Grading the Severity of Damage to the Perijunctional Actomyosin Ring and Zonula Occludens-1 of the Corneal Endothelium by Ensemble Learning Methods.

Purpose: In many epithelia, including the corneal endothelium, intracellular/extracellular stresses break down the perijunctional actomyosin ring (PAMR) and zonula occludens-1 (ZO-1) at the apical junctions. This study aims to grade the severity of damage to PAMR and ZO-1 through machine learning. Methods: Immunocytochemical images of PAMR and ZO-1 were drawn from recent studies on the corneal endothelium subjected to hypothermia and oxidative stress. The images were analyzed for their morphological (e.g., Hu moments) and textural features (based on gray-level co-occurrence matrix [GLCM] and Gabor filters). The extracted features were ranked by SHapley analysis and analysis of variance. Then top features were used to grade the severity of damage using a suite of ensemble classifiers, including random forest, bagging classifier (BC), AdaBoost, extreme gradient boosting, and stacking classifier. Results: A partial set of features from GLCM, along with Hu moments and the number of hexagons, enabled the classification of damage to PAMR into Control, Mild, Moderate, and Severe with the area under the receiver operating characteristics curve (AUC) = 0.92 and F1 score = 0.77 with BC. In contrast, a bank of Gabor filters provided a partial set of features that could be combined with Hu moments, branch length, and sharpness for the classification of ZO-1 images into four levels with AUC = 0.95 and F1 score of 0.8 with BC. Conclusions: We have developed a workflow that enables the stratification of damage to PAMR and ZO-1. The approach can be applied to similar data during drug discovery or pathophysiological studies of epithelia.

Automated Image Segmentation of the Corneal Endothelium in Patients With Fuchs Dystrophy.

Purpose: To perform segmentation of specular microscopy (SM) images of the corneal endothelium for comparing average perimeter length (APL) between Fuchs endothelial corneal dystrophy (FECD) patients and healthy subjects. Methods: A retrospective review of clinical records of FECD patients and those with healthy endothelium was carried out to collect images of the endothelium. The images were segmented by modified U-Net, a deep learning architecture, followed by the Watershed algorithm to resolve merged cell borders (<5%). The segmented images were analyzed for endothelial cell density (ECDUW) and APL. Results: The combination of the U-Net and Watershed algorithm, referred to as the UW approach, enabled a complete segmentation of the endothelium. In healthy, ECDUW was close to estimates by SM and manual segmentation (31 subjects; P > 0.1). However, in FECD, ECDUW was closer to estimates by manual segmentation but not by SM (27 patients; P < 0.001). ECDUW in FECD (2547 ± 499 cells/mm2; 60 patients) was smaller compared to that in the healthy (2713 ± 401 cells/mm2; 70 subjects) (P < 0.001). APL in the healthy was 66.87 ± 7.68 µm/cell (70 subjects), but it increased with %Guttae in FECD (56.60-195.30 µm/cell; 60 patients) (P < 0.0001). Conclusions: The UW approach is precise for the segmentation of SM images from the healthy and FECD. Our analysis has revealed that APL increases with %Guttae. Translational Relevance: The average perimeter length of the corneal endothelium, which represents the length of the paracellular pathway for fluid flux into the stroma, is increased in Fuchs dystrophy.

Robust and computationally efficient superresolution algorithm.

Superresolution is the process of combining information from multiple subpixel-shifted low-resolution images to form a high-resolution image. It works quite well under ideal conditions but deteriorates rapidly with inaccuracies in motion estimates. We model the original high-resolution image as a Markov random field (MRF) with a discontinuity adaptive regularizer. Given the low-resolution observations, an estimate of the superresolved image is obtained by using the iterated conditional modes (ICM) algorithm, which maximizes the local posterior conditional probability sequentially. The proposed method not only preserves edges but also lends robustness to errors in the estimates of motion and blur parameters. We derive theoretically the neighborhood structure for the posterior distribution in the presence of warping, blurring, and downsampling operations and use this to effectively reduce the overall computations. Results are given on synthetic as well as real data to validate our method.