Nuclear magnetic resonance spectrum denoising method based on neural network algorithm

Abstract

The invention discloses a nuclear magnetic resonance spectrum denoising method based on a neural network algorithm, and relates to a nuclear magnetic resonance spectrum denoising method. The method comprises the steps of simulating noiseless and noisy FID signals by utilizing matlab program codes, and generating a nuclear magnetic resonance simulation pure wave spectrum and a noise wave spectrum after Fourier transform; establishing a spectrogram data set of a plurality of samples; dividing the whole data set into a training set, a verification set and a test set; designing a neural network structure and selecting neural network hyper-parameters by using the verification set; testing and checking the neural network model by using the test data; further testing and inspecting the neural network model by using the experimental data; the denoising function of the nuclear magnetic resonance noise spectrum can be quickly realized, and the experimental time for collecting the nuclear magnetic resonance spectrum is saved. Noise of one-dimensional and two-dimensional nuclear magnetic resonance spectra can be removed, and spectral peaks of the spectra are reserved; the method is also suitable for removing noise of noise spectra of multiple signal-to-noise ratios, and spectral peaks of the spectra can be reserved. And the method has better robustness for various different noise wave spectrums.

Description

technical field [0001] The invention relates to a nuclear magnetic resonance spectrum denoising method, in particular to a nuclear magnetic resonance spectrum denoising method based on a neural network algorithm. Background technique [0002] Nuclear Magnetic Resonance (NMR) spectroscopy is widely used in the study of the structure, interaction and dynamics of biological macromolecules. However, NMR spectroscopy is less sensitive than other spectroscopic methods such as mass spectrometry, making NMR analysis require larger sample volumes or longer sampling times to obtain spectra with sufficient signal-to-noise ratio (SNR) picture. Therefore, when the sample amount is small or the sample is not stable enough, NMR detection will encounter difficulties. The SNR of the NMR spectrum is determined by two factors: signal strength and noise level. Therefore, ways to improve the SNR include signal enhancement and noise suppression: 1) Improving the spectrum SNR through NMR signal ...

AI Technical Summary

Problems solved by technology

However, the sensitivity of NMR spectroscopy is low compared to other spectroscopic methods such as mass spectrometry, which makes NMR analysis require larger sample volumes or longer sampling times to obtain spectra with sufficient signal-to-noise ratio (SNR). picture

Therefore, when the sample amount is small or the sample is not stable enough, NMR detection will encounter difficulties

Images