Doctors often use MRI scans to look at the brain and find problems like strokes, tumors, or diseases such as multiple sclerosis.
But reading these scans takes time and there are not enough radiologists in many hospitals. This means people may have to wait longer for a diagnosis and treatment, which can affect their health. A new study from King’s College London shows that artificial intelligence (AI) might help solve this problem.
In the study, researchers created an AI tool that can look at brain scans and quickly tell whether they are normal or abnormal. The AI was trained using more than 60,000 brain MRI scans along with the written reports made by radiologists.
This means the AI learned not only from the images but also from the words doctors used to describe what they saw. This helped the AI understand what brain problems look like in real life.
The special thing about this AI is that it didn’t need humans to label each image manually. Normally, training AI like this takes a lot of time and money because expert doctors have to go through thousands of scans and mark them.
But in this case, the AI used both the scan and its matching report to teach itself. This saved time and allowed the system to learn from a large number of cases.
Once trained, the AI was tested on new MRI scans it had never seen before. These included images showing strokes, brain tumors, and multiple sclerosis. The AI was able to identify these problems correctly, which shows that it could be helpful in real hospitals.
The AI system can also be used in other ways. For example, if a doctor wants to look at all the past cases of a brain tumor called glioma, they can type this word into the system. The AI can then find and show similar scans, which could help doctors study cases or teach medical students.
This new tool could be used during a patient’s scan to immediately flag anything that looks wrong. This can speed up the process and help doctors make decisions faster. It can also suggest what the problem might be, check if a report is missing something important, or find similar past cases for comparison. All of these features could make doctors’ work easier and improve care for patients.
The researchers say the next step is to test this AI system in real hospitals across the UK. They plan to run a large trial starting in 2026 to see how the tool works in everyday practice.
This study shows that AI can become a helpful assistant to radiologists. It can look at scans faster, help spot mistakes, and reduce delays.
However, the final decisions still need to be made by doctors. The goal is to support them, not replace them. If used well, this kind of technology could lead to quicker diagnoses and better outcomes for many people who need brain scans.
The study is published in Radiology: Artificial Intelligence.
Copyright © 2025 Knowridge Science Report. All rights reserved.
