Model-based Noise Cancelling of Multichannel Magnetic Resonance Sounding Measurements

The signals recorded in a multichannel magnetic resonance sounding measurement are often heavily dominated by noise, in particular spikes and powerline harmonics. The efficiency of the standard techniques for noise cancelling; multichannel adaptive and Wiener filtering are limited by the fact that they fail to properly account for the presence of multiple noise sources. In this paper, we address this problem through a model-based noise cancelling approach. A model of the powerline harmonic interference is constructed using a weighted least squares method to estimate the fundamental powerline frequency and the amplitude and phase of all relevant harmonics. The model is subsequently subtracted from the signal records. The proposed method efficiently removes the large contribution from powerline harmonics in the MRS signals. Simulations with synthetic signals embedded in real noise records show that the synthetic signals can be recovered with a fidelity similar to or exceeding the standard noise cancelling methods.