A new modification of Singular Value Decomposition using Independent Component Analysis with applications to MRS imaging quantification

Abstract

A modification of Singular Value Decomposition (SVD) is proposed in this aiming at de-noising and finding more accurately the statistically independent spectra of metabolite sources in Magnetic Resonance Spectroscopy (MRS) quantification. Although SVD is known in MRS applications and several efficient algorithms exist for estimating SVD summation terms in which the raw MRS data is analyzed, however, modifications of the main methodology incorporating techniques for calculating the assumed statistically independent spectra have not been employed so far. The goal of this paper is to present such a modification based on applying SVD on the MRS spectrogram. A methodology combining SVD on spectrogram followed by an iterated application of Independent Component Analysis (ICA) on the principle components ICA will also be tested. The methodologies are tested and results are discussed by conducting an experimental study based on synthetic MRS signals.