Creeping crawlers win award

October 18, 2006, volume 37, no. 4
By Barry Shell



Document Tools

Print This Article

E-mail This Page

Font Size
S      M      L      XL

Related Stories

Imagine an intelligent microscopic worm that could painlessly crawl through blood vessels, mapping every branch and passage. Such a creature could guide brain surgeons cutting out cancerous growths, or cardiologists searching for narrowed arteries.

While it doesn't exist in the real world, computer scientists at SFU have created virtual animals that do travel the complex tree-like vessels in three-dimensional medical images from computer-aided tomography (CAT) scans, or magnetic resonance images (MRI).

Since the images are merely vast amounts of data representing different intensities of light passing through the body, these virtual creatures can decipher the information to map out complicated networks of nerves or air passages in lungs and similar systems in other organs.

"The virtual worms have artificial bodies, movements, sensors, memory, behaviours and decision-making abilities," says Ghassan Hamarneh, assistant professor of computing science at SFU.

For his PhD thesis, the Jordanian-born Hamarneh created virtual animals that could move around two dimensional medical images such as X-rays.

"Our virtual worms would detect and analyse anatomical structures in the images," says Hamarneh, "but they were restricted to two-dimensional worlds."

Now Chris McIntosh, one of his graduate students, is extending the work to 3D images.

The 3D crawlers don't just map out the passages they traverse. They also provide immediate quantitative and qualitative analysis of the vessels.

The information is useful to doctors and researchers looking for cures for illnesses like heart disease or nerve damage.

"A major advantage of our vessel crawlers is their ability to encode prior knowledge," says McIntosh.

The creatures remember features and structures they have encountered and use this information to figure out where the vessel goes next or where it branches. It's a powerful technique that can extract fine detail from image data.

McIntosh's work earned him a Canadian Institute of Health Research graduate scholarship master's award, the only one that went to a computer scientist in 2006.

Visit the group's medical image analysis lab website at http://mial.fas.sfu.ca/ and click on Artificial Life.

Search SFU News Online