Corneal opacities are important causes of blindness, and their major etiology is infectious keratitis. Slit-lamp examinations are commonly used to determine the causative pathogen; however, their diagnostic accuracy is low even for experienced ophthalmologists. To characterize the “face” of an infected cornea, we have adapted a deep learning architecture used for facial recognition and applied it to determine a probability score for a specific pathogen causing keratitis. To record the diverse features and mitigate the uncertainty, batches of probability scores of 4 serial images taken from many angles or fluorescence staining were learned for score and decision level fusion using a gradient boosting decision tree. A total of 4306 slit-lamp images including 312 images obtained by internet publications on keratitis by bacteria, fungi, acanthamoeba, and herpes simplex virus (HSV) were studied. The created algorithm had a high overall accuracy of diagnosis, e.g., the accuracy/area under the curve for acanthamoeba was 97.9%/0.995, bacteria was 90.7%/0.963, fungi was 95.0%/0.975, and HSV was 92.3%/0.946, by group K-fold validation, and it was robust to even the low resolution web images. We suggest that our hybrid deep learning-based algorithm be used as a simple and accurate method for computer-assisted diagnosis of infectious keratitis.

The purpose of this study was to identify the inflammatory cytokines that were associated with pachychoroid neovasculopathy (PNV). Seventy-five eyes of 75 patients with PNV, 145 eyes of 145 patients with neovascular age-related macular degeneration without pachyvessels, and 150 eyes of 150 normal subjects were examined for the levels of intraocular cytokines. In eyes with PNV, the levels of IL-1α, IL-1β, IL-2, IL-4, IL-10, and VEGF were significantly higher than that of the controls. Logistic regression analysis showed that the highest association with the pachyvessels was found for IL-4, IL-2, and IL-1α. In eyes with PNV, the levels of IL-4, IL-2, IL-5, IL-13, IL-1α, and IL-1β were significantly higher in eyes with both increased choroidal thickness and choroidal vessel diameter. The strongest correlation with the choroidal thickness and vessel diameter was observed for IL-4. In PNV eyes with polypoidal lesions, the levels of IL-4, IL-17, and TNFβ were significantly correlated with the number of polypoidal lesions. Of these cytokines, IL-4 was especially associated with the thickness of the choroidal vessels and the formation of polypoidal lesions. We conclude that IL-4 is most likely involved in establishing the clinical characteristics of PNV and polypoidal vascular remodeling.

The purpose of this study was to identify inflammatory cytokines that are associated with pachychoroid neovasculopathy (PNV). Seventy-five eyes of 75 patients with PNV, 145 eyes of 145 patients with neovascular age-related macular degeneration without pachyvessels, and 150 eyes of 150 normal subjects were examined for the levels of intracameral cytokines. In eyes with PNV, IL-1α, IL-1β, IL-2, IL-4, IL-10, and VEGF were significantly elevated compared to controls. Logistic regression analysis indicated highest association with pachyvessels was observed for IL-4, IL-2, and IL-1α. In eyes with PNV, IL-4, TNFα, IL-17, IL-2, IL-12, IL-15, IL-5, IL-13, IL-1α, and IL-1β significantly increased choroidal thickness. Highest correlation with choroidal thickness was observed for IL-4. In PNV eyes with polypoidal lesions, the level of IL-4, IL-17, and TNFβ significantly correlated with the number of polypoidal lesions. We determined how the different disease characteristics of PNV were associated with the elevated cytokines. Of all these cytokines, IL-4 contributed significantly to the thickening of the choroidal vessels and to the formation of polypoidal lesions. We conclude that IL-4 is most likely involved in the establishing the clinical characteristics of PNV and polypoidal vascular remodeling. This may help to establish future therapeutic strategy for PNV.

Ocular cytomegalovirus (CMV) infections in immunocompetent individuals are rare, but its activation can cause chronic and relapsing inflammation in anterior segment of the eye resulting in loss of corneal clarity and glaucoma. Fifty five patients with anterior segment CMV infection were assessed for their clinical characteristics, and CMV corneal endotheliitis was found to cause significant loss of corneal endothelial cells. The disease duration with recurrences was significantly correlated with the maximum intraocular level of CMV DNA. To examine why CMV is activated in healthy immunocompetent individuals and causing corneal endothelial cell damage, assays of cytotoxic T cells (CTLs) which directly target infected corneal endothelial cells were performed for 9 HLA-matched CMV corneal endotheliitis patients (HLA-A*2402). When the cell loss was analyzed for associations with CTL responses, CMV-induced endothelial cell damage was mitigated by pp65-specific CTL induction. The recurrence-free time was also prolonged by pp65-specific CTL induction (hazard ratio (HR): 0.93, P = 0.01). In contrast, IE1-specific CTL was associated with endothelial cell damage and reduced the time for corneal transplantation (HR: 1.6, P = 0.003) and glaucoma surgery (HR: 1.5, P = 0.001). Collectively, induction of pp65-specific CTL was associated with improved visual prognosis. However, IE1-specific CTL without proper induction of pp65-specific CTL can cause pathological damage leading to the need of surgical interventions.

Corneal opacities are an important cause of blindness, and its major etiology is infectious keratitis. Slit-lamp examinations are commonly used to determine the causative pathogen; however, their diagnostic accuracy is low even for experienced ophthalmologists. To characterize the “face” of an infected cornea, we have adapted a deep learning architecture used for facial recognition and applied it to determine a probability score for a specific pathogen causing keratitis. To record the diverse features and mitigate the uncertainty, batches of probability scores of 4 serial images taken from many angles or fluorescence staining were learned for score and decision level fusion using a gradient boosting decision tree. A total of 4306 slit-lamp images and 312 images obtained by internet publications on keratitis by bacteria, fungi, acanthamoeba, and herpes simplex virus (HSV) were studied. The created algorithm had a high overall accuracy of diagnosis, e.g., the accuracy/area under the curve (AUC) for acanthamoeba was 97.9%/0.995, bacteria was 90.7%/0.963, fungi was 95.0%/0.975, and HSV was 92.3%/0.946, by group K-fold validation, and it was robust to even the low resolution web images. We suggest that our hybrid deep learning-based algorithm be used as a simple and accurate method for computer-assisted diagnosis of infectious keratitis.