Digitizing the diagnostic process

University of Waterloo professor Alex Wong has developed biomarkers that have allowed radiologists to better identify lung and prostate cancer on medical images.</p>

In an interview with Imprint, Wong explained that the goal of the project was to improve medical diagnosis. “One of the things with doing diagnosis on images … [whether it be] x-ray images, CT, MRI, the information is actually not very clear.”

According to Wong, it can take several years of experience to determine if an anomaly on a medical image is actually a tumour, and the means of determination can be subjective. Due to the variability of the observations, radiologists are likely to have different opinions on the same image. This is not the ideal — with medical tests, it is important that the factors contributing to a diagnosis are reliable and valid, considering treatment will depend on the answer.

“We built a new, more quantitative kind of approach to help them make better clinical decisions. It’s something we call Discovery Radiomics.” Discovery Radiomics is a program that works so that when a patient undergoes a diagnosis, their data is stored with the hospital. This way, the hospital has a mass of information for reference. Discovery Radiomics then takes this data and uses it to determine quantitative biomarkers for recognizing cancer. “We have found it to be very successful,” Wong stated, “and [it] gave us very good detection rates.”

Wong and his team were able to successfully implement their program via extensive research and data collection. Instead of trying to mimic the intuition of a radiologist and quantify the irregularity seen in an image, the team flipped their thinking, as Wong explained:

 “‘Why don’t we try to look for biomarkers directly based on data?’ So based on that notion, we came up with a mathematical model that searches through radiomics features and then we went through a comprehensive analysis based on data we received from hospitals [from] past patients to see if it works.” It did – quite well.

Discovery Radiomics is still in its first phase, but it is working with doctors and patients now for testing purposes.

The program is used, at the time of publishing, for lung or prostate cancer, but Wong said he would like to expand. By implementing the Discovery Radiomics in several other hospitals, more data can be collected and thus more biomarkers can be discovered. Wong commented his hope to go beyond the health sector. “Can we figure out the best products out of these products based on the characteristics within them, applying them to remote sensing in the environment? To figure out, is there a climate change aspect we’re looking for? Is there an oil spill? So beyond that, instead of quantitative biomarkers, can we look for more features or other tasks? Essentially, [we’re] trying to spread this research into a variety of fields.”

Wong and his team have a few other projects being worked on, such as a new MRI that captures cancer better and improvements to a blood-flow imaging system of theirs. “These kinds of projects are good at bringing staff and students together to try and solve a real-world problem,” Wong said. Most of the team members are former UW undergrads.

Wong believes that these projects aid in applying theory learnt in lecture to real world situations, such as pattern recognition or machine intelligence courses. “With these kinds of projects, you think ‘oh yeah, I could actually apply [this class] to [this project] and make a difference for a lot of people suffering from these disease.’ So I’m always recruiting undergraduate research assistants to help with these kinds of projects.”