NMR Data Processing Based On Compressed Sensing

Nuclear magnetic resonance (NMR) spectroscopy is an important part of analytical chemistry, which has become an indispensable analytical method in structural biology, metabolomics and other hot subjects. However, the experimental time of the mD NMR spectroscopy accumulates dramatically as the indirect dimensions increase, hence obtaining enough spectral resolution or signal-to-noise ratio(SNR) often needs a longer sampling time. This greatly limits the application of NMR spectroscopy, especially for those samples that can only be obtained in small amounts or only exist for a short time, in addition to high-flux experiments. Numerous approaches have been developed to improve the spectral quality in the aspects of hardware and software, which consequently increases the experimental cost and difficulty. By comparison, data post-processing is more cost-effective, and has wider developing prospective. This paper puts forward new methods of NMR data processing based on compressed sensing(CS), to improve the SNR of NMR spectra without increasing the sampling time, and eliminate the truncation artifacts caused by the signal truncation. The main contents of this paper are as follows:(1) Based on statistical resampling principle, we have proposed a data post-processing method named NASR (An Effective Approach for Simultaneous Noise and Artifact Suppression in NMR Spectroscopy), that can be used to reduce unwanted noises and suppress artifacts in 1D and mD NMR spectroscopy effectively. However, the optimal parameter setting for it is usually difficult to achieve. In the manuscript, Compressed sensing (CS) has been introduced into NASR, This improvement could remove subjective factors of the handlers effectively, and would be a novel robust noise suppression method in 1D and mD NMR.(2) CS has also been used to eliminate thetruncation artifacts caused by the truncated free induction decay (FID). The truncated FID is usually produced when the sampling time or the number of the sampling points is limited. The main achievement of this method is that it can retain the spectral resolution without broadening the spectral linewidth, which makes it especially suitable for the construction of the crowded spectra.