Real and imaginary part symmetries (RIPS) and artifacts removal in 1H magnetic resonance spectroscopy without water suppression

TL;DR

This paper introduces a software technique based on symmetry properties to effectively eliminate sideband artifacts in 1H MRSI data acquired without water suppression, improving spectral quality without additional hardware or prolonged acquisition.

Contribution

A novel SVD-based method utilizing symmetry features to remove sideband artifacts in non-water suppressed 1H MRSI data, outperforming existing hardware/software approaches.

Findings

Significant reduction of sideband artifacts in real and imaginary spectra.

Residual sidebands smaller than those in water-suppressed spectra.

Method performs well on both phantom and high-quality in vivo data.

Abstract

Purpose: Proton MR spectroscopic imaging (1H MRSI) without water suppression (WS) possess some distinct advantages over the conventionally used 1H MRSI with WS. However, the sideband artifacts in the non-water suppressed spectra hinder the applications of the 1H MRSI without WS. Although many hardware or software techniques to tackle the sidebands have been developed, they suffer from various shortcomings, such as the prolonged data acquisition time, insufficient elimination of the sidebands, or loss of some spectral information. Here, we present a software method that allows the elimination of sideband artifacts from 1H MRSI data acquired without WS. Methods The presented method is based on the following observations: (1) the real part of the sidebands is anti-symmetric, and the imaginary part of sidebands is symmetric about the unsuppressed water signal, which is at the center of the spectrum, and (2) there is no observable metabolite resonances in the downfield region spectrum acquired with conventional MRSI pulse sequences. We model the complex sidebands in the downfield region using a singular value decomposition-based method, reconstruct from them the upfield sidebands according to the symmetry features of the sidebands, and subtract the sidebands from the spectrum. We demonstrate the method using phantom and in vivo data. Results The method can remarkably reduce sideband artifacts in both the real and the imaginary parts of the spectrum. In both phantom data and high-quality in vivo data, the residual sidebands are smaller than those in the WS spectra. Conclusion The method allows to remove sideband artifacts in both the real and the imaginary parts of the spectrum. The performance of the method on phantom data and high-quality in vivo data outperforms on poor quality in vivo data.