December 11, 2020

As an undergraduate, Vector scholarship recipient Rachel Theriault brought AI and analytical chemistry together to investigate breast cancer. Her early work has gained recognition. Theriault’s undergraduate thesis – The Use of Sparse Subspace Clustering to Detect Breast Cancer from DESI-MS Scans – was recently honored for ranking among the top ten percent of entries in this year’s Global Undergraduate Awards Programme Computer Science category.

Theriault’s paper explains how clustering – a machine learning technique – can help pathologists distinguish between cancerous and benign tissues in samples taken from a lumpectomy.

“When we have computational strategies that can do intraoperative analysis or help guide a pathologist to do faster analysis by flagging certain samples, we hope that we can decrease the need for second surgeries,” says Theriault. After a lumpectomy – an operation removing only the tumour, not the breast – a pathologist determines if the cancerous tissue has been entirely extracted or if a second surgery is needed, an unfortunate requirement in over 20% of cases. The analysis can require weeks to perform, and the delay can be costly. Cancer missed in the initial surgery can continue to grow in the interim.

In her effort to improve this, Theriault researched how machine learning can analyze images produced through mass spectrometry, a technique that measures the mass of molecules within tissue samples and which has proven useful for identifying cancer biomarkers. These images present enormous amounts of granular data about the metabolite present in tissues: one pixel may contain a thousand values that require analysis, and each image is composed of nearly a hundred thousand pixels. Further inhibiting rapid analysis is the fact that cancer tissue is often heterogeneous, presenting differences that are challenging to capture and categorize.

“We needed a complex algorithm to solve this complex problem,” says Theriault. Her thesis supervisor, Professor Randy Ellis at Queen’s University, suggested she consider sparse subspace clustering, a clustering technique designed for high-dimensional data. “I applied an algorithm that’s normally used for facial recognition and video processing to detect cancer, and I got lucky that it was successful.”

Theriault is now pursuing a master’s degree at Queen’s University in a Vector-recognized master’s program where she continues the project but with an expanded scope. She’s now also designing ways to visualize data to provide richer information about the presence of cancer in a pixel or cell and examining how machine learning can also improve the analysis of skin, liver, and prostate cancers.

Theriault has been awarded a Vector Scholarship in Artificial Intelligence to support her graduate work. The entrance scholarship is valued at $17,500 and open to graduate students studying in AI-related master’s programs in Ontario. Theriault says, “My thesis supervisor brought the Vector Scholarship to me and all the undergrads at the time. I actually didn’t realize that what I was doing was considered AI. He told me that it’s AI in health care, and that I should consider it  if I want to continue my research. So I did, and I got it, which was beyond exciting, and now that has helped me fund all of my master’s work.”

Being part of the Vector community also provided value in other ways. Theriault joined Vector-hosted expert talks, including some that brought health perspectives to AI. “I learned a lot about AI and AI in the real world through them,” says Theriault.

When asked about what’s next, Theriault says, “I know I love the research I’m doing. I know I feel like I have a purpose while I’m doing it. I’m getting good feedback, and I feel happy because I’m in a place with a lot of collaboration. I get to go to meetings where I talk with surgeons, chemists, and other computer scientists, and I get to be the one that pulls all the ideas together, along with my supervisor of course. And that makes me very happy.”

She continues: “The motivation is to just keep learning and keep discovering as much as I can.”