Differentiation of Active Corneal Infections from Healed Scars Using Deep Learning.

PURPOSE: To develop and evaluate an automated, portable algorithm to differentiate active corneal ulcers from healed scars using only external photographs. DESIGN: A convolutional neural network was trained and tested using photographs of corneal ulcers and scars. PARTICIPANTS: De-identified photographs of corneal ulcers were obtained from the Steroids for Corneal Ulcers Trial (SCUT), Mycotic Ulcer Treatment Trial (MUTT), and Byers Eye Institute at Stanford University. METHODS: Photographs of corneal ulcers (n = 1313) and scars (n = 1132) from the SCUT and MUTT were used to train a convolutional neural network (CNN). The CNN was tested on 2 different patient populations from eye clinics in India (n = 200) and the Byers Eye Institute at Stanford University (n = 101). Accuracy was evaluated against gold standard clinical classifications. Feature importances for the trained model were visualized using gradient-weighted class activation mapping. MAIN OUTCOME MEASURES: Accuracy of the CNN was assessed via F1 score. The area under the receiver operating characteristic (ROC) curve (AUC) was used to measure the precision-recall trade-off. RESULTS: The CNN correctly classified 115 of 123 active ulcers and 65 of 77 scars in patients with corneal ulcer from India (F1 score, 92.0% [95% confidence interval (CI), 88.2%-95.8%]; sensitivity, 93.5% [95% CI, 89.1%-97.9%]; specificity, 84.42% [95% CI, 79.42%-89.42%]; ROC: AUC, 0.9731). The CNN correctly classified 43 of 55 active ulcers and 42 of 46 scars in patients with corneal ulcers from Northern California (F1 score, 84.3% [95% CI, 77.2%-91.4%]; sensitivity, 78.2% [95% CI, 67.3%-89.1%]; specificity, 91.3% [95% CI, 85.8%-96.8%]; ROC: AUC, 0.9474). The CNN visualizations correlated with clinically relevant features such as corneal infiltrate, hypopyon, and conjunctival injection. CONCLUSIONS: The CNN classified corneal ulcers and scars with high accuracy and generalized to patient populations outside of its training data. The CNN focused on clinically relevant features when it made a diagnosis. The CNN demonstrated potential as an inexpensive diagnostic approach that may aid triage in communities with limited access to eye care.

Development and Validation of a Natural Language Processing Algorithm to Extract Descriptors of Microbial Keratitis From the Electronic Health Record.

PURPOSE: The purpose of this article was to develop and validate a natural language processing (NLP) algorithm to extract qualitative descriptors of microbial keratitis (MK) from electronic health records. METHODS: In this retrospective cohort study, patients with MK diagnoses from 2 academic centers were identified using electronic health records. An NLP algorithm was created to extract MK centrality, depth, and thinning. A random sample of patient with MK encounters were used to train the algorithm (400 encounters of 100 patients) and compared with expert chart review. The algorithm was evaluated in internal (n = 100) and external validation data sets (n = 59) in comparison with masked chart review. Outcomes were sensitivity and specificity of the NLP algorithm to extract qualitative MK features as compared with masked chart review performed by an ophthalmologist. RESULTS: Across data sets, gold-standard chart review found centrality was documented in 64.0% to 79.3% of charts, depth in 15.0% to 20.3%, and thinning in 25.4% to 31.3%. Compared with chart review, the NLP algorithm had a sensitivity of 80.3%, 50.0%, and 66.7% for identifying central MK, 85.4%, 66.7%, and 100% for deep MK, and 100.0%, 95.2%, and 100% for thin MK, in the training, internal, and external validation samples, respectively. Specificity was 41.1%, 38.6%, and 46.2% for centrality, 100%, 83.3%, and 71.4% for depth, and 93.3%, 100%, and was not applicable (n = 0) to the external data for thinning, in the samples, respectively. CONCLUSIONS: MK features are not documented consistently showing a lack of standardization in recording MK examination elements. NLP shows promise but will be limited if the available clinical data are missing from the chart.

Classification Criteria for Syphilitic Uveitis.

