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.

Ophthalmic manifestations of paranasal sinus disease: a clinical grading system.

BACKGROUND: Ophthalmic complications are common manifestations of paranasal sinus disease. We propose a clinical grading system to standardize the category, treatment, and outcome of these complications. METHODS: Forty-two patients with paranasal sinus disease-induced ophthalmic complications were included in the study. The patients were categorized according to their clinical presentation: Grade I, patients with anatomical disturbance; Grade II, patients with ophthalmic functional involvement; Grade III, patients with orbital infection; and Grade IV, patients with visual impairment. RESULTS: The age of the patients ranged from 8 to 65 years (mean = 30 years). Overall, 36% patients had ophthalmologic manifestations with anatomical disturbance (Grade I), 26% had functional disturbance (Grade II), 26% had orbital infections (Grade III), and 12% had visual loss (Grade IV). The cause of the ophthalmic manifestations was allergic fungal sinusitis in 50%, chronic rhinosinusitis in 36%, acute sinusitis in 10%, and mucocele in 4%. The most common ophthalmic manifestation was proptosis (36%), followed by orbital infection (26%), functional involvement (26%), and visual impairment (12%). The patients underwent functional endoscopic sinus surgery and medical treatment when indicated. The outcome in Grades I, II, and III was favorable; the outcome in Grade IV was not favorable. CONCLUSION: This simple clinical grading system provides a useful tool for assessing the overall status of ophthalmic complications of Acute and chronic paranasal sinus disease at the initial clinical assessment and for subsequent management of these complications.

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.

Prognosis of endogenous fungal endophthalmitis and utility of Ishibashi's classification.

PURPOSE: Evaluations of visual acuity outcomes of eyes with endogenous fungal endophthalmitis were made retrospectively, according to the classification proposed by Ishibashi. METHODS: We surveyed endogenous fungal endophthalmitis cases at the 4 Nihon University Hospitals and 20 affiliated hospitals. Sixty eyes of 34 patients were classified into five stages according to Ishibashi's system, and therapeutic methods and visual outcomes in each stage were then evaluated. RESULTS: Systemic antifungal drugs were efficacious in 82% of stage II and 69% of stage IIIa cases. Antifungal drugs were even efficacious in 42% of stage IIIb cases. Among the unresponsive cases, only half had been given the maximal dosage of antifungal drugs. Half of the eyes in which vitrectomy had been performed at stage IIIb achieved a postoperative visual acuity of 0.5 or better and none had a visual acuity of less than 0.03. CONCLUSION: Based on the above results, we concluded that systemic antifungal drugs should be administered at the maximal dosage to stage II and IIIa cases. If these eyes progress to stage IIIb despite receiving the maximal dosage, vitrectomy is indicated. For stage IIIb eyes, the maximal dosage should be administered first. If not efficacious, vitrectomy should be carried out before progression to stage IV.

[Prognosis of endogenous fungal endophthalmitis and utility of Ishibashi's classification].

PURPOSE: Prognostic evaluations of eyes with endogenous fungal endophthalmitis were made according to the classification proposed by Ishibashi. PATIENTS AND METHODS: We surveyed endogenous fungal endophthalmitis cases at four Nihon University Hospitals and 20 branch hospitals. Sixty eyes of 34 patients were classified into five stages according to Ishibashi's proposal, and therapeutic methods and visual prognosis in each stage were then evaluated. RESULTS: Systemic antifungal drugs were efficacious in 82% of stage II and 69% of stage IIIa cases. Even among stage III b cases, antifungal drugs were efficacious in 42%. Among the unresponsive cases, only half had been given the maximum dosage of antifungal drugs. Half of the eyes in which vitrectomy was performed at stage III b achieved a postoperative visual acuity of 0.5 or better and none had a visual acuity of less than 0.03. CONCLUSION: Based on the above results, we concluded that systemic antifungal drugs should be administered at the maximum dosage in stage II and III a cases. If these eyes progress to stage III b despite the maximum dosage, vitrectomy is indicated. For stage III b eyes, the maximum dosage should be administered first. If the maximum dosage is not efficacious, vitrectomy should be carried out before progression to stage IV.