AI Spots Pancreatic Cancer and Alzheimer's Years Before Doctors Can
Two new studies show AI reading routine scans and patient records can flag disease years early — but neither tool has been tested in real clinics yet.
What happened: Mayo Clinic researchers built an AI model called REDMOD that scans routine abdominal CT images and flagged 73% of pancreatic cancers an average of 16 months before diagnosis, nearly double the rate of specialists reading the same scans. Separately, researchers in China built a deep learning model that fuses brain MRI scans with ordinary clinical records to classify Alzheimer's stages, hitting about 95% accuracy distinguishing Alzheimer's patients from healthy people, and 87-90% accuracy on the trickier call between mild cognitive impairment and early dementia.
Why it matters: Both diseases are hard to catch early with today's tools. Pancreatic cancer is usually found late, when treatment options are limited, and Alzheimer's diagnosis relies on costly PET scans or spinal fluid tests that aren't available everywhere. If AI can pull extra warning signs out of scans and records clinics already collect, it could widen access to early detection without new equipment or extra procedures for patients.
How it works, plainly: The pancreatic cancer model looks for subtle texture changes in tissue on CT scans that are invisible to the human eye, especially useful in patients with new-onset diabetes, a known risk flag. The Alzheimer's model runs brain scans through neural networks built to notice faint patterns in regions like the hippocampus, then combines that with information like age, education and symptoms, rather than judging the scan alone.
The rollout: Both results come from retrospective research, not real-world clinical trials. The Mayo study is published in Gut; the Alzheimer's work in BMC Medical Imaging, using a US research database called NACC. Researchers say the next step is testing these models on new scanners, hospitals, and patient groups before doctors could rely on them, and neither study yet proves that catching disease earlier actually changes patient outcomes.