AI Pathology Tool Predicts Relapse Risk in Stage II Colorectal Cancer

In stage-two disease, determining which patients will relapse remains a persistent diagnostic challenge, with Australian guidelines recommending chemotherapy after surgery only for those considered high risk. Improved risk stratification could therefore help better target monitoring and adjuvant treatment. Researchers now report an AI approach that analyzes routine pathology data to predict relapse risk in stage-two bowel cancer.

La Trobe University (Victoria, Australia) researchers developed SÉMIL (Semantically-Enhanced Multiple Instance Learning), an AI algorithm for risk stratification in stage-two bowel cancer. The system uses images and written descriptions derived from routine pathology slides to assess tumor architecture. Specifically, it evaluates growth patterns at the invasive front—a feature recognized as prognostically important but difficult for pathologists to classify consistently—and then assigns tumors to higher- or lower-risk groups.

Published in Gastroenterology on July 29, 2026, the study documented the creation of SÉMIL and analyzed more than 1,600 pathology slides. Findings were validated across 1,220 stage-two bowel cancer patients drawn from three independent cohorts and multiple Australian institutions. The study also found that the most accurate risk ratings occurred when the AI-based assessment aligned with pathologist evaluation.

According to the authors, SÉMIL could be integrated into existing digital pathology workflows without requiring additional tissue or new tests, supporting more consistent and objective pathology assessments. They also noted the potential to combine AI-based pathology assessments with emerging biomarkers to refine risk stratification and treatment planning. The research involved collaborators from the La Trobe Centre for Molecular Science, La Trobe School of Cancer Medicine, WEHI, Monash University, University of Melbourne, UNSW Sydney, Peter MacCallum Cancer Centre, and other Australian cancer centers.

“This information could be used to assist pathologists and clinicians to identify which stage-two cancer patients are at higher risk of relapse and may need closer monitoring or additional treatment such as chemotherapy,” said Francis Magisson, PhD candidate and lead author, La Trobe University’s School of Computing, Engineering and Mathematical Sciences.

“SÉMIL could be integrated into existing digital pathology workflows without requiring expensive new tests or tissue samples to provide more consistent and objective assessments of cancer pathology,” said Associate Professor Zhen He, who leads the Digital Biology program at La Trobe’s Institute for Molecular Science, the School of Computing, Engineering and Mathematical Sciences, and the Australian Centre for Artificial Intelligence in Medical Innovation.

Related Links
La Trobe University