In this study, we aimed to develop a deep learning model for identifying bacterial keratitis (BK) and fungal keratitis (FK) by using slit-lamp images. We retrospectively collected slit-lamp images of patients with culture-proven microbial keratitis between 1 January 2010 and 31 December 2019 from two medical centers in Taiwan. We constructed a deep learning algorithm consisting of a segmentation model for cropping cornea images and a classification model that applies different convolutional neural networks (CNNs) to differentiate between FK and BK. The CNNs included DenseNet121, DenseNet161, DenseNet169, DenseNet201, EfficientNetB3, InceptionV3, ResNet101, and ResNet50. The model performance was evaluated and presented as the area under the curve (AUC) of the receiver operating characteristic curves. A gradient-weighted class activation mapping technique was used to plot the heat map of the model. By using 1330 images from 580 patients, the deep learning algorithm achieved the highest average accuracy of 80.0%. Using different CNNs, the diagnostic accuracy for BK ranged from 79.6% to 95.9%, and that for FK ranged from 26.3% to 65.8%. The CNN of DenseNet161 showed the best model performance, with an AUC of 0.85 for both BK and FK. The heat maps revealed that the model was able to identify the corneal infiltrations. The model showed a better diagnostic accuracy than the previously reported diagnostic performance of both general ophthalmologists and corneal specialists.

We conducted a retrospective, cross-sectional study to analyze predisposing factors, clinical features, and microbiological characteristics of patients with microbial keratitis hospitalized over 10 years. The medical records of 558 patients who were diagnosed with microbial keratitis and admitted to Chang Gung Memorial Hospital (CGMH), a referral center in Taiwan, from January 1, 2003 to December 31, 2012 were reviewed. Demographics, predisposing factors, isolated organisms, treatment, and hospital stay were recorded. Yearly trends were tested using a linear-by-linear association. Contact lens wear was the most common predisposing factor (31.4%), followed by ocular and systemic diseases (26.3%) and trauma (23.5%). Contact lens-related infectious keratitis increased year by year (P = 0.011). Pseudomonas aeruginosa was the most commonly isolated organism (28%), followed by fungi (17.6%) and coagulase-negative Staphylococcus (5.4%). Except for Serratia marcescens, the identified organisms did not change over 10 years. Most bacterial infections were controlled using antimicrobial treatment, but more than half of patients with fungal keratitis required surgical interventions. The mean hospital stay was 13.7 ± 11.5 days. Previous ocular surgery, large ulcer size, nontuberculous myycobacteris infection, and surgery during admission were related to prolonged hospital stay. In Taiwan, contact lens-related pseudomonal keratitis remained the most common cause of microbial keratitis in patients hospitalized from 2003 to 2012.

Methicillin-resistant Staphylococcus aureus (MRSA) infection is an important public health issue. The study aimed to characterize the patient demographics, clinical features, antibiotic susceptibility, and clinical outcomes of keratitis caused by S. aureus, and to make a comparison between MRSA and methicillin-sensitive S. aureus (MSSA) isolates. Patients (n = 59) with culture-proven S. aureus keratitis treated in Chang Gung Memorial Hospital between January 1, 2006, and December 31, 2010, were included in our study. Patients' demographic and clinical data were retrospectively reviewed. Twenty-six MRSA (44%) and 33 MSSA (56%) isolates were collected. The MRSA keratitis was significantly more common among the patients with healthcare exposure (P = 0.038), but 46.2% (12/26) of patients with MRSA keratitis were considered to have community-associated infections. All isolates were susceptible to vancomycin. MRSA isolates were significantly more resistant to clindamycin, erythromycin, and sulfamethoxazole/trimethoprim. Ocular surface disease was a significant risk factor for MRSA keratitis (P = 0.011). Visual outcome did not differ significantly between the MRSA and MSSA groups. However, age (B = 0.01, P = 0.035, 95% confidence interval [CI]: 0.001–0.019) and visual acuity at presentation (B = 0.749, P<0.001, 95% CI: 0.573–0.926) were significantly correlated with visual outcome. Ocular surface disease is an important predisposing factor for S. aureus keratitis, especially for MRSA infections. Advanced age and poor visual acuity at presentation are important prognostic indicators for poor visual outcome in S. aureus keratitis. Oxacillin resistance may not be a significant prognostic indicator.