Screening for diabetic retinopathy in rural area using single-field, digital fundus images

To evaluate the practicability of using single-field, 2.3 million-pixel, digital fundus images for screening of diabetic retinopathy in rural areas.

All diabetic patients who regularly attended the diabetic clinic at Kabcheang Community Hospital, located at 15 kilometers from the Thailand-Cambodia border, were appointed to the hospital for a 3-day diabetic retinopathy screening programme. The fundi of all patients were captured in single-field, 45 degrees, 2.3 million-pixel images using nonmydriatic digital fundus camera and then sent to a reading center in Bangkok. The fundi were also examined through dilated pupils by a retinal specialist at this hospital. The grading of diabetic retinopathy from two methods was compared for an exact agreement.

The average duration of single digital fundus image capture was 2 minutes. The average file size of each image was 750 kilobytes. The average duration of single image transmission to a reading center in Bangkok via satellite was 3 minutes; via a conventional telephone line was 8 minutes. Of all 150 patients, 130 were assessed for an agreement between dilated fundus examination and digital fundus images in diagnosis of diabetic retinopathy. The exact agreement was 0.87, the weighted kappa statistics was 0.74. The sensitivity of digital fundus images in detecting diabetic retinopathy was 80%, the specificity was 96%. For diabetic macular edema the exact agreement was 0.97, the weighted kappa was 0.43, the sensitivity was 43%, and the specificity was 100%.

The image capture of the nonmydriatic digital fundus camera is suitable for screening of diabetic retinopathy and single-field digital fundus images are potentially acceptable tools for the screening. The real-time image transmission via telephone lines to remote reading center, however, may not be practical for routine diabetic retinopathy screening in rural areas.