Deep learning segmentation of endothelial cell images using an active learning paradigm with guided label corrections.

Purpose: To create Guided Correction Software for informed manual editing of automatically generated corneal endothelial cell (EC) segmentations and apply it to an active learning paradigm to analyze a diverse set of post-keratoplasty EC images. Approach: An original U-Net model trained on 130 manually labeled post-Descemet stripping automated endothelial keratoplasty (EK) images was applied to 841 post-Descemet membrane EK images generating "uncorrected" cell border segmentations. Segmentations were then manually edited using the Guided Correction Software to create corrected labels. This dataset was split into 741 training and 100 testing EC images. U-Net and DeepLabV3+ were trained on the EC images and the corresponding uncorrected and corrected labels. Model performance was evaluated in a cell-by-cell analysis. Evaluation metrics included the number of over-segmentations, under-segmentations, correctly identified new cells, and endothelial cell density (ECD). Results: Utilizing corrected segmentations for training U-Net and DeepLabV3+ improved their performance. The average number of over- and under-segmentations per image was reduced from 23 to 11 with the corrected training set. Predicted ECD values generated by networks trained on the corrected labels were not significantly different than the ground truth counterparts (p=0.02, paired t-test). These models also correctly segmented a larger percentage of newly identified cells. The proposed Guided Correction Software and semi-automated approach reduced the time to accurately segment EC images from 15 to 30 to 5 min, an ∼80% decrease compared to manual editing. Conclusions: Guided Correction Software can efficiently label new training data for improved deep learning performance and generalization between EC datasets.

Subclinical increased anterior stromal reflectivity with topical taprenepag isopropyl.

PURPOSE: To assess the effect of topical taprenepag isopropyl on each layer of the cornea by confocal microscopy. METHODS: Thirty-two ocular hypertensive or glaucoma patients were randomized into a 2-period, crossover study of 14 days of 0.1% taprenepag alone and in unfixed combination with 0.005% latanoprost (combination therapy). Baseline and sequential slit-lamp biomicroscopy, fluorescein staining, central ultrasonic pachymetry, and confocal microscopy were performed. Confocal images were analyzed for the density of the central superficial and basal epithelium, midstromal keratocytes, and endothelium, as well as endothelial coefficient of variation and percentage of hexagonal cells, and reflectivity of anterior stromal and midstromal layers. RESULTS: Corneal staining increased from baseline, reaching a peak at day 13 (69% and 63% of subjects treated with monotherapy and combination therapy, respectively), which resolved by day 35. A statistically significant increase in mean corneal thickness for both eyes and both treatments occurred on days 7 and 13 (range, 20-27 µm; P < 0.001) but recovered (≤ 6 µm) by day 35. No statistically significant changes were observed in the basal epithelial, midstromal, or endothelial cells. Mean ratio of average reflectivity of anterior stroma to midstroma increased on days 13 and 35 in period 1 for each treatment (range, 1.2-1.9; P < 0.001), and this increase persisted during period 2. CONCLUSIONS: Anterior stromal reflectivity may remain increased even when biomicroscopic and confocal images of corneal layers remain normal or have recovered after topical taprenepag. This subclinical measure may be useful to detect a persistent adverse effect of a topical agent on the cornea.