Automated Photoreceptor Cell Identification on Nonconfocal Adaptive Optics Images Using Multiscale Circular Voting

Abstract

PurposeAdaptive optics scanning light ophthalmoscopy (AOSLO) has enabled quantification of the photoreceptor mosaic in the living human eye using metrics such as cell density and average spacing. These rely on the identification of individual cells. Here, we demonstrate a novel approach for computer-aided identification of cone photoreceptors on nonconfocal split detection AOSLO images.MethodsAlgorithms for identification of cone photoreceptors were developed, based on multiscale circular voting (MSCV) in combination with a priori knowledge that split detection images resemble Nomarski differential interference contrast images, in which dark and bright regions are present on the two sides of each cell. The proposed algorithm locates dark and bright region pairs, iteratively refining the identification across multiple scales. Identification accuracy was assessed in data from 10 subjects by comparing automated identifications with manual labeling, followed by computation of density and spacing metrics for comparison to histology and published data.ResultsThere was good agreement between manual and automated cone identifications with overall recall, precision, and F1 score of 92.9%, 90.8%, and 91.8%, respectively. On average, computed density and spacing values using automated identification were within 10.7% and 11.2% of the expected histology values across eccentricities ranging from 0.5 to 6.2 mm. There was no statistically significant difference between MSCV-based and histology-based density measurements (P = 0.96, Kolmogorov-Smirnov 2-sample test).ConclusionsMSCV can accurately detect cone photoreceptors on split detection images across a range of eccentricities, enabling quick, objective estimation of photoreceptor mosaic metrics, which will be important for future clinical trials utilizing adaptive optics.

Cone Photoreceptor Cell Segmentation and Diameter Measurement on Adaptive Optics Images Using Circularly Constrained Active Contour Model

“… 34 Image sequences were corrected for eye motion 35 and manually assembled into montages that included both confocal and split detection images, as previously described. 17 In this study, we imaged 10 subjects with no history of systemic or ocular disease (five female, five male; age range, 22–40 years; mean ± SD, 26.3 ± 5.6 years; additional information in Supplementary Table S1 ) and also a patient with late-onset retinal degeneration (male, 55 years). For each subject, over 100 retinal locations were imaged, from the fovea out to an eccentricity of approximately 6 mm in the temporal direction.…”

“… 1 – 3 Quantitative assessment of the mosaic through metrics on AO retinal images, such as cone density and spacing, has shown potential for clinical application 2 , 4 with substantial efforts already realized toward assembling normative databases. 5 – 12 To overcome the tedious task of manually identifying cones and to remove the variability of human graders, various automated algorithms have been developed for two types of AO modalities: confocal 13 – 16 and nonconfocal 17 – 19 AO scanning light ophthalmoscopy (AOSLO). However, most quantitative metrics have been based on representing each cone as a point.…”

“…This three-dimensional appearance with oblique illumination also poses an interesting dilemma on defining the precise location of the boundary, which we will discuss later. To date, there have been very few articles addressing image analysis of split detection images, 17 , 18 , 22 , 31 , 32 with only one report of cone photoreceptor segmentation 22 (defined as the extraction of the region occupied by cells as opposed to identification of cell centers).…”

Longitudinal adaptive optics fluorescence microscopy reveals cellular mosaicism in patients

“…Eye motion was corrected after image acquisition (47) using one of the simultaneously acquired channels (3). Averaged AO images were assembled into montages of overlapping retinal regions; then, ROIs were selected and scaled for analysis, as previously described (48). Longitudinal imaging datasets were registered within visits and then manually overlaid on top of each other based on vascular features and other landmarks.…”

Noninvasive near infrared autofluorescence imaging of retinal pigment epithelial cells in the human retina using adaptive optics

“…Cone photoreceptors were identified semi-automatically, based on previously-published algorithms combined with manual correction [27][28][29][30]. In the fovea, a smaller sampling window (37µm × 37µm) within the initial ROI was selected to mitigate the effects of the large variation in cone density in the fovea [31,32].…”