Illuminated Camera is feasible, quick, and reliable in the clinic setting. As a result, we now have the potential to utilize AO in direct patient care and research. Future studies will aim to compile an age-adjusted database of retinal montages from normal subjects to be used with clinical trials studying retinal disorders.
Keywords: imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound) • image processing • retina