Published in a magazine GastroenterologyThis study documents the creation of SÉMIL (Semantically-Enhanced Multiple Instance Learning), an AI algorithm that uses images and written descriptions to analyze routine pathology slides and identify patients at high risk for recurrent bowel cancer.
The researchers analyzed more than 1,600 pathology slides and validated their results in 1,220 patients with stage 2 colorectal cancer from three independent cohorts and multiple centers in Australia.
The study also found that concordance between AI-based and pathologist assessments resulted in the most accurate risk assessment for stage 2 cancer patients.
Lead author and PhD candidate Frances Madison, from La Trobe University’s School of Computing, Engineering and Mathematical Sciences (SCEMS), said the technique assesses the tumor’s growth pattern at the invasive front, a feature that is prognostically important but difficult for pathologists to classify consistently. Each tumor was then assigned to a high-risk or low-risk group.
“This information could help pathologists and clinicians identify patients with stage 2 cancer who are at high risk of recurrence and may require close monitoring and additional treatments such as chemotherapy,” Francis said.
Current Australian clinical guidelines recommend only postoperative chemotherapy for patients with high-risk stage 2 colorectal cancer. Therefore, it would be of great value to more accurately determine which patients are at high risk.
Associate Professor Jen He, head of the Digital Biology Program at the La Trobe Institute for Molecular Sciences (LIMS), SCEMS and the Australian Center for Medical Innovation and Artificial Intelligence (ACAMI), said the research highlighted the growing role of AI in supporting precision medicine.
“SÉMIL can be integrated into existing digital pathology workflows to provide a more consistent and objective assessment of cancer pathology, without the need for expensive new tests or tissue samples,” Professor He said.
“In the future, AI-based pathology assessment could be combined with other emerging biomarkers to improve risk stratification and treatment planning.”
Associate Professor David Williams, an anatomical pathologist at the Olivia Newton-John Cancer Institute, which is affiliated with Austin Health and the La Trobe School of Cancer Medicine, said the tool was designed to support, not replace, clinical decision-making.
“One of the biggest challenges in stage 2 colorectal cancer is identifying which high-risk patients need treatment and weighing the potential benefits and side effects of administering chemotherapy,” Associate Professor Williams said.
“Our study shows that AI-based assessment has the potential to provide pathologists with additional information that may help clinicians make more informed decisions about next-step treatment options.”
Bowel cancer is the fourth most commonly diagnosed cancer in Australia and the second leading cause of cancer death. It is the third most common cancer worldwide.
reference: Magisson F, He Z, Millward J et al. AI-assisted risk stratification in stage II colorectal cancer: multicenter validation of semantically enhanced deep learning. Gastroenterology. 2026:S0016508526070691. doi: 10.1053/j.gastro.2026.07.009
This article has been reprinted from the following material: Note: Materials may be edited for length and content. Please contact the citation source for more information. You can access our press release publishing policy here.
