Making it harder for cancer to hide

Alexander Wong could have spun what he learned as a University of Waterloo student into a career in consumer electronics.

Instead, he enlisted in the war on cancer.

An assistant professor in systems design engineering, Wong co-directs the Vision and Image Processing Lab with colleagues David Clausi and Paul Fieguth.
 Remote sensing, via satellite, is one side of his work. The other involves changing the way magnetic resonance imaging (MRI) works to reveal hard-to-find cancerous tissue.

“If we can pinpoint where the cancer is, it facilitates minimally invasive treatment,” Wong says. “We can destroy the cancerous region and leave the organ intact, with fewer side effects.”

Wong’s research currently focuses on prostate and skin cancers, conditions of particular concern to senior Canadians. Besides modulating MRI functions to “light up” cancerous tissue that is tricky to spot, his research group is developing accompanying clinical decision support software to improve how scans are read and interpreted by patient-care teams.

Wong expects clinical testing to begin next year at Sunnybrook Health Sciences Centre in Toronto.

“Our goal is to help doctors make consistent, well-informed decisions,’’ he says.

Cutting confusion

Hamid Tizhoosh, an associate professor in systems design engineering, focuses on quality assurance in cancer treatment. He and his students developed software to not only speed up the “segmentation” of medical scans, but also to improve the analysis that follows from it.

Photo credit: Dan Epstein

Segmentation sorts an image into different regions, based on such things as intensity and texture. It’s a slow, labour-intensive process that, for example, helps distinguish diseased tissue from healthy tissue for diagnostic and treatment-planning purposes.

Trouble is, says Tizhoosh, different doctors will come up with different segmentations of the same scan.

But if that range of approaches could be collected and catalogued, “we could build a consensus contour, and hence, establish a baseline for accuracy measurement,’’ he says.

Through cloud computing and algorithms, his research group built “knowledge maps” — based on the input of many doctors, potentially from around the world — to achieve more consistent segmentations.

“This,” Tizhoosh says, “has the potential to practically eliminate the error in medical- image analysis, as the technology is based on collective wisdom of doctors.”