Determination of probability of causative pathogen in infectious keratitis using deep learning algorithm of slit-lamp images.

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.

Distinguishing Features of Anterior Uveitis Caused by Herpes Simplex Virus, Varicella-Zoster Virus, and Cytomegalovirus.

PURPOSE: To determine distinguishing features of the clinical characteristics of anterior uveitis (AU) caused by herpes simplex virus (HSV), varicella-zoster virus (VZV), and cytomegalovirus (CMV). DESIGN: Retrospective, multicenter case series. METHODS: Consecutive patients with herpetic AU examined at 11 tertiary centers in Japan between January 2012 and December 2017 and who were followed for ≥3 months were evaluated. Diagnosis was made by polymerase chain reaction (PCR) for HSV, VZV, or CMV in the aqueous humor, or classical signs of herpes zoster ophthalmicus. RESULTS: This study enrolled 259 herpetic AU patients, including PCR-proven HSV-AU (30 patients), VZV-AU (50), and CMV-AU (147), and herpes zoster ophthalmicus (32). All HSV-AU and VZV-AU patients were unilateral, while 3% of CMV-AU patients were bilateral. Most HSV-AU and VZV-AU patients were sudden onset with an acute clinical course, while CMV-AU had a more insidious onset and chronic course. There were no significant differences for all surveyed symptoms, signs, and complications between HSV-AU and VZV-AU. However, significant differences were detected for many items between CMV-AU and the other two herpetic AU types. Ocular hyperemia and pain, blurring of vision, ciliary injection, medium-to-large keratic precipitates (KPs), cells and flare in the anterior chamber, and posterior synechia significantly more often occurred in HSV-AU and VZV-AU vs CMV-AU. In contrast, small KPs, coin-shaped KPs, diffuse iris atrophy, elevated intraocular pressure, and glaucoma surgery were significantly more frequent in CMV-AU vs HSV-AU and VZV-AU. CONCLUSION: This multicenter, retrospective study identified distinguishing features of HSV-AU, VZV-AU, and CMV-AU.

Effectiveness of 16S ribosomal DNA real-time PCR and sequencing for diagnosing bacterial keratitis.

PURPOSE: To evaluate the efficacy of real-time PCR for 16S ribosomal DNA (16S r-DNA) and sequencing for diagnosing microbial keratitis. METHODS: We studied 272 eyes of 272 patients with keratitis. Eyes with keratitis were classified as "definite" (N = 118), "likely" (N = 71), or "non-bacterial" (N = 83) to have bacterial keratitis. The diagnostic efficacy of real-time PCR and conventional testing was determined by receiver operating characteristic analysis. The copy numbers of bacterial DNA and clinical characteristics were retrospectively analyzed for association with concordant culture results in the "definite" cases. RESULTS: The level of bacterial DNA was significantly associated with the diagnostic probability of the three diagnostic categories. The level of bacterial DNA had comparable diagnostic efficacy with the area under the curve (AUC) at 0.67, by culture at 0.65, and by smear testing at 0.73. The efficacy was significantly improved by combining the DNA level with the conventional culture testing with an AUC of 0.81. Analysis of the "definite" cases showed culture positivity in 51.8% (58 eyes), and of these, 41 eyes (70.7%) were higher than the cutoff PCR values and 40 eyes were identified by 16S r-DNA sequencing. In the culture-negative eyes, the level of bacterial DNA was significantly lower (P = 0.0008). Eyes with higher bacterial DNA levels had significantly concordant outcomes with sequencing and culture results (P = 0.006). Previous antibiotic treatments decreased the bacterial DNA amount by 0.09-fold, and it was a significant factor for discordance (P = 0.006). CONCLUSION: Quantification of the bacterial DNA level and conventional testing improves the diagnostic efficacy of infectious bacterial keratitis.

High interleukin-8 level in aqueous humor is associated with poor prognosis in eyes with open angle glaucoma and neovascular glaucoma.

Glaucoma is a leading cause of blindness worldwide. Purpose of this study was to identify molecular markers that were significantly correlated with presence of glaucoma and outcome of glaucoma surgery. To accomplish this, we determined the profiles of pro-inflammatory cytokines in the aqueous humor of 101 glaucoma patients; 31 primary open angle glaucoma (POAG), 38 pseudoexfoliation glaucoma (PEG), and 32 neovascular glaucoma (NVG). We also studied 100 normal subjects as controls. In eyes with POAG or PEG, the level of interleukin (IL)-1α, IL-2, IL-4, IL-8, IL-23, and CCL2 were significantly elevated. In the NVG eyes, many inflammatory cytokines were also highly elevated. IL-8 had the highest odds ratio, and levels of IL-8 and CCL2 were significantly correlated with preoperative IOP or visual field defects in PEG eyes. Principal component analysis showed that IL-8 had the highest association to the IOP-cytokine component, and Cox proportional hazard model indicated that an elevation of IL-8 was a significant risk of filtering surgery failure. Together with modeling of their interactions and prognosis, IL-8 elevation is a significant risk factor both for detecting and managing glaucoma and may serve as a therapeutic target candidate to improve the prognosis of glaucoma surgery.