Abstract
Abstract: :
Purpose: This study examined a remote site ophthalmologist’s confidence in grading 560x720 pixel, 24-bit color composite digital imagery. It also assessed diabetic inmate vision levels affecting the usefulness of store-and-forward telemedicine evaluation of diabetic retinopathy. Methods: Two hundred consecutive diabetic inmates retinas were imaged with a nonmydriatic retinal camera. Each eye’s nine 45-degree fields were merged into a single, large composite image using computer tools and graded as: no retinopathy, mild nonproliferative, moderate nonproliferative, severe nonproliferative or proliferative. Grader’s confidence level was rated as no, low or high confidence. The number of useful retina fields of each eye was recorded. Results: The ophthalmologist reader recorded high confidence level in 88.79% of eyes graded as no retinopathy, 77.42% of eyes graded as mild non-proliferative and 64.86% of eyes graded as proliferative. Severe non-proliferative and moderate non-proliferative eyes had the lowest percentage of eyes graded with a high level of confidence: respectively 42.86% and 17.65%. No field was missing in 90.71% of eyes with vision ≷=20/50, 71.47% of 20/60 - 20/200 and 50% of <20/200. Conclusion: Lowest confidence was associated with grading moderate or severe non-proliferative retinopathy using 560x720 pixel, 24-bit color, non-stereo imagery when inspecting for signs of retinopathy with 4-2-1 rules applied in modified quadrants. Lower vision eyes had less number of useful retinal fields for telemedicine evaluation or follow-up of retinopathy.
Keywords: 388 diabetic retinopathy • 430 imaging/image analysis: clinical • 356 clinical (human) or epidemiologic studies: systems/equipment/techniques