AI Breakthrough dramatically Improves Clarity of Brain Scans wiht fMRI Technology
Table of Contents
A new artificial intelligence-powered method promises to revolutionize functional magnetic resonance imaging (fMRI) by considerably reducing image distortions and unlocking new potential for understanding the brain and its disorders. Researchers at Boston College detailed their findings in a recent report published in Nature Methods.
fMRI, a widely utilized neuroimaging technique, relies on detecting brain activity through blood flow changes. Though, the data collected is frequently enough compromised by “noise” stemming from sources like patient movement and physiological processes such as heartbeat. With tens of thousands of studies published in 2024 alone, improving the clarity of these scans is paramount.
The research team, led by Boston College Associate Professor of Psychology Stefano Anzellotti, developed a generative AI approach that demonstrably outperforms existing methods for removing this interference. According to Anzellotti, the new technique effectively “triples the performance of previous approaches.”
“We wanted to improve the removal of noise from fMRI data,” Anzellotti stated. “What is new about our work is that thanks to the use of generative AI we were able to improve by more than 200 percent over previous methods.”
DeepCor: A New Standard in fMRI Denoising
The newly developed method, dubbed DeepCor, has shown remarkable results in both simulated and real-world fMRI data.It surpasses other state-of-the-art denoising techniques,including a commonly used method called CompCor. In tests analyzing brain responses to facial stimuli, DeepCor achieved a 215% improvement in noise reduction. Moreover, when applied to synthetic data mimicking real fMRI datasets, the improvement reached an astounding 339%.
The AI’s success lies in its ability to differentiate between patterns unique to brain regions containing neurons and those found in areas without neurons, such as the ventricles. “Noise typically affects both sets of regions, thus removing the patterns they have in common makes the unique patterns of the regions that contain neurons stand out,” Anzellotti explained.
The research team included post-doctoral researcher Aidas Aglinskas and undergraduate student Yu Zhu.
Unexpected Results and Future Implications
The magnitude of the improvement achieved by DeepCor took the researchers by surprise. “We were surprised by how big the improvement was,” Anzellotti said. “We expected the method to do better, but we anticipated an improvement in the range of 10 percent to 50 percent. Improving by 200 percent was beyond our most optimistic expectations.”
Looking ahead, Anzellotti’s team is focused on two key objectives: increasing accessibility to the DeepCor method for other researchers and applying it to large, publicly available datasets. This woudl allow the broader scientific community to benefit from cleaner, more reliable fMRI data as quickly as possible.
“we are looking at two key next steps: making the method as easy to access for as manny ot
Why: The clarity of fMRI scans was limited by noise from patient movement and physiological processes. Researchers sought to improve the removal of this noise to gain more accurate insights into brain activity.
Who: The research was led by Stefano anzellotti, Associate Professor of Psychology at Boston College, along with post-doctoral researcher Aidas Aglinskas and undergraduate student Yu Zhu.
What: The team developed a new AI-powered method called DeepCor, which utilizes generative AI to significantly reduce noise in fMRI data. DeepCor outperforms existing methods like compcor,achieving improvements of 215% in real-world data and 339% in simulated data.
