AI radiomics flags pancreatic cancer years before diagnosis
Researchers at Mayo Clinic and MD Anderson developed REDMOD, an AI radiomics model that analyzes CT scans to detect pancreatic cancer years before diagnosis, flagging 46 of 63 pre-diagnosis cases (about 73%) an average of 16 months early (some up to 3 years) and outperforming radiologists, who identified signs in 38.9% of cases. The system also produced 81 false positives among 430 healthy controls. While promising across multiple datasets, researchers caution that larger, more diverse validation and smoother clinical integration are needed before use in routine care.