Constructing the human brain metabolic connectome with MR spectroscopic imaging reveals cerebral biochemical organization.

Abstract

Network science has mapped brain structural and functional organization, yet capturing metabolic contrast remains a major gap in connectomics. Using fast, high-resolution 3D whole-brain proton magnetic resonance spectroscopic imaging (¹H-MRSI), we derive a within-subject metabolic connectome in 51 healthy subjects, defined as pairwise correlations among five metabolites (tCr, tNAA, Glx, Ins, Cho) across gray-matter parcels. Results show stability, consistency and replicability, including validation in an independent sample (N = 13) scanned at a different site. A dimensionality reduction analysis shows that the leading metabolic similarity mode forms a continuous caudal-to-rostral gradient across gray-matter regions. We show that this progression is reflective of a balance between local metabolic homogeneity and global metabolic diversity, and can be summarized by a principal path through the metabolic network. While the most metabolically active regions overlap with structural hubs, overall metabolic similarity aligns weakly with tractography-based structural connectivity but more closely with cytoarchitectonic similarity and gene co-expression matrices. These findings introduce the metabolic similarity gradient as a signature of the brain's biochemical organization and position MRSI as a biologically grounded dimension for connectomics in health and disease.