An Engineering and Ophthalmology Collaboration

This research in computer-based retinal diagnostics is a result a collaboration between the Department of Ophthalmology at the University of Tennessee Health Science Center (UTHSC) in Memphis, Tennessee, and the Image Science and Machine Vision Group at the Oak Ridge National Laboratory. This team brings together a unique and complimentary group of scientists to address the research necessary to develop effective computer-based diagnostic capabilities for early detection of diabetic retinopathy. Dr. Chaum and his team have extensive experience in diagnosing and treating diabetic retinopathy and they are directing the construction of a comprehensive database of diagnosed retinal images. Dr. Tobin and his team have proven experience in electronic imaging encompassing scene analysis and pattern recognition for applied computer vision applications. He and his team are providing the computer science and architecture skills and experience required to effectively leverage the repository of human patient fundus imagery data to develop a telemedicine-ready, computer-aided diagnostic capability.

	Department of Ophthalmology, Hamilton Eye Institute, Memphis, Tennessee

Drs. Tobin and Chaum originally established a strong collaboration through a Laboratory Directed Research and Development project funded by the U.S. Department of Energy and ORNL, entitled “An Image-Based Method for Screening and Diagnosis of Blinding Eye Diseases” (through July 2005). The goal of this project was to develop, test, and validate the fundamentals of image-based retinal screening and diagnostic methods that leverage content-based image retrieval (CBIR) with fundus photography, fluorescein angiography, and optical coherence tomography images.

Dr. Karen Fox (UTHSC) is serving this effort by facilitating and directing the implementation and networking requirements. Dr Fox is Assistant Dean at the UTHSC College of Medicine and Director of the Center for Health Innovation and Community Outreach. She is the principal investigator on telemedicine grants from HRSA, the Departments of Commerce and Agriculture and the Internet 2 Consortium for medical middleware and protocol development. The UTHSC Telehealth Network currently has 67 telehealth locations in Tennessee, Mississippi and Arkansas, and mobile medical vans providing 1,250 clinical encounters annually. Through our collaboration with the Telehealth Network we will be able to utilize an established telemedical network maintained by the UTHSC. This dedicated IP network is built upon a secure infrastructure of dedicated T-1 lines, with transmission speeds of 1.544Mbps. The UTHSC employs dedicated personnel to operate and maintain this network including bridge operations, assuring clinical access, maintenance and network management. Data security is maintained along the fixed network by the physical nature of the dedicated T1s, and along outside connectivity by the extensive firewall system already in place in the UTHSC server network. Additional security is provided by Polycom's codecs, which offer embedded encryption and by physical measures to facilitate privacy and compliance with HIPAA regulations. All members of the UT Telehealth Department have been certified in a UTHSC HIPAA training program, and the department's operations have been reviewed for compliance with these measures. The Network employs biostatisticians who will participate in the statistical review and analysis of the research data generated by this project.

	Oak Ridge National Laboratory, Oak Ridge, Tennessee

This collaboration has resulted in an NIH National Eye Institute funded program titled "Automated Screening for Diabetic Retinopathy by Content" (R01 EY017065). The goal of this program is to investigate the feasibility of a CBIR method to accurately describe and index human retinal images of diabetic retinopathy collected from low-cost, non-dilated retinal photographic examinations. Our goal is to demonstrate the feature-based indexing and retrieval process of CBIR and verify our hypothesis that retinal pathology can be identified and quantified from visually similar retinal images assembled from a large database comprising images of diabetic retinopathy. This research extends our previous investigations by incorporating intrinsic and extrinsic patient data to provide a diagnostic method in a telemedicine environment.