PURPOSE: To determine classification criteria for syphilitic uveitis. DESIGN: Machine learning of cases with syphilitic uveitis and 24 other uveitides. METHODS: Cases of anterior, intermediate, posterior, and panuveitides were collected in an informatics-designed preliminary database, and a final database was constructed of cases achieving supermajority agreement on the diagnosis, using formal consensus techniques. Cases were analyzed by anatomic class, and each class was split into a training set and a validation set. Machine learning using multinomial logistic regression was used on the training set to determine a parsimonious set of criteria that minimized the misclassification rate among the different uveitic classes. The resulting criteria were evaluated on the validation set. RESULTS: Two hundred twenty-two cases of syphilitic uveitis were evaluated by machine learning, with cases evaluated against other uveitides in the relevant uveitic class. Key criteria for syphilitic uveitis included a compatible uveitic presentation (anterior uveitis; intermediate uveitis; or posterior or panuveitis with retinal, retinal pigment epithelial, or retinal vascular inflammation) and evidence of syphilis infection with a positive treponemal test. The Centers for Disease Control and Prevention reverse screening algorithm for syphilis testing is recommended. The misclassification rates for syphilitic uveitis in the training sets were as follows: anterior uveitides 0%, intermediate uveitides 6.0%, posterior uveitides 0%, panuveitides 0%, and infectious posterior/panuveitides 8.6%. The overall accuracy of the diagnosis of syphilitic uveitis in the validation set was 100% (99% confidence interval 99.5, 100)-that is, the validation set's misclassification rates were 0% for each uveitic class. CONCLUSIONS: The criteria for syphilitic uveitis had a low misclassification rate and seemed to perform sufficiently well for use in clinical and translational research.

Association of single nucleotide polymorphisms of interleukins-1ß, -6, and -12B with contact lens keratitis susceptibility and severity.

PURPOSE: To investigate whether single nucleotide polymorphisms (SNPs) in interleukin (IL)-1ß, IL-6, and IL-12ß are associated with the susceptibility and severity of contact lens-related keratitis. DESIGN: Retrospective, case control study. PARTICIPANTS: One hundred twelve cases of keratitis and 225 controls were recruited from studies conducted at Moorfields Eye Hospital and in Australia during 2003 through 2005. METHODS: Buccal swab samples were collected on Whatman FTA cards and were mailed by post for analysis. IL-1ß (-31), IL-6 (-174, -572, -597), and IL-12B (3'+1158) genotypes were analyzed with pyrosequencing and analyzed using a regression model for susceptibility (sterile, microbial keratitis, controls) and severity. Statistical significance was set at 0.05. MAIN OUTCOME MEASURES: The relative risk of developing contact lens-related keratitis and more severe forms of the disease based on allele, genotype, and haplotype associations. RESULTS: Carriers of IL-6 SNPs were more likely to experience moderate and severe events compared with those with nonmutated genotypes (-174 heterozygous: odds ratio [OR], 3.1; 95% confidence interval [CI], 1.1-8.3; homozygous: OR, 6.4; 95% CI, 1.4-28.4; -174/-597: OR, 4.1; 95% CI, 1.6-11.0). More severe keratitis and microbial keratitis were less likely to occur in wearers with the nonmutated IL-6 haplotype (severity OR, 0.4 [95% CI, 0.2-0.7]; microbial OR, 0.6 [95% CI, 0.4-0.9]). Wearers carrying an IL-12B SNP had an increased risk of sterile keratitis (OR, 9.7; 95% CI, 1.2-76.9) compared with controls. CONCLUSIONS: The IL-6 SNPs are known to reduce protein expression of this cytokine and thus ocular immune defense, and carriers of these SNPs were more likely to experience more severe and microbial keratitis, suggesting that IL-6 decreases the severity and susceptibility of contact lens-related keratitis. Carriers of a functional SNP of IL-12B that is known to increase IL-12 expression and stability are more likely to experience sterile keratitis, suggesting that this is associated with the intense inflammatory reaction that occurs in this condition.

Acute periorbital infections: who needs emergent imaging?

OBJECTIVES: Computed tomography (CT) is used often in the evaluation of orbital infections to identify children who are most likely to benefit from surgical intervention. Our objective was to identify predictors for intraorbital or intracranial abscess among children who present with signs or symptoms of periorbital infection. These predictors could be used to better target patients for emergent CT. METHODS: This was a retrospective cohort study of all patients admitted to an urban pediatric tertiary care emergency department between 1995 and 2008. We included otherwise healthy patients with suspected acute clinical periorbital or orbital cellulitis without a history of craniofacial surgery, trauma, or external source of infection. Immunocompromised patients and patients with noninfectious causes of periorbital swelling were excluded. Variables analyzed included age, duration of symptoms, highest recorded temperature, previous antibiotic therapy, physical examination findings, laboratory results, and interpretation of imaging. CT scans of the orbit were reread by a neuroradiologist. RESULTS: Nine hundred eighteen patients were included; 298 underwent a CT scan, and of those, 111 were shown to have an abscess. Although proptosis, pain with external ocular movement, and ophthalmoplegia were associated with presence of an abscess, 56 (50.5%) patients with abscess did not experience these symptoms. Other variables associated with the presence of an abscess in multivariate analysis were a peripheral blood neutrophil count greater than 10,000/microL, absence of infectious conjunctivitis, periorbital edema, age greater than 3 years, and previous antibiotic therapy (P < .05 for all). Our recursive partitioning model identified all high-risk (44%) patients as well as a low-risk (0.4%-2%) group (Rsq = 0.27). CONCLUSIONS: We confirmed that patients with proptosis and/or pain or limitation of extraocular movements are at high risk for intraorbital abscess, yet many do not have these predictors. Other features can identify patients who do not have such obvious predictors but do have significant risk of disease. A recursive partitioning model is presented.

