Reliability testing of in vivo 1H-MRS-signals and elimination of signal artifacts by median filtering

In order to make SV-MRS clinically viable, assessment of spectral quality and reliability of the data should preferably be handled by the MR-scanner system rather than by medical staff. The reliability of the data is most affected by patient motion during acquisition. In this paper, we devise, apply, and test two statistical methods that can be used for automated data reliability- and quality-assessment. Next, we establish that, in order to assess the reliability of SV-MR-data, it is essential to store the data of each acquisition separately, rather than averaging the data of all acquisitions irreversibly prior to storage. We propose a) statistical tests on these separately stored acquisitions that can reveal artifacts, b) application of a special type of order-statistics filtering, namely median filtering, once artifacts have been detected. Furthermore, we develop computer algorithms that provide automated, fully user-independent information on spectral quality and reliability of the acquired spectra. Finally, we conclude that combination of statistical tests with median filtering can provide user-free quality and reliability assessment and improvement of SV-MR-spectra.