A paper published in Nature Communications presents an artificial intelligence test designed to predict the risk that breast cancer will recur. The work was led by Krzysztof J. Geras, a visiting scholar at New York University’s Center for Data Science and an adjunct assistant professor at NYU Grossman School of Medicine.
Yann LeCun, a co-author and professor at New York University, says the model benefits from self-supervised pretraining that helps it learn useful representations before the final prediction. The team developed a multi-modal test that combines routine clinical data with pathology slides. Clinical information included tumor stage, patient age and hormone-receptor status.
The researchers drew on data from 15 patient populations across seven countries and evaluated the test with more than 3,500 patients. They measured accuracy with standard statistics such as the C-Index and a Hazard Ratio. The AI distinguished higher-risk from lower-risk patients, performed well for triple-negative and HER2-positive cancers, and matched or outperformed a widely used genomic test.
The authors stress that completed randomized clinical trials are needed before the AI can guide treatment. Some authors hold equity in Ataraxis AI; Geras is co-founder and chief scientific officer, and New York University maintains financial and intellectual property interests.
Difficult words
- recur — to happen again after earlier treatment
- pretraining — initial learning phase for a machine model
- multi-modal — using more than one type of data
- pathology slide — thin tissue sample on a glass for studypathology slides
- hazard ratio — measure comparing risk between two groups
- genomic test — exam that looks at genes in a tumor
- randomized clinical trial — study randomly assigning treatments to patient groupsrandomized clinical trials
Tip: hover, focus or tap highlighted words in the article to see quick definitions while you read or listen.
Discussion questions
- How could combining clinical data and pathology slides affect patient treatment decisions?
- What concerns would you have about using an AI system to guide cancer treatment?
- Why do you think randomized clinical trials are important before adopting this AI in hospitals?
Related articles
Researchers find 'vibe coding' linked to insecure AI-written code
A research team found that a programming style called "vibe coding" is producing insecure code with help from generative AI tools. A new radar scans public vulnerability data and flags cases that show AI signatures or risky patterns.
