When Francis Magisson started building a new computer program to study bowel cancer, he had a simple goal: help doctors figure out which patients might get sick again after treatment.
Three years later, that program — called SÉMIL — is showing real promise. Researchers at La Trobe University in Melbourne, Australia published their findings in the journal Gastroenterology. Their AI tool, which learned to read medical slides, could help predict which patients with stage 2 bowel cancer are most at risk of their cancer coming back.
Bowel cancer is a serious disease. In Australia, it is the fourth most commonly diagnosed cancer and the second most common cause of cancer death. Around the world, it ranks as the third most common cancer. For patients with stage 2 bowel cancer — where the disease has grown but not spread far — doctors often struggle to know who really needs extra treatment like chemotherapy and who does not.
SÉMIL works by analyzing pictures of tumor cells taken during routine lab tests. The computer looks at how the tumor is shaped and growing, especially at its outer edge, a region called the invasive front. Magisson, a Ph.D. candidate who led the study, said this area is important for predicting outcomes, but it is hard for pathologists — the doctors who study tissue samples — to evaluate consistently every time.
To build and test their tool, the research team studied more than 1,600 pathology slides from patients. They then checked their results against three separate groups of 1,220 stage 2 bowel cancer patients from multiple hospitals and institutions across Australia. The AI performed well: when its assessment matched a pathologist's evaluation, the combined result was the most accurate of all.
"This information could be used to assist pathologists and clinicians to identify which stage 2 cancer patients are at higher risk of relapse and may need closer monitoring or additional treatment such as chemotherapy," Magisson said.
The current approach in Australia recommends chemotherapy only for high-risk stage 2 patients after surgery. That means more accurate risk sorting could spare some patients unnecessary treatment while making sure others get it when they truly need it.
Associate Professor Zhen He, who oversees the digital biology program at La Trobe, said the tool could fit into existing hospital lab systems without requiring new expensive tests. "SÉMIL could be integrated into existing digital pathology workflows without requiring expensive new tests or tissue samples," he explained.
Associate Professor David Williams, an anatomical pathologist who also worked on the study, said the technology is meant to support doctors, not replace them. "One of the biggest challenges in stage 2 bowel cancer is identifying which higher-risk patients require treatment and weighing the potential benefits of administering chemotherapy against side effects," he said.
The research team is hopeful that in the future, AI tools like SÉMIL could be combined with other emerging tests to further improve how doctors plan treatment for bowel cancer patients.