Douglas Flora, Executive Medical Director of Yung Family Cancer Center at St. Elizabeth Healthcare, President-Elect of the Association of Cancer Care Centers, and Editor in Chief of AI in Precision Oncology, shared an article on LinkedIn about a recent article by Sovanlal Mukherjee’s et al, published in Gut:
“Sixteen Months Earlier
Pancreatic cancer still kills most people who get it. We almost never find it in time.
The five-year survival rate for pancreatic adenocarcinoma sits around 12%. That number has barely moved in decades. Not for lack of trying. Because by the time a mass is visible on a CT scan, the opportunity for cure has usually gone.
A study just published in Gut is worth our attention.
Researchers from Mayo Clinic and MD Anderson trained an AI model called REDMOD to read CT scans — routine scans that radiologists had already reviewed and cleared. It wasn’t looking for a tumor. It was looking for radiomic patterns: subtle disruptions in tissue texture and structure, invisible to the human eye.
In 73% of cases — 46 out of 63 patients who would later develop pancreatic cancer — REDMOD flagged the scan as suspicious. On average, 16 months before diagnosis.
Every one of those scans had received an all-clear from specialists. This isn’t at all to suggest criticism of our Radiologists, they are still great, there just wasn’t a mass to see.
Of 430 healthy individuals in the study, REDMOD flagged 81 as suspicious. False positives have consequences: anxiety, additional procedures, cost. This is not a screening tool ready for universal deployment tomorrow. The researchers themselves are calling for larger, more diverse validation studies.
We have spent most of medicine’s history as historians of the cell — waiting for disease to become visible before we could name it. REDMOD suggests we might be able to read the early manuscript instead.
The goal is getting to the conversation earlier. More time to decide. More time to plan. More time to fight on a patient’s own terms.
Not just a better scan. A different kind of hope.”
Title: Next-generation AI for visually occult pancreatic cancer detection in a low-prevalence setting with longitudinal stability and multi-institutional generalisability
Authors: Sovanlal Mukherjee, Ajith Antony, Nandakumar Patnam, Kamaxi Trivedi, Aashna Karbhari, Khurram Bhinder, Armin Zarrintan, Joel Fletcher, Mark Truty, Matthew Johnson, Suresh Chari, Ajit Goenka
Read the Full Article on Gut
More posts featuring Douglas Flora.
