February Fourier Talks 2010

Robert Bonner


Multispectral Noninvasive Retinal Imaging


High quality retinal digital photographs have long been critical to clinical practice and clinical trials of therapies for retinal diseases. Fundus cameras routinely acquire RGB reflectance images of a white flashlamp, reflectance images of the spectrally filtered flash, and more recently autofluorescence images that arise from fluorophores in the retinal pigment epithelium at the back of the retina. These high resolution (~10 micron) images of the retina exhibit contrast from local absorbers such as hemoglobin in the retinal arteries and veins on the retinal surface, macular pigment largely in the photoreceptor axons, and melanin and lipofuscin granules within the RPE. Rhodopsin, the photoreceptor pigment in the rods, absorbs strongly in the blue-green when "dark-adapted". Sub-RPE deposits called drusen commonly seen in older eyes, particularly in those with age-related macular degeneration, increase local intensity in retinal images by enhanced backscatter of both incident light and autofluorescence. The local variations in contrast arising from these optically active molecules routinely allow subjective and semi-quantitative clinical assessment of retinal disease. Our goal is to develop quantitative separation of the different intrinsic molecular contrast agents using computer algorithms on multispectral retinal image sets.