Sparsity-Based Thresholding Criterion for Spurious Echo Removal and Denoising Magnetic Resonance Spectra Using Rational-Dilation Wavelet Transform

Abstract

Biological signals such as magnetic resonance spectroscopy signals are susceptible to noise and artifacts. The information obtained from these signals is significant in analyzing human physiological conditions. MRS, a nonionizing and noninvasive method, presents an effective alternative method to biopsy for diagnosis and analysis from generated signals that are rich in chemical information of the tissues in the region of interest. A persisting problem of this method is the presence of noise and artifacts causing misinterpretation and subsequent incorrect diagnosis. The present research proposes a denoising strategy using the rational-dilation wavelet transform-based signal decomposition and a thresholding criterion designed using the Lpq-norm-based sparsity measure of the decomposition levels of the signal. Compared with the standard state-of-the-art methods, which are effective in denoising but can cause distortion of the signal at discontinuities, the proposed method can remove artifacts such as spurious echoes present in the magnetic resonance signals and improve the signal-to-noise ratio without distorting the signal. © 2022, Springer Science+Business Media, LLC, part of Springer Nature.