Computer Aided Detection of Diabetic Retinopathy in Veterans with Diabetes

Iowa City Va Medical Center, Iowa City IA

Investigators

Linked publications & trials

Abstract

DESCRIPTION (provided by applicant): This is a resubmission of a well-developed research proposal to compare human assessment of diabetic retinopathy to that of computer assisted assessment that generated a lot of interest and discussion. The applicant provided a good response to the previous critiques. This proposal has the potential to result in immediate benefits to efficiency of the health care system and reduce overall costs, while having no undesirable effect on patients themselves. There remain concerns about if or how the algorithm will be improved as a result of this study, and questions about intellectual property. PUBLIC HEALTH RELEVANCE: Computer Aided Detection of Diabetic Retinopathy (DR) in Veterans with Diabetes DR is a feared complication of diabetes and an important cause of blindness in veterans. The VA, through the Office of Care Coordination, has been rapidly moving to photoscreening for all veterans with diabetes, using digital cameras and licensed independent practitioners as readers. Expert reading suffers from intra-and inter-observer variability, and expert readers are scarce, costly and have high turnover rate. Computer detection of DR as developed by the applicant has the potential to increase the cost-effectiveness, scalability, reproducibility and sustainability of photoscreening programs. However, skeptics remain within the scientific community and computer detection of DR is unproven in clinical practice. The applicant will compare the performance of human experts, human experts aided by computer, and standalone computer on sensitivity, specificity, time to diagnosis, and also perform a cost-effectiveness analysis of DR detection on a large number of veterans.