Comparisons of deep learning algorithms for diagnosing bacterial keratitis via external eye photographs

Bacterial keratitis (BK), a painful and fulminant bacterial infection of the cornea, is the most common type of vision-threatening infectious keratitis (IK). A rapid clinical diagnosis by an ophthalmologist may often help prevent BK patients from progression to corneal melting or even perforation, but many rural areas cannot afford an ophthalmologist. Thanks to the rapid development of deep learning (DL) algorithms, artificial intelligence via image could provide an immediate screening and recommendation for patients with red and painful eyes. Therefore, this study aims to elucidate the potentials of different DL algorithms for diagnosing BK via external eye photos. External eye photos of clinically suspected IK were consecutively collected from five referral centers. The candidate DL frameworks, including ResNet50, ResNeXt50, DenseNet121, SE-ResNet50, EfficientNets B0, B1, B2, and B3, were trained to recognize BK from the photo toward the target with the greatest area under the receiver operating characteristic curve (AUROC). Via five-cross validation, EfficientNet B3 showed the most excellent average AUROC, in which the average percentage of sensitivity, specificity, positive predictive value, and negative predictive value was 74, 64, 77, and 61. There was no statistical difference in diagnostic accuracy and AUROC between any two of these DL frameworks. The diagnostic accuracy of these models (ranged from 69 to 72%) is comparable to that of the ophthalmologist (66% to 74%). Therefore, all these models are promising tools for diagnosing BK in first-line medical care units without ophthalmologists.

Contact Lens Associated Bacterial Keratitis: Common Organisms, Antibiotic Therapy, and Global Resistance Trends: A Systematic Review

IntroductionContact lens wearing has been increased globally during recent decades, which is one of the main risk factors for developing microbial keratitis. Microbial keratitis is a severe and dangerous condition that causes cornea inflammation. It can lead to corneal scarring and perforation or even endophthalmitis and visual loss if it remains untreated. Among bacterial, fungal, protozoal, and viral agents which can cause microbial keratitis, bacteria are the most common cause. Therefore, in this study, we aim to find common causative bacteria, sensitivity, and resistance to antibiotics and the outcome of antibiotic therapy in contact lens-related bacterial keratitis.MethodsA systematic search was carried out in PubMed/Medline, EMBASE, and Web of Science for published studies and medRxiv for preprints up to February 30, 2021, and May 14, 2021, respectively. A combination of the following keywords was used: “Infection”, “Corneal infection”, “Keratitis”, “Microbial keratitis”, and “Contact lens”, Also, we used the “Contact lenses” MeSH term. Lists of references for each selected article and relevant review articles were hand-searched to identify further studies.ResultsTwenty-six articles were included. From 1991 to 2018, 2,916 episodes of contact lens-related microbial keratitis) CLMK(with 1,642 episodes of proven bacterial keratitis have been reviewed in these studies. Studies were conducted in 17 countries with different geographical regions, and four studies were conducted in Iran, which is the highest number of studies among these countries. According to 20 studies, the mean age of patients was 30.77 years. Females with 61.87% were more than males in 19 studies. A percentage of 92.3% of patients used soft contact lenses, and 7.7% of patients used hard contact lenses (including RGP), according to 16 studies. Pseudomonas aeruginosa, Staphylococcus spp., and Serratia marcescens were the three most common bacteria isolated from samples of patients with contact lens-related bacterial keratitis. Overall, isolated bacteria were most sensitive to fluoroquinolones and aminoglycosides, especially ciprofloxacin and gentamicin respectively, and most resistant against penicillin and cephalosporins especially cefazolin and chloramphenicol. Almost all patients responded well to antibiotic therapy, with some exceptions that needed further surgical interventions.ConclusionAntibiotics are efficient for treating almost all patients with contact lens-related bacterial keratitis if they are appropriately chosen based on common germs in every geographical region and the sensitivity and resistance of these germs against them. In this regard, Pseudomonas aeruginosa is the most common causative germ of contact lens-associated bacterial keratitis all over the world and is almost fully sensitive to ciprofloxacin. Because of some different results about the sensitivity and resistance of germs against some antibiotics like gentamicin, vancomycin, and chloramphenicol in the Middle East region, especially Iran, more in vitro and clinical studies are suggested.

