A method and system for improving sensitivity of multidimensional spectra in nuclear magnetic resonance

Abstract

The application discloses a method for improving sensitivity of multi-dimensional spectrum in nuclear magnetic resonance, comprising the following steps: collecting time domain data of multi-dimensional experiment according to preset sampling parameters and by using a nuclear magnetic resonance instrument; performing Fourier transform on direct dimensions of the collected time domain data; performing phase correction on the Fourier transform result to obtain corrected data; performing data reconstruction on indirect dimensions of the corrected data column by column by using a non-uniform reconstruction algorithm to obtain data after reconstruction; performing finite impulse response digital filtering on the indirect dimensions of the data after reconstruction column by column to obtain filtered data; and performing Fourier transform on the indirect dimensions of the filtered data to obtain transformed data. The application can solve the technical problem that the existing method for improving atomic polarization degree has special requirements for samples and application conditions, and is difficult to popularize as a universal nuclear magnetic resonance measurement method.

Description

Technical Field

[0001] This invention belongs to the field of nuclear magnetic resonance technology, and more specifically, relates to a method and system for improving the sensitivity of multidimensional spectra in nuclear magnetic resonance. Background Technology

[0002] Nuclear magnetic resonance spectroscopy (NMR) is a crucial research tool in analytical chemistry, providing abundant information on molecular structure and has been widely applied in chemistry, biology, and medicine. However, due to the increasingly complex molecular structures of the substances being analyzed, severe peak overlap in NMR spectra has significantly reduced the resolution of NMR spectra. Multidimensional NMR techniques can significantly improve spectral resolution; however, due to the limitation of low NMR polarizability, sensitivity remains a significant challenge for multidimensional NMR spectroscopy.

[0003] To improve the sensitivity of nuclear magnetic resonance spectra, researchers have proposed several different methods: (1) methods to enhance nuclear polarization, typical techniques include DNP (Dynamic Nuclear Polarization, for details please refer to the website https: / / baike.baidu.com / item / %E5%8A%A8%E6%80%81%E6%A0%B8%E6%9E%81%E5%8C%96?fromModule=lemma_search-box), optical pumping (Pang Zhenfeng, Guan Hanxi, Gao Lina, et al.). The principle and application of polarized nuclear magnetic resonance method [J]. Acta Physico-Chimica Sinica, 2020, 36(4):50-68.DOI:10.3866 / PKU.WHXB201906018.), secondary hