A computer tool that can say it is not sure about a brain scan
Original title: Uncertainty-Aware Selective Prediction in Brain-Tumour MRI Classification: Selective Risk Reduction and Component-Wise Reliability Findings.
How far along is this research?
- Lab cells
- Animals
- Review
- Tested in people
This was done on cells in a lab, not in people. It is a very early step.
This was only tested on stored scan files, not on patients.
The short version
A computer program that reads brain MRI scans did better when it was allowed to skip the scans it was unsure about.
What was studied. Researchers tested a computer program on 1000 stored brain MRI images. The program could hold a scan back when it was not confident, instead of guessing.
What they found. When the program held back 1.80% of scans, its accuracy on the rest rose to 97.76%. Its error rate dropped from 3.40% to 2.24%. The scans it held back were the ones it was most likely to get wrong. Another part of the work did worse. Sorting gliomas into high grade or low grade was no better than guessing.
What this means, and what it doesn't
What it could mean: A tool that admits when it is unsure may be safer than one that always gives an answer. Those unclear scans could be passed to a person to read. This work also shows that one part of a tool working well does not mean all of it works.
What it doesn't mean: This is early computer work, not care you can get. The program was tested on saved image files, not on patients in a clinic. The authors call this finding exploratory and say they could not run all the checks they wanted. It does not diagnose anyone. It is not a treatment and it is not a cure.
Source: PubMed, October 12, 2026 · Read the original
This plain-language summary was written by AI and published automatically after passing our automatic safety checks. How we write.