A team of neuroscientists from the University of California San Diego School of Medicine has introduced a novel neuroimaging approach that promises to revitalize the research efforts of many brain scientists. Their findings, published in Cerebral Cortex, describe a method for using neuroimaging data to predict cognitive elements and behavioral variables. This technique involves analyzing widely used magnetic resonance images (MRIs) of brains to identify behavioral associations, then applying these predictions to independent, unseen samples.
Carolina Makowski, Ph.D., a postdoctoral fellow in the Department of Radiology at UC San Diego School of Medicine, explains that researchers typically use MRIs to study specific brain structures to understand their role in behaviors or abilities like short-term memory and problem-solving. Makowski, the first and corresponding author of the paper, highlighted that their work challenges the common belief that neuroimaging studies need thousands of participants to be meaningful. “It’s not always possible to have thousands of individuals,” said Makowski. “Many studies and grants are based on smaller datasets.”
According to co-author Terry L. Jernigan, Ph.D., professor of cognitive science, psychiatry, and radiology at UC San Diego, and director of the Center for Human Development, the study team found that multivariate methods could produce satisfactory results with just a few dozen participants. Typical neuroimaging studies rely on univariate analyses, which test correlations between a single brain point and one behavior, often requiring thousands of participants. In contrast, multivariate studies examine patterns of associations, enabling predictions from smaller datasets. The team saw the most substantial effects by focusing on task-based functional MRI data, particularly those related to working memory processes. These patterns were predictive of general cognition and relevant task-related behavior.
The team trained and tuned large datasets using the Adolescent Brain Cognitive Development (ABCD) study database, which includes MRIs from about 12,000 nine- and ten-year-olds. Jernigan noted that multivariate methods, often referred to as AI, help train large datasets, enhancing predictive power when applied to smaller studies. The results show that reproducible brain-wide association studies can be achieved without thousands of participants.
Makowski added that a larger training sample allows for smaller actual studies. Even with strong initial effects, training samples don’t necessarily need thousands of individuals. For example, a training sample of 5,000 children can predict cognition from working-memory-related brain activation with just 40 children in the replication study. Even with only 100 children in the training sample, they could still predict cognition well with 60 children in the replication.
Senior and co-corresponding author Anders M. Dale, Ph.D., professor of neurosciences, psychiatry, cognitive science, and data science at UC San Diego, believes this method will help scientists fully utilize the ABCD database by focusing on smaller relevant segments for different health outcomes. The large size of the ABCD study allows for examining effects in subsets of individuals, essential for studying outcomes like substance use, psychiatric conditions, and dementia. Dale added that the predictive power of this small-sample approach could be further enhanced by incorporating other analyses, such as genetics studies or different imaging types.
Co-author Tim Brown, Ph.D., a researcher in the Department of Neurosciences, emphasized the significance of their publication, which reopens opportunities for brain researchers worldwide. This comes after a 2022 paper suggested that only large databases could yield valuable results, leaving many researchers with smaller studies at a disadvantage. “You have a whole bunch of grant holders who don’t have thousands of individuals in their studies,” said Brown. “And they’re suddenly being told: ‘You can’t do reproducible work.’ What we’re showing here is that reproducible brain-wide association studies do not require thousands of individuals.”
Source: University of California – San Diego
- : public
