Introduction
In general, proton MRS, as implemented on modern clinical MR machines, provides a reliable adjunct to the growing cadre of imaging methods. Pitfalls can be minimized using automation and standard protocols.[1] Given that the spectral patterns are well known, minor artifacts are relatively easy to identify and read through, at least for the large signals in the spectrum, such as choline (Cho), creatine (Cr) and N-acetyl aspartate (NAA). However, the current trend is toward the incorporation of second tier markers, such as lactate (Lac), glutamate (Glu), glutamine (Gln) and myo-inositol (mI), as well as a demand for longitudinal studies with narrow repeatability requirements. These applications require an understanding of potential artifacts, and the limits of the existing remedies. To achieve repeatability at the limit of biological variation may, in fact, require the development of new artifact reduction algorithms. This chapter details artifacts, remedies, and trade-offs that impact the quantitative use of in vivo MRS. The focus is on proton spectroscopy and spectroscopic imaging (SI) of cerebral metabolites, but the basic principles can be applied to other nuclei, and other parts of the body. With a few notable exceptions, most of the advances in MRS artifact reduction were developed for proton neurological applications at 1.5 T using orthogonal-slice localization methods, most notably stimulated echo acquisition mode (STEAM) and double spin echo (SE) point resolved spectroscopy (PRESS).
Pre-scan
Pre-scan operations, including prescription, sequence, variable selection, shimming, transmit gain and water suppression, all impact the limits of detection and repeatability of the examination. Even with automation and reproducible, predefined protocols, attention to the details of the pre-scan may improve repeatability. Some applications still benefit from manual pre-scan operations. Routine maintenance and quality-control checks using a standard MRS phantom will also reduce the chance for system contributions to variance.