Abstract
Mobile ocular telemedicine is potentially an effective method to provide service in medically underserved areas and to screen large populations for abnormalities. Currently, digital images are acquired, stored, and transferred to readers for evaluation, after which the results are provided to the subjects. The transfer of large image files and the timeliness of the subsequent reading of images are significant factors for practical implementation of effective telemedicine screening. This work examines the feasibility of in situ real-time computer analysis of digital images to determine and classify the image results as normal and abnormal. This retrospective study used a photoscreening database of 360 patients ranging in ages from 6 months to 18 years. Computer analysis automatically classified the binocular photorefraction(PR) images, and these PR results were compared to those of the subjective clinical eye examinations provided. With an average processing time of approximately 15 seconds per examinee, the analysis found that the PR results can be categorized as: a positive group that requires referral (186 cases) with a predictive value of 98.9% (2 false-positives); a negative group(144 cases) with a predictive value of 89.6% (15 false-negatives); and an uncertain group (30 cases or 8.3%) that required resolution by readers. The real-time analysis code reduces by approximately 92% the manpower for image grading and electronic transmission at this stage of ocular evaluation. These results indicate the feasibility of this approach.
Get full access to this article
View all access options for this article.