Separation of Macromolecules and Metabolites in Ultrashort-TE MRSI Data with Learned Probabilistic Subspaces

Separation of macromolecules and metabolites in ultrashort-TE MRSI data has been very difficult due to limited SNR and strong spectral overlap. In this work, we proposed a new solution to the problem using a subspace-based approach aided with long-TE navigator signals. Physics-based prior information was incorporated through pre-learned spectral bases and probability distributions of spatial coefficients. The proposed method has been validated using experimental data from healthy and brain tumor subjects, producing impressive results.