Fast Volumetric Mapping of Brain NAD Levels Using 7T 31P-MRSI and Learned Probabilistic Subspaces

Motivation: Measuring brain intracellular NAD levels has long been of interest, but the current 31P-MRSI methods would take prohibitively long scan times for mapping NAD. Goal(s): To present a method for fast volumetric NAD mapping of the entire human brain. Approach: In vivo 31P-MRSI scans were performed at 7T with a nominal resolution of 1.0 cc within 20 minutes. A probabilistic subspace-based method integrating spectral prior, spatial constraint, and statistical distributions was applied for denoising. Results: The proposed method successfully provided high-resolution brain NAD mapping within 20-minute scans. The results also showed promises in revealing metabolic tissue heterogeneity and age correlation of NAD. Impact: This work demonstrates the feasibility of volumetric brain NAD mapping with a nominal resolution of 1.0 cc within 20 minutes. It may provide a powerful metabolic imaging tool for many applications.