Abstract
Purpose :
While posterior eye diseases cause the majority of visual impairment in the US, anterior segment (AS) eye diseases cause the majority of visits to eye-care professionals. We evaluated the diagnostic accuracy of detecting corneal ulcers, opacities, and abrasions using external photographs from two portable AS cameras for telemedicine purposes.
Methods :
A prospective study of patients >= 18 years old with a clinical diagnosis in one eye of corneal pathology. When possible, photographs of both eyes were obtained. Gold-standard diagnosis was determined from a slit-lamp exam by a cornea specialist. iTouch 5S and Nidek VersaCamm cameras were used to obtain a series of images including multiple gazes with white light and fluorescein with blue light. Three cornea specialists interpreted the images for presence, suspicion, or absence of AS pathology. Sensitivity and specificity to detect AS disease compared to gold standard diagnosis were calculated, stratified by camera and grader. Reliability of diagnosis was evaluated with weighted kappa statistics. Grader confidence in diagnosis was assessed on a scale from 1 (not at all) to 10 (most confident).
Results :
188 eyes (108 patients) were photographed. By gold standard diagnosis, 59 eyes (31%) had corneal scars, 34 (18%) had ulcers, 13 (7%) had abrasions, and 82 (44%) were normal. Sensitivity to detect AS pathology ranged from 51-69% for iTouch and 62-73% for Nidek, across graders; specificity ranged from 82-96% for iTouch and 92-98% for Nidek. The positive and negative predictive values of the cameras were 83-95% and 60-67% for iTouch, and 92-97% and 67-72% for Nidek, respectively. Inter-grader reliability was moderate to strong with weighted kappa values ranging from 0.51-0.70 for iTouch and 0.71–0.79 for Nidek. On average, graders were confident in their diagnosis (mean, standard deviation) that ranged from 6.8 (1.7) to 7.7 (1.9), over all graders and cameras. Grader confidence was the highest when pathology was present.
Conclusions :
External photographs taken by standard, non-enhanced portable cameras and interpreted remotely by ophthalmologist graders yielded sensitivity values that are likely not yet suitable for telemedicine applications, even in the absence of visible pathology on photographs. Additional work is needed to improve the ability to detect AS pathology remotely.
This is an abstract that was submitted for the 2017 ARVO Annual Meeting, held in Baltimore, MD, May 7-11, 2017.