Search Papers | Poster Sessions | All Posters
Poster B141 in Poster Session B - Thursday, August 8, 2024, 1:30 – 3:30 pm, Johnson Ice Rink
Connectome Fingerprinting Predicts Prefrontal Cortical Activation During Abstract Reasoning
Kylie Isenburg1,2 (), Yixiang Liu3, Thomas Morin4,5, Chantal Stern2,3; 1Graduate Program for Neuroscience, Boston University, 2Cognitive Neuroimaging Center, Boston University, 3Department of Psychological and Brain Sciences, Boston University, 4Department of Radiology, Massachusetts General Hospital, 5Department of Psychology, Brandeis University
A key goal of cognitive neuroscience is to understand meaningful differences in brain activity and how this activity maps onto human behavior. Functional Magnetic Resonance Imaging (fMRI) can be used to non-invasively assess human cognition, but it is vulnerable to the blurring of individual differences due to group averaging. Connectome Fingerprinting (CF) is a machine learning technique that uses resting-state brain connectivity profiles to make predictions about individual brain activity patterns. This is useful in brain areas including the prefrontal cortex (PFC), where activity patterns are highly variable across individuals. In this study, we used ridge-regression CF to predict activation in the lateral PFC during an abstract reasoning task. Our results demonstrate that CF is better able to predict individually specific activation patterns compared to the group average. Additionally, the results suggest that model accuracy is influenced by within-participant activation variability. In summary, our study used CF to predict task-evoked activation in the lateral PFC at the individual participant level during an abstract reasoning task. The results showed that CF results in a more accurate prediction of individual brain activity compared to the group average.
Keywords: connectome ridge-regression fMRI machine learning