Purpose

To validate computer software developed to assess digital corneal photographs of fungal keratitis in clinical research.

Methods

A cornea specialist and five medical students (after training) graded on two occasions 100 corneal photographs of patients with fungal keratitis using Optscore software. Variables assessed were lesion area, location, degree of opacity, percentage of the ulcer lying within a central 4 mm circle of the cornea. Intraclass correlation coefficients (ICCs) were used to assess intragrader reliability, agreement of the students with the corneal specialist, and the reliability of the group mean of the student raters. The area determined using Optscore was compared to the area estimated from slit lamp and to visual acuity.

Results

As a group, medical students achieved an ICC greater than 0.9 for five out of the seven assessed variables. Similar levels of consistency were found after analyzing the graders' individual results compared to the specialist. The area estimated using slit lamp examination was highly correlated with the mean area determined by Optscore, as was the logarithm of the minimum angle of resolution visual acuity at enrollment.

Conclusions

Non-expert graders using Optscore to assess digital photographs of fungal keratitis are self-consistent, agree with an expert grader both as a group and individually, and measurements of ulcer area obtained from Optscore are highly correlated with measurements of the same patients obtained on clinical examination. These observations support the validity of Optscore for assessing corneal pathology associated with fungal keratitis and make it a promising clinical research tool.

Objective

To assess the trends in microbiological organisms identified from corneal scrapings from patients with infectious keratitis at a tertiary care medical centre in South India.

Methods

We reviewed the records of the microbiology laboratory at Aravind Eye Hospital in Madurai, India, from 2002 until 2012. We identified the microbiological causes of all corneal ulcers from the culture and smear results, and assessed for trends in bacterial and fungal keratitis over time.

Results

Of 23 897 corneal patients with ulcer with a corneal smear from 2002 to 2012 a fungal organism was identified in 34.3%, a bacterial organism in 24.7% and no organism in 38.3%. During this period, the annual number of keratitis cases due to bacteria decreased from 677 to 412, and the annual number due to fungus increased from 609 to 863. In analyses accounting for the total number of outpatients seen each year, the decline in number of smears positive for bacteria was statistically significant (p<0.001) but the increase in the number positive for fungus was not (p=0.73). The relative frequency of individual bacterial or fungal organisms remained relatively stable over this time.

Conclusions

At a tertiary eye care centre in South India, there has been a reduction in the numbers of smear-positive bacterial keratitis over the past 11 years. This decline likely reflects economic development in India and increased access to antibiotics.

Aims

To report trends in antibiotic resistance in cases of bacterial keratitis from a large eye hospital in South India.

Methods

In this retrospective cross-sectional study, the microbiology laboratory records of patients with infectious keratitis diagnosed at an eye hospital in South India from 2002 to 2013 were reviewed to determine the proportion with antibiotic non-susceptibility.

Results

3685 bacterial isolates had susceptibility testing performed over the 12-year period. The two most common organisms with resistance were Streptococcus pneumoniae (n=1204) and Pseudomonas aeruginosa (n=894). Antibiotic non-susceptibility was generally uncommon for these two organisms and no significant trends were detected over the course of the study. In contrast, Staphylococcus aureus (N=211) isolates demonstrated a significant increase in fluoroquinolone non-susceptibility over the 12-year study period. This coincided with a significant increase in methicillin-resistant S. aureus (MRSA) during the study period, though the increase in fluoroquinolone resistance was likewise seen in methicillin-sensitive S. aureus (MSSA). For example, ofloxacin resistance in MSSA increased from 11.1% in 2002 to 66.7% in 2013 (p=0.002). No trends were apparent for the aminoglycosides, cefazolin or vancomycin, for which in vitro non-susceptibility generally appeared to be low.

Conclusion

Resistance to antibiotics was generally stable for infectious keratitis isolates from a large eye hospital in South India, except for S. aureus, which experienced a significant increase in fluoroquinolone resistance from 2002 to 2013. Fluoroquinolone antibiotics currently have poor in vitro activity against both MRSA and MSSA in South India and are therefore not the ideal therapy for Staphylococcal corneal ulcers.

Purpose

We described the change in visual acuity experienced by eyes successfully treated for bacterial keratitis.

Methods

This was a prospective cohort study of a subset of study participants who had previously enrolled in the Steroids for Corneal Ulcers Trial (SCUT). All study participants had been diagnosed with culture-proven bacterial keratitis before enrollment in SCUT and subsequently were randomized to adjunctive topical corticosteroids or placebo. During SCUT, we monitored study participants at enrollment, 3 weeks, 3 months, and 12 months. We invited a subset to complete a comprehensive eye examination approximately 4 years after enrollment in SCUT. Certified refractionists assessed best spectacle-corrected visual acuity (BSCVA) using the same protocol at each study visit.

Results

We examined 50 SCUT participants at 4 years after enrollment. Among those in this cohort, mean logMAR BSCVA at enrollment was 0.85 (Snellen equivalent, 20/160; 95% confidence interval [CI], 0.71-0.99). On average, visual acuity improved by 2.9 logMAR lines from enrollment to 3 weeks (P < 0.001), 1.2 lines from 3 weeks to 3 months (P = 0.002), and 0.8 lines from 3 to 12 months (P = 0.01). The BSCVA did not change significantly between 12 months and 4 years (0.04-line improvement, P = 0.88). After controlling for visual acuity at enrollment, BSCVA was not significantly different between the corticosteroid and placebo groups at 4 years (P = 0.53).

Conclusions

Cases of bacterial keratitis may continue to demonstrate improvements in visual acuity up to 12 months following diagnosis, but further improvements are unlikely. These findings may guide the appropriate timing of surgical intervention in these patients. (ClinicalTrials.gov number, NCT00324168.).

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.

Purpose

To compare longitudinal outcomes of visual acuity after fungal corneal ulcers with those of bacterial ulcers.

Design

Prospective cohort study.

Methods

This study was conducted in a tertiary eye hospital in South India. The population consisted of 100 of 152 individuals whose fungal or bacterial keratitis had been diagnosed 4 years prior and had been enrolled in 1 of 2 concurrent randomized trials. Causative organisms of infectious keratitis were either bacterial or fungal. Presenting visual acuity consisted of best spectacle corrected visual acuity (BSCVA) and hard contact lens-corrected visual acuity (CLVA).

Results

Fifty study participants with prior fungal keratitis and 50 with prior bacterial keratitis were enrolled. Four years after treatment for keratitis, participants' presenting vision in the better eye was worse than 20/60 for 12 individuals (24.0%) in the fungal group and 10 individuals (20.0%) in the bacterial group. Median BSCVA in the affected eye at the 4-year visit in the fungal group was similar to that in the bacterial group (Snellen equivalent, 20/32 for each), although vision worse than 20/400 was more common in the fungal ulcer group after spectacle correction (odds ratio [OR] 4.19; 95% confidence interval [CI], 1.11-15.8) and contact lens correction (OR, 5.74; 95% CI, 1.37-24.1).

Conclusions

In this South Indian population with a previous episode of fungal or bacterial keratitis, correctable bilateral visual impairment was common. Although long-term visual outcomes were, on average, similar between fungal and bacterial ulcers, fungal ulcers were more likely to produce severe visual impairment.