A computer model that reads brain scans in glioblastoma
Original title: Multiscale Multiparametric MRI Deep Learning for Short-Term Survival Assessment in Glioblastoma.
How far along is this research?
- Lab cells
- Animals
- Review
- Tested in people
This was tested in people. That is the most reliable kind of research we share.
This was a look back at old brain scans, and the model is not used in care yet.
The short version
A computer model read brain scans to spot who might live 9 months or less.
What was studied. Researchers looked back at MRI brain scans from 728 adults with new glioblastoma: The fastest-growing type of glioma (grade 4). Treatment usually starts soon after diagnosis: surgery first, then radiation and chemotherapy. See the glossary. They built the model using 290 people, then tested it on three other groups of 225, 182, and 31 people.
What they found. The model looked at the whole brain and at the tumor. It did better at spotting short survival than brain scan shape and patient facts alone. That gain held up in the first outside group after extra checks. The model's answers also lined up with genes tied to the immune system and to cell growth.
What this means, and what it doesn't
What it could mean: In time, a tool like this might help doctors see who is at higher risk before surgery. That could help with planning care and talks about it. It does not change treatment today.
What it doesn't mean: This is not a new treatment. The model only sorts risk, it does not treat the tumor. It was tested on scans that were already taken, not on patients in real time. It is not used in normal care, and it is not a promise of a cure. A model score cannot tell one person how long they will live.
Source: PubMed, August 15, 2026 · Read the original
This plain-language summary was written by AI and published automatically after passing our automatic safety checks. How we write.