AI-driven pathology predicts colorectal cancer relapse and guides irinotecan benefit

A newly developed artificial‑intelligence system analyses routine digitised pathology slides stained with haematoxylin and eosin to generate a risk score for colorectal cancer recurrence. The tool identifies subtle morphological patterns in the tumour micro‑environment, offering a non‑genomic method to stratify patients for follow‑up care and adjuvant therapy.

Separately, researchers at University College London applied a comparable AI computational‑pathology platform to biopsy specimens from the ARISTOTLE phase‑III trial. By quantifying tumour cell density, the algorithm pinpointed patients with locally advanced rectal cancer who derived a substantial benefit from adding the chemotherapy drug irinotecan to standard chemoradiotherapy, reducing recurrence by about 43% and death by roughly 50% in this subgroup. Both approaches aim to personalise treatment, sparing low‑risk patients from unnecessary toxicity while directing more intensive therapy to those most likely to profit.

The technologies are undergoing broader validation across multiple international centres and will require regulatory approval before routine clinical deployment.