Abstract
In vivo NMR spectroscopy (MRS) is a technique that enables non-invasive measurement of metabolite concentrations in patients. Assessment of signal-quality and reliability of clinical MRS data is a task which requires much expert knowledge and cannot be handled by clinicians. To bridge this gap, statistical methods are proposed enabling unsupervised testing of spectral quality and reliability as well as a reconstructive filtering technique for signal-artifact reduction. Separate storage of MRS data is a prerequisite of reliability testing, which includes the computation of the first four central moments (mean, variance, skewness and excess kurtosis). Reliability test-parameters related to the estimated central moments are introduced. Reconstructive filtering by means of all rank selection order statistic filtering (ARSOS) applied to unreliable MRS data is proposed, aiming for minimization of signal artifacts. The properties of ARSOS filtering are described. A parameter accounting for the MRS signal quality is proposed. Illustrative examples of the proposed methods applied to in vivo MRS signals are presented. The mathematics of unsupervised methods for quality and reliability testing of localized clinical MRS signals are described. The described methods are generally applicable to any type of repetitively measured signal.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call