Use of the Cadzow procedure in 2D NMR for the reduction of t1 noise

Abstract

A data processing approach is proposed for reducing the t(1) noise observed in multidimensional NMR spectra. This method is based on the use of the Cadzow procedure [Cadzow, J.A. (1988) IEEE Trans. Acous. Speech Signal Proc., 36, 49-62], and is demonstrated to be efficient for simulated cases as well as real experiments.

Two‐dimensional Fourier transform ion cyclotron resonance mass spectrometry: reduction of scintillation noise using Cadzow data processing

“…A second method, relying on the singular value decomposition of matrices and less specific to NMR, was developed by Cadzow et al and used successfully to reduce t 1 noise. [29,30] A third method, developed specifically for metabolomics and called correlated trace de-noising, was developed by Poulding et al [31] Of these three noise reduction methods, the Cadzow algorithm is the most general: it is based on Singular Value Decomposition, does not take into account any assumption on the nature of the data, except that it can be decomposed as a sum of exponentially damped sinusoids, which means it can be used on any data that can be Fourier transformed. There is no correlation hypothesis, and the calculation only relies on the internal coherence of the signal.…”

“…the vertical dimension. [15,30] Therefore, the time transient is Fourier transformed first along the measurement date dimension (t 2 ) and the algorithm is then applied for each f 2 frequency. The de-noised data is then Fourier transformed along t 1 in order to get the clean 2D mass spectrum.…”

Removal of t1 noise from metabolomic 2D 1H–13C HSQC NMR spectra by Correlated Trace Denoising

“…The Cadzow procedure can be used to directly denoise the FIDs acquired in a 2D NMR experiment using a process described by Brissac et al [27]. It leverages mathematical properties of a matrix derived from the time-domain signal to remove all signals apart from those resulting from a specified number of resonance frequencies.…”

“…In [27], the peak count is obtained from a simple threshold-based peak picker, and it is unclear whether small genuine peaks would be identified by such a peak picker were they to be lower in intensity than the t 1 noise. In comparison, the Correlated Trace Denoising algorithm does not require the number of peaks to be known, and it is specifically designed to retain small 'genuine' peaks of intensity lower than nearby t 1 noise.…”

“…A number of highly effective denoising algorithms have been described-such as Reference Deconvolution [24][25][26] and the Cadzow procedure [27]-that might fulfil this role. However, although these techniques are successfully applied to 2D HSQC experiments in many other contexts, their use in the domain of metabolomics is constrained by specific pre-requisites of the algorithms.…”