Accurate Correspondence of Cone Photoreceptor Neurons in the Human Eye Using Graph Matching Applied to Longitudinal Adaptive Optics Images

Automated Iterative Label Transfer Improves Segmentation of Noisy Cells in Adaptive Optics Retinal Images

High quality data labeling is essential for improving the accuracy of deep learning applications in medical imaging. However, noisy images are not only under-represented in training datasets, but also, labeling of noisy data is low quality. Unfortunately, noisy images with poor quality labels are exacerbated by traditional data augmentation strategies. Real world images contain noise and can lead to unexpected drops in algorithm performance. In this paper, we present a non-traditional, purposeful data augmentation method to specifically transfer high quality automated labels into noisy image regions for incorporation into the training dataset. The overall approach is based on the use of paired images of the same cells in which variable image noise results in cell segmentation failures. Iteratively updating the cell segmentation model with accurate labels of noisy image areas resulted in an improvement in Dice coefficient from 77% to 86%. This was achieved by adding only 3.4% more cells to the training dataset, showing that local label transfer through graph matching is an effective augmentation strategy to improve segmentation.

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

Purpose

Cone photoreceptor cells can be noninvasively imaged in the living human eye by using nonconfocal adaptive optics scanning ophthalmoscopy split detection. Existing metrics, such as cone density and spacing, are based on simplifying cone photoreceptors to single points. The purposes of this study were to introduce a computer-aided approach for segmentation of cone photoreceptors, to apply this technique to create a normal database of cone diameters, and to demonstrate its use in the context of existing metrics.

Methods

Cone photoreceptor segmentation is achieved through a circularly constrained active contour model (CCACM). Circular templates and image gradients attract active contours toward cone photoreceptor boundaries. Automated segmentation from in vivo human subject data was compared to ground truth established by manual segmentation. Cone diameters computed from curated data (automated segmentation followed by manual removal of errors) were compared with histology and published data.

Results

Overall, there was good agreement between automated and manual segmentations and between diameter measurements (n = 5191 cones) and published histologic data across retinal eccentricities ranging from 1.35 to 6.35 mm (temporal). Interestingly, cone diameter was correlated to both cone density and cone spacing (negatively and positively, respectively; P < 0.01 for both). Application of the proposed automated segmentation to images from a patient with late-onset retinal degeneration revealed the presence of enlarged cones above individual reticular pseudodrusen (average 23.0% increase, P < 0.05).

Conclusions

CCACM can accurately segment cone photoreceptors on split detection images across a range of eccentricities. Metrics derived from this automated segmentation of adaptive optics retinal images can provide new insights into retinal diseases.

Longitudinal adaptive optics fluorescence microscopy reveals cellular mosaicism in patients

The heterogeneity of individual cells in a tissue has been well characterized, largely using ex vivo approaches that do not permit longitudinal assessments of the same tissue over long periods of time. We demonstrate a potentially novel application of adaptive optics fluorescence microscopy to visualize and track the in situ mosaicism of retinal pigment epithelial (RPE) cells directly in the human eye. After a short, dynamic period during which RPE cells take up i.v.-administered indocyanine green (ICG) dye, we observed a remarkably stable heterogeneity in the fluorescent pattern that gradually disappeared over a period of days. This pattern could be robustly reproduced with a new injection and follow-up imaging in the same eye out to at least 12 months, which enabled longitudinal tracking of RPE cells. Investigation of ICG uptake in primary human RPE cells and in a mouse model of ICG uptake alongside human imaging corroborated our findings that the observed mosaicism is an intrinsic property of the RPE tissue. We demonstrate a potentially novel application of fluorescence microscopy to detect subclinical changes to the RPE, a technical advance that has direct implications for improving our understanding of diseases such as oculocutaneous albinism, late-onset retinal degeneration, and Bietti crystalline dystrophy.

COMPUTER-AIDED DETECTION OF PATTERN CHANGES IN LONGITUDINAL ADAPTIVE OPTICS IMAGES OF THE RETINAL PIGMENT EPITHELIUM

Retinal pigment epithelium (RPE) defects are indicated in many blinding diseases, but have been difficult to image. Recently, adaptive optics enhanced indocyanine green (AO-ICG) imaging has enabled direct visualization of the RPE mosaic in the living human eye. However, tracking the RPE across longitudinal images on the time scale of months presents with unique challenges, such as visit-to-visit distortion and changes in image quality. We introduce a coarse-to-fine search strategy that identifies paired patterns and measures their changes. First, longitudinal AO-ICG image displacements are estimated through graph matching of affine invariant maximal stable extremal regions in affine Gaussian scale-space. This initial step provides an automatic means to designate the search ranges for finding corresponding patterns. Next, AO-ICG images are decomposed into superpixels, simplified to a pictorial structure, and then matched across visits using tree-based belief propagation. Results from human subjects in comparison with a validation dataset revealed acceptable accuracy levels for the level of changes that are expected in clinical data. Application of the proposed framework to images from a diseased eye demonstrates the potential clinical utility of this method for longitudinal tracking of the heterogeneous RPE pattern.