Evaluation of the Potential Impact of Telecytology of Cell Block Preparations from Pap Smears
Project Award Date: 06-01-2010
The Pap smear has been proven to be the most useful test for the detection of precancerous and cancerous lesions of the cervix. Unfortunately, pap smear preparations produce thick slides with many layers of cells. This complicates diagnosis, and because there are many focal planes, digitization of slides is impractical.
Our research studies the benefits of producing cell blocks from Pap smears, which would allow samples to be condensed and sliced thinly. Thin slices simplify diagnosis and permit accurate digitization. Digital slide images make it possible to consult with expert pathologists remotely and can be better used in education, either in published material or online. Archival of digital images is substantially easier than glass slides. Finally, digital slide images permit computer-assisted diagnosis. Automated computer vision techniques are able to prescreen digital images to flag suspicious regions and to discard areas of images that contain clearly healthy tissue. Automated prescreening alleviates the fatigue faced by professional pathologists, who must often spend hours peering through microscopes. Visual fatigue is known to reduce diagnosis accuracy. Our research investigates the use of hierarchical feature extraction techniques to learn complex visual features from slide images that can be used to better prescreen digital cell block images.
Primary Sponsor(s): Institute for Advancing Medical Innovation (IAMI) at University of Kansas Medical Center