Abstract
Purpose :
Malarial retinopathy (MR) is a highly specific indicator for cerebral malaria (CM). Retinal cameras allow documentation of MR for remote reading as an alternative to an onsite examination by ophthalmic experts. However, current retinal cameras face issues with affordability and lack of user expertise in the low-resource settings of Africa where CM is prevalent. Consequently, we investigated the performance of two retinal cameras and a smartphone adapter for detecting MR and in terms of usability in African clinical settings.
Methods :
We analyzed retinal images from 23 pediatric patients with clinically diagnosed CM. An ophthalmologist used two retinal cameras (Pictor-plus, Topcon-50EX) and a smartphone adapter (Prototype Peek Retina) to image the patients. The images were graded for the presence/absence of MR and specific MR lesions (whitening, hemorrhages, and vessel discoloration). The ground truth was determined by an ophthalmologist based on binocular indirect ophthalmoscopy (BIO) findings. The image grading was compared with BIO in terms of accuracy of MR detection and Kappa agreement on severity grades assigned to each MR lesion.
Results :
Compared to BIO, the sensitivity and specificity for MR detection was 87%/75% for the Peek adapter, 100%/87% for Pictor-plus, and 100%/75% for Topcon-50EX, respectively. The sensitivity and specificity of Pictor-plus for detection of individual MR lesions was consistently comparable or superior to the other two systems. In terms of severity grading of MR lesions, Pictor-plus gave an average Kappa agreement with BIO of 0.68 compared to 0.51 by Peek adapter and 0.61 by Topcon-50EX. Vessel discoloration could not be detected by any of the systems.
Conclusions :
Pictor-plus offered the best overall performance compared to other systems in terms of usability, image quality, and MR detection accuracy. It was also found by the ophthalmologist to be stable, user-friendly, and ergonomically suitable for imaging comatose children. The Topcon camera, although providing high accuracy, is bulky and difficult to handle. The Peek adapter is inexpensive, small and lightweight, but its structure makes it unstable for imaging that results in lower image quality. The ergonomics of each camera and ease of training health care workers for quality imaging with each camera may be a factor in the choices of retinal cameras.
This is an abstract that was submitted for the 2016 ARVO Annual Meeting, held in Seattle, Wash., May 1-5, 2016.