Interdisciplinary Initiatives Program Round 5 – 2010

Daniel Rubin, Radiology
Mark Blumenkranz, Ophthalmology

Vision is crucial to quality of life, and retinal disease is a major source of disability. While great advances have been made in retinal diagnostic imaging procedures that provide exquisite visualization of retinal architecture and disease, assessment of retinal imaging results is largely a subjective task, subject to inter-observer variation and potential oversight of features in the images that could indicate disease response or progression. A new imaging modality called optical coherence tomography (OCT) produces high resolution images of the retina, and it could dramatically improve the assessment of eye disease; however, its impact is limited by the subjective assessment of the images. Simple visual inspection of these images could overlook important early indicators of disease or change in disease status that could impact care early in the course of treatment. There is growing excitement in the use of quantitative methods of image analysis in radiology to enable “personalize medicine,” and sophisticated image processing and quantitative image analysis methods are being developed and applied in Radiology; however, to date, such methods have not been thoroughly explored in ophthalmologic imaging.

Through a collaboration between Radiology and Ophthalmology (likely the first such collaboration to be taking place), Stanford researchers are developing novel quantitative image analysis methods in retinal images to improve the characterization of retinal disease and to improve its treatment. The researchers are building on the substantial and related work that has been done on image processing in Radiology, applying similar techniques to retinal images. This is motivated by the fact that like Radiology images, retinal images contain anatomic structures of interest that may be altered by diseases and recognized by quantitative image processing.

The Stanford researchers are pursuing this work through studying a particular retinal disease called Age-related Macular Degeneration (AMD), a disease that is characterized by abnormal deposits (called “drusen” under the light-sensitive retinal layer, and progressive distortion and loss of visual acuity. While their work is focused on this single retinal disease, they believe their methods will be extensible to, and benefit many eye diseases. To date, they have created methods to automatically recognize the drusen abnormalities and measure a variety of their quantitative characteristics, such as area, volume, and height. They are now collecting a large dataset of patients with AMD to study whether the quantitative features their methods detect in images will be better predictors of disease progression or treatment response than is possible at present without such computer-assisted methods. IF successful, their work could enable ophthalmologists “profile” patients, providing them the best treatment for their particular disease based on how it is characterized by the quantitative image analysis methods the Stanford team is creating.

The collaboration has been invaluable to the progress of the research to date. Drs. Rubin and Bluenkranz are working together closely on this challenging problem; Dr. Rubin is brining image processing and informatics expertise and Dr. Blumenkranz is contributing clinical experience and clinical use cases. The teams of Dr. Rubin and Blumenkranz have regular lab meetings and they are collaborating on several joint grant submissions to expand their collaborations.