Summer/Fall 2009

STORY COMPONENTS   Using the Power of Gold Against Cancer Biomarker Identification Tools Earn Certification Microdevices Separate & Analyze Cancer Cells Breath Test Studied for Detecting Breast Cancer Creating an Ovarian Cancer Detection Tool Robotic Image-Guided Surgical Procedures

Robots are being used more frequently today in hospitals around the country. Many of these robots, like the one developed by Queen’s University associate professor Gabor Fichtinger to perform needle-based prostate biopsy and therapy procedures, require medical images to accurately guide the surgical tool to the desired target.

photo by Gary Meek Professor Allen Tannenbaum displays the computer program he developed to extract the prostate (shown in blue) from magnetic resonance images. (Download 300-dpi JPEG)

“Magnetic resonance imaging enables real-time scanning of the needle from its insertion through the skin to contact with the target, but the difficulty lies in being able to develop algorithms that immediately display and analyze the images while the patient is in the imaging scanner,” says Allen Tannenbaum, who holds a joint appointment as the Julian Hightower Chair in the Georgia Tech School of Electrical and Computer Engineering and the Wallace H. Coulter Department of Biomedical Engineering at Georgia Tech and Emory University.

To visualize the prostate in real-time during biopsy and radioactive seed-placement procedures, Tannenbaum and graduate student Yi Gao developed fast image segmentation and registration algorithms to locate the prostate in magnetic resonance images and correct for movement during the procedure. The algorithms have been integrated into the transrectal prostate magnetic resonance imaging module of Slicer3, an open-source surgical navigation software.

Tannenbaum employed two methods to “extract” the prostate from the magnetic resonance images: a shape-based algorithm and a semi-automatic method. The shape-based algorithm required inputting manually segmented three-dimensional prostate images into an artificial intelligence program. Then, given a new image, the program was able to isolate the prostate from nearby structures. For the semiautomatic method, users selected points inside and outside of the prostate and the program used that information to decide whether a pixel belonged to the organ or the background.

In addition to segmentation, images of the same patient taken at different points in time require registration to cope with deformation of the organ.

“Imagine you have a balloon – that’s the prostate – and you take a needle and push on the balloon. Pushing on it deforms the prostate and these changes have to be accounted for,” explains Tannenbaum.

The prostate presents a number of difficulties for traditional image registration approaches because there are no easily discernable landmarks. However, because the surface of the prostate is almost half convex and half concave, Tannenbaum was able to capture the concave region in each image and use it to register the whole prostate.

“Our segmentation and registration algorithms provide much greater accuracy for the robot to stick a needle in the prostate, while also requiring less than a second for computation and no special supercomputers,” adds Tannenbaum.

This research was funded by the National Institutes of Health (NIH) Roadmap Initiative called The National Alliance for Medical Imaging Computing (NA-MIC). The content is solely the responsibility of the principal investigator and does not necessarily represent the official view of the NIH.

Contact: Allen Tannenbaum (404-894-0290); E-mail: (allen.tannenbaum <at> ece.gatech.edu).