This study tested whether AI risk scores could pick out Lung-RADS 1 and 2 scans safe for biennial screening. Sybil, an imaging deep learning model, and the Brock model were tested on 3,120 scans from one safety-net hospital. Cut-offs were calibrated to move 30% of negative screens to biennial screening in a National Lung Screening Trial subset. Sybil's overall AUC was 0.819 to 0.912, and 0.746 at 2 years for Lung-RADS 1 and 2. Its cut-off moved 33.8% of hospital negative screens with no delayed cancers, while Brock's AUC was 0.513. The analysis is retrospective and from one hospital, with 72 cancers over a median 2.3 years of follow-up.
Why it is interesting: Lung-RADS 1 and 2 made up 90.5% of scans, so interval decisions drive much of the screening workload. Sybil held up on external data where the established Brock model performed near chance.
Cite this item
BibTeX
Free sign-up. A digest of the new papers twice a week, grouped by domain.
Other drivers in Data, models and systems
-
Forecasts and their record
How the forecasts are made, how well they explained the past, and how well mortality can be forecast — and by whom.
3 this week -
What linked data reveals
What linked registries make it possible to know about whole populations
9 this week -
AI in medicine
Whether machine learning predicts, detects and decides better than the tools we have, and whether that has yet changed an outcome for a population.
4 this week -
Health systems and access
What the health system can provide and who gets it.
4 this week -
Value in health
What a treatment, health programme or policy is worth against what it costs, and how far the models that decide it can be trusted.
3 this week -
Revisions and new sources
What the figures are built from, what is newly available, and how much the numbers move when they are revised.
2 this week