Albumin in Tears Modulates Bacterial Susceptibility to Topical Antibiotics in Ophthalmology

Bacterial keratitis is a serious and vision-threatening condition in veterinary and human patients, one that often requires culture and susceptibility testing to adjust therapy and improve clinical outcomes. The present study challenges the antimicrobial susceptibility testing (AST) paradigm in ophthalmology, enabling more accurate in vitro-to-in vivo translation by incorporating factors normally present during host-pathogen interactions in clinical patients. Thirty bacteria (10 Staphylococcus pseudintermedius, 10 Streptococcus canis, 10 Pseudomonas aeruginosa) were isolated from canine patients with infectious keratitis. For each isolate, commercial plates (Sensititre™ JOEYE2) were used to assess the minimal inhibitory concentration (MIC) of 17 different antibiotics in the absence (0% albumin, control) or presence of canine albumin (0.01–2%). For Staphylococcus pseudintermedius, the experiment was repeated with actual tear fluid collected from canine eyes with ocular surface inflammation. Kruskal-Wallis, Wilcoxon signed rank test and Spearman's correlation tests were used for statistical analysis. Clinical outcomes were unfavorable in selected canine patients with bacterial keratitis (e.g., globe perforation, graft dehiscence) despite standard AST (i.e., 0% albumin in test medium) confirming that most corneal infections (93%) were susceptible to ≥1 topical antibiotics used at the initial visit. Albumin levels ≥0.05% increased MICs in a dose-dependent, bacteria-specific, and antibiotic-specific manner. No significant differences (P = 1.000) were noted in MICs of any antibiotic whether albumin or tear fluid was added to the Mueller-Hinton broth. Percent protein binding inherent to each antibiotic was associated with clinical interpretations (Spearman's rho = −0.53, P = 0.034) but not changes in MICs. Albumin in tears impacted the efficacy of selected ophthalmic antibiotics as only the unbound portion of an antibiotic is microbiologically active. The present findings could improve decision making of clinicians managing bacterial keratitis, reduce development of antimicrobial resistance, influence current guidelines set by the Clinical and Laboratory Standards Institute, and serve as a reference for bacteriological evaluations across medical fields and across species.

Comparative Clinical Analysis of Polymicrobial and Monomicrobial Bacterial Keratitis

Purpose: We comparatively analyzed the microbiological profiles, predisposing factors, clinical aspects, and treatment outcomes of patients with polymicrobial and monomicrobial bacterial keratitis.Methods: A total of 194 cases of culture-proven bacterial keratitis treated between January 2007 and December 2016 were reviewed. Microbiological profiles, the epidemiology, predisposing factors, clinical characteristics, and treatment outcomes were compared between the polymicrobial group (polymicrobial bacterial keratitis [PBK]; 29 eyes, 62 isolates) and monomicrobial (monomicrobial bacterial keratitis [MBK]; 165 eyes, 165 isolates) group.Results: The most common isolates were Enterobacter (24%) in the PBK group and Staphylococcus (22%) in the MBK group. There were no significant differences between the two groups in previous ocular surface disease, previous ocular surgery, prior topical steroid use, epithelial defect size, and hypopyon. Age ≥60 years (PBK vs. MBK, 31% vs. 51%, p = 0.048), symptom duration (4.7 days vs. 8.0 days, p = 0.009), and contact lens use (34% vs. 18%, p = 0.036) were significantly different between the two groups. Regarding treatment outcomes, epithelial healing time ≥10 days, the final best-corrected visual acuity (BCVA), a need for surgical intervention, and the rate of poor clinical outcome were not significantly different between the two groups. Significant risk factors for a poor clinical outcome in all patients were an initial BCVA <0.1 (Z = 6.33, two-proportion Z-test), an epithelial defect size ≥5 mm2 (Z = 4.56), and previous ocular surface disease (Z = 4.36).Conclusions: Polymicrobial bacterial keratitis, compared to monomicrobial bacterial keratitis, was more significantly associated with younger age, contact lens use, and shorter symptom duration.