Anyone who has tried to sit completely still knows how hard it is. Even tiny movements, a heartbeat or a breath can interfere with delicate measurements.
In brain research, those small disturbances have long stood between scientists and a clearer understanding of how the brain truly works. Now, researchers say artificial intelligence may finally be tipping the balance.
A long-standing problem in brain scans
Functional MRI, or fMRI, is one of the most widely used tools in neuroscience. It allows scientists to observe brain activity without surgery and has been central to tens of thousands of studies in recent years.
Yet the technology comes with a persistent flaw: signals from actual brain activity are mixed with distortions caused by motion, blood flow and other physiological factors.
Also read: A targeted method improving early melanoma detection
According to researchers at Boston College, this “noise” has limited how precisely scientists can interpret fMRI data, especially when studying subtle brain responses or neurological disorders.
How artificial intelligence changes the picture
In research published in Nature Methods, a team led by associate professor Stefano Anzellotti developed an AI-based approach called DeepCor.
The method uses generative artificial intelligence to distinguish between patterns linked to active brain tissue and patterns that appear in regions without neurons, such as fluid-filled spaces.
By identifying and removing overlapping patterns, the system isolates signals that are more likely to reflect real neural activity.
Also read: These ingredients can cause blood pressure to spike within minutes, experts say
In tests on both simulated and real fMRI datasets, the AI method significantly outperformed existing techniques, improving noise removal by more than 200 percent in some cases.
The work was carried out with contributions from postdoctoral researcher Aidas Aglinskas and then-undergraduate Yu Zhu.
What it could mean for neuroscience
Cleaner brain imaging data could accelerate research into perception, cognition and brain disorders.
The researchers also see potential for improving large public brain datasets, allowing other scientists to benefit quickly without changing how scans are collected.
Also read: Chronic pain does not necessarily age the brain, study shows
Next steps include making the method easier to access and applying it broadly across existing research data.
Key potential benefits include:
- clearer detection of brain responses
- more reliable comparisons across studies
- improved research into neurological conditions
Sources: News Medical and Nature
Also read: New research sheds light on the effects of Dry January
Also read: Scientists find link between daily steps and reduced Alzheimer’s risk
This article is made and published by Mie Hermansen, who may have used AI in the preparation.