Use of azithromycin ophthalmic solution in the treatment of chronic mixed anterior blepharitis.

We tested the efficacy of azithromycin ophthalmic solution for the treatment of chronic mixed anterior blepharitis. The findings suggest that patients with chronic mixed anterior blepharitis can be more effectively treated with azithromycin ophthalmic solution than erythromycin ophthalmic ointment. Patients treated with azithromycin ophthalmic solution show an extraordinary clinical response with shorter treatment duration.

Classifying the severity of corneal ulcers by using the "1, 2, 3" rule.

PURPOSE: To propose and define the "1, 2, 3" rule as a system of categorizing the severity of bacterial keratitis and to determine its value in predicting the likelihood of visual loss after healing of bacterial corneal infection. METHODS: Seventy patients were evaluated for infectious corneal ulcers by 2 cornea subspecialists (J.D.L., M.C.V.) at a tertiary facility between October 30, 2001, and November 1, 2004. Forty-one patients met the inclusion criteria. Visual acuities, ulcer characteristics, culture results, and treatment data were collected. RESULTS: The sensitivity of the "1, 2, 3" rule for detecting potentially sight-threatening (PST) ulcers that would result in 0.20 logMAR (2 Snellen lines) or greater loss in best spectacle-corrected visual acuity (BSCVA) was 100%. The specificity of the "1, 2, 3" rule in correctly labeling an ulcer that did not cause vision loss as rarely sight threatening (RST) was 57.14%. The positive predictive value of the "1, 2, 3" rule was 28.57%. The negative predictive value of the "1, 2, 3" rule was 100%. CONCLUSIONS: Corneal ulcers classified as PST are at significantly higher risk for sustaining loss of BCSVA compared with RST corneal ulcers. The "1, 2, 3" rule is sensitive and specific enough to be clinically useful in predicting which ulcers are more likely to have vision loss and which ulcers are not.

Neural network approach to classify infective keratitis.

PURPOSE: Infective keratitis is a major sight-threatening condition in developing countries like India. An early diagnosis of infective keratitis is critical to its treatment. Epidemiological trends, morphological features of corneal ulceration and presence of other risk factors often dictate choice of initial treatment. This work assesses the usefulness of classification of infective keratitis by artificial neural network (ANN). METHODS: Forty input variables from each of the sixty-three known bacterial or fungal ulcers provided the basis for training a three layer feed-forward neural network. The trained neural network classified another set of forty-three corneal ulcers. RESULTS: Trained artificial neural network could classify correctly all sixty-three cornea ulcers in the training set. In the test set, the artificial neural network correctly classified 39 out of 43 cornea ulcers. Specificity for bacterial and fungal categories was 76.47% and 100% respectively. Accuracy of classification by neural network was 90.7% and compared significantly better than clinicians' prediction of 62.8% (p < 0.01). CONCLUSION: ANN has the potential to help clinicians classify corneal ulcers more accurately.

Avaliaçäo clínica do tratamento tópico das úlceras de córnea bacterianas com ciprofloxacina a 0,3 por cento / Clinical evaluation of ciprofloxacin 0,3 per cent ophthalmic solution for treatment of bacterial keratitis

Foi realizado um estudo clínico, prospectivo, envolvendo 30 pacientes com úlcera de córnea bacteriana cultura positiva, tratados aleatoriamente com colírio de ciprofloxacina a 0,3 por cento ou com colírios fortificados de cefalotina (50 mg/ml) e gentamicina (14 mg/ml). O tratamento com ciprofloxacina alcançou 86,7 por cento de cura enquanto os pacientes tratados com a associaçäo dos antibióticos fortificados alcançou 80 por cento de cura. Näo ocorreram efeitos adversos graves com nenhum dos regimes terapêuticos utilizados. Estes achados sugerem que, nas condiçöes desse estudo, a ciprofloxacina a 0,3 por cento é efetiva como agente único para o tratamento de úlcera de córnea bacteriana

Comparison of the old and new W.H.O. leprosy disability grading scheme for ocular disabilities.

We compared the old (1970) and new (1988) World Health Organization schemes for classifying the ocular disabilities in leprosy patients. 509 leprosy patients from eight resettlement villages in central South Korea were examined and graded by eye according to both of the schemes. A more liberal definition of severely disabled in 1988 resulted in a 119% increase in eyes graded as severely disabled in this population. 59 eyes were graded as severely disabled by the old scheme and 129 eyes were so graded according to the new scheme. Keratitis, one of three measures of moderate disability in the old scheme, was replaced by corneal anaesthesia in the new scheme, but this change did not make a substantial difference in the number of patients in the moderately impaired category. In the absence of longitudinal studies documenting the significance of keratitis, it is unclear whether the change in an improvement. The new disability scheme improved upon the old by removing the criteria for mild impairment.