Blind spectra decomposition of MRSI of the brain with tumor by sparse component analysis

Abstract

Magnetic Resonance Spectroscopy Imaging (MRSI) is suitable for analyzing brain tumor metabolites in vivo. Several tissue types may exist in the tumor lesion area, and thus the measured MR spectra are a linear superposition of the constituent spectra of each of the different tissues. In this abstract we describe a novel approach, based on sparse component analysis, to recover the constituent spectra from the measured mixtures. The method is applied to experimental MRSI data of a brain with tumor lesion, and promising result is obtained. Compared to the existing BSD and cNMF approaches, this method is relatively simple and very fast.