Brain Atlas of Macromolecular Spectroscopic Signals from Ultrashort-TE MRSI

Motivation: Macromolecules generate broad baselines in short-TE MRSI data, complicating the quantification of metabolite signals. Goal(s): To develop a brain atlas of macromolecular signals for modelling and effective removal of macromolecule signals in ultrashort-TE FID MRSI at 3T. Approach: We learned the spectroscopic patterns of macromolecule signals in the human brain across different subjects in the standard MNI space. The learned macromolecule atlas is used as probabilistic distribution constraints for subspace-based macromolecule removal. Results: Macromolecule atlas was built from a group of healthy subjects (n=45). It was tested using simulation data and experimental data acquired from healthy subjects and tumor patients, producing encouraging results. Impact: The proposed method may lead to more accurate modelling of macromolecule signals, thus improving the quantification of metabolites in short-TE MRSI.