How Firm Is a Pancreatic Cancer Score? Reading a CT Scoring AI at Several of Its Own Settings

Radiologists read a routine CT scan as normal, yet the pancreas may already carry the first signs of cancer. An AI from the Mayo Clinic (REDMOD) scores such scans. On 493 test scans it found 46 of the 63 that later proved to be cancer (73.0%) and cleared 349 of the 430 healthy scans (81.1%), while about 36 of every 100 flags were right (36.2%). The same study found that 3 of 10 healthy scans wrongly flagged at first were not flagged again on a repeat scan. This note describes a controller that tells, for each scan and on the same day, whether the flag is firm. The controller moves a few of the AI's own non-learned settings a little (such as filter width, gray-level bin width, resampling size and outline margin), reads the spread of the scores, and, if the spread is large, shunts to another validated setting by itself, without the intervention of a human operator. The scan, the trained model and the cutoff are never changed, nothing is retrained, and no new hardware is needed. The note explains the idea in simple steps, compares it with earlier robustness studies and with Bayesian and Monte Carlo dropout methods, and names who can take it up. Keywords: pancreatic cancer, CT, early detection, radiomics, firmness of AI score, robustness, controller, shunting action

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-16
DOI
https://doi.org/10.5281/zenodo.24290518
Primary Topic
AI in cancer detection
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

How Firm Is a Pancreatic Cancer Score? Reading a CT Scoring AI at Several of Its Own Settings

Prakash Vaithyanathan
4 links
Zenodo (CERN European Organization for Nuclear Research)
AI in cancer detection
article

How Firm Is a Pancreatic Cancer Score? Reading a CT Scoring AI at Several of Its Own Settings

Prakash Vaithyanathan
article en

Abstract

Radiologists read a routine CT scan as normal, yet the pancreas may already carry the first signs of cancer. An AI from the Mayo Clinic (REDMOD) scores such scans. On 493 test scans it found 46 of the 63 that later proved to be cancer (73.0%) and cleared 349 of the 430 healthy scans (81.1%), while about 36 of every 100 flags were right (36.2%). The same study found that 3 of 10 healthy scans wrongly flagged at first were not flagged again on a repeat scan. This note describes a controller that tells, for each scan and on the same day, whether the flag is firm. The controller moves a few of the AI's own non-learned settings a little (such as filter width, gray-level bin width, resampling size and outline margin), reads the spread of the scores, and, if the spread is large, shunts to another validated setting by itself, without the intervention of a human operator. The scan, the trained model and the cutoff are never changed, nothing is retrained, and no new hardware is needed. The note explains the idea in simple steps, compares it with earlier robustness studies and with Bayesian and Monte Carlo dropout methods, and names who can take it up. Keywords: pancreatic cancer, CT, early detection, radiomics, firmness of AI score, robustness, controller, shunting action

Zenodo (CERN European Organization for Nuclear Research)
Good health and well-being
AI in cancer detection
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon