Precise maps of the brain’s deepest corners are made possible through tools developed by these Northeastern researchers
The NeuroPRISM lab, led by assistant psychology professor Stephanie Noble, makes tools that pave the way for reliable and reproducible neuroimaging of the brain.
Scientists who study the brain strive to detect very specific signals from among the billions of neurotransmitter molecules present. Precise and sensitive imaging methods are crucial because some signals are very small.
Northeastern University’s NeuroPRISM lab, led by assistant psychology professor Stephanie Noble, makes tools that pave the way for reliable and reproducible neuroimaging of the brain’s deepest recesses.
“The focus of the lab is both on developing these tools and then showing the community what sort of knowledge can be gained from them,” Noble says. After developing algorithms to help researchers estimate how precise their study findings will be, the lab is now building tools to sharpen precision even further. “This is the most comprehensive version of this, where we can show lots of different people how different types of studies can be improved.”
The lab is focused on “precision neuroscience,” a kind of meta-research to help scientists publish results that other scientists can duplicate. Their data-driven tools can enhance signals collected during functional MRIs — imaging that is like a movie of blood flow in the brain. More broadly, the lab’s work aims to help all neuroscientists do more rigorous research.
“We want our work to both help address problems with reproducibility that are widely acknowledged in the field,” Noble says, “and also to better characterize the individual so we can get closer to personalized ‘precision medicine’ solutions.”
The team received a 2025 National Institute of Neurological Disorders and Stroke Rigor Champions Prize for work “calling attention to issues of statistics, analytical reliability/validity, multiple comparisons, and power analysis.”
Noble uses “precision” to describe both study outcomes and the potential for individualized medicine.
To make fMRI scans more accurate, her lab created a tool that identifies and filters extraneous environmental “noise,” including artifactual signals from within the body, that can drown out signals that clinicians are looking for.
“One of the biggest challenges is separating neural signals from noise,” says Alexandra Fischbach, a graduate student researcher in Noble’s lab. “If you don’t correct for this, you could either miss true effects or risk reporting false positives.”
Fischbach developed a computational algorithm to find physiological noise, including the signal arising from cerebral spinal fluid as it pulsates with a person’s breathing and heartbeat. This vibration can interfere with measuring brain activity.
Editor’s Picks
This noise is a problem for fMRIs of the deep brain because those pulses are happening nearby.
“You don’t really know which is which,” Fischbach says. “Is this a signal of noise, or is this a signal of interest?”
Her open-source tool, available on the lab’s website, helps researchers answer that question.
The lab built two other tools — an effect size explorer and a power calculator — using very large fMRI datasets. The effect size explorer, built by graduate student Hallee Shearer, helps researchers see how different variables will change the precision of a study’s outcome. That information feeds the power calculator, which estimates the sample size that a study might need.
This is important because using human subjects is expensive, says Fabricio Cravo, a computational neuroscientist in Noble’s lab. When a researcher is planning a study, “one of the first things that comes to mind is how many subjects I need,” he says. “If we can give scientists direction of how many subjects they need to do an experiment they want to do, this would help them budget and properly plan a study beforehand.”
The lab also led workshops to educate researchers about how new machine learning algorithms and computational tools can be applied to many fields that suffer from similar reproducibility issues, Noble says.
Better precision at the research level is one of Noble’s objectives. Another is to use the same techniques to develop better insights into an individual patient’s needs.
“That is something that has echoed through some of our previous work to understand test/retest reliability,” Noble says. “We want our work to both help address problems with reproducibility that are widely acknowledged in the field, and also to better characterize the individual so we can get closer to personalized ‘precision medicine’ solutions.”
