HMN 2026: How AI gives a clearer picture of functional MRI brain data

Obtaining clearer functional MRI data about the brain and its disorders is possible using artificial intelligence, according to Boston College researchers who report in Nature Methods that they have developed an AI-assisted method to remove “noise,” or image distortions, caused by movement, heartbeat, and other factors.

How AI improves fMRI data quality

Functional neuroimaging, also known as fMRI, is one of the most widely used noninvasive methods in neuroscience, with tens of thousands of studies published just in 2024. A major obstacle for fMRI research is that MRI data about brain responses is mixed with noise from movements and other sources.

Removing noise more effectively could pave the way to new discoveries about the brain and its disorders, according to Boston College Associate Professor of Psychology Stefano Anzellotti, the paper’s senior author. The new method developed by Anzellotti and two other researchers used generative AI to triple the performance of previous approaches.

The findings could open new doors for brain research, Anzellotti said.

“We wanted to improve the removal of noise from fMRI data,” said Anzellotti. “Other work had attempted to do this before. What is new about our work is that, thanks to the use of generative AI, we were able to improve by more than 200% over previous methods.”

DeepCor’s performance and future plans

The method developed by the researchers, known as DeepCor, outperforms other state-of-the-art denoising approaches on a variety of simulated datasets. In real fMRI data, DeepCor outperforms another widely-used method, known as CompCor, by 215% in removing noise from face responses, and by 339% in clarifying realistic synthetic data, generated to imitate the properties of a real fMRI dataset, according to Anzellotti.

The AI learns which patterns are unique to brain regions that contain neurons and the unique patterns within regions of the brain that do not contain neurons, like the ventricles, said Anzellotti.

“Noise typically affects both sets of regions, therefore removing the patterns they have in common makes the unique patterns of the regions that contain neurons stand out,” Anzellotti said.

The team, including postdoctoral researcher Aidas Aglinskas and Yu Zhu, then an undergraduate student, studied the human brain with functional magnetic resonance imaging.

Anzellotti said the scope of improvement was not expected.

“We were surprised by how big the improvement was,” he said. “We expected the method to do better, but we anticipated an improvement in the range of 10% to 50%. Improving by 200% was beyond our most optimistic expectations.”

Anzellotti’s research will continue to explore improvements in fMRI readings.

“We are looking at two key next steps: making the method as easy to access for as many other researchers as possible, and using it to denoise large public datasets so that the field can start benefiting from cleaner data as soon as possible,” he said.

The content is provided for information purposes only.