Inference of functional connectivity from direct and indirect structural brain connections

• Published in: 2011 IEEE International Symposium on Biomedical Imaging: From Nano to Macro pp. 849 - 852
• Main Authors: Deligianni, F, Robinson, E, Beckmann, C F, Sharp, D, Edwards, A D, Rueckert, D
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.03.2011
• Subjects: Brain connectivity; Brain modeling; Correlation; Diffusion tensor imaging; functional connectivity; indirect structural connections; Input variables; Predictive models; Principal component analysis; rs-fMRI; structural connectivity; whole-brain connectivity matrices
• ISBN: 1424441277; 9781424441273
• ISSN: 1945-7928
• DOI: 10.1109/ISBI.2011.5872537

We propose statistical inference based on the Least Absolute Shrinkage and Selective Operator (Lasso) regression as a framework to investigate the relationship between structural brain connectivity data (DTI) and functional connectivity data (fMRI). Regions of interest (ROIs) are obtained from an accurate atlas-based segmentation. We use direct structural connections to model indirect (higher-order) structural connectivity. Subsequently, we use Lasso to associate each functional connection with a subset of structural connections. Lasso offers the advantage of simultaneous dimensionality reduction and variable selection. We use a cohort of 22 subjects with both resting-state fMRI and DTI and we provide both qualitative and quantitative results based on leave-one-out cross validation. The results demonstrate that the performance of prediction is enhanced through the incorporation of indirect connections. In fact, the mean explained variance was improved from 54%±6.53 to 58%±4.31 when indirect connections of up to second order are added and the improvement in performance was statistically significant (p <; 0.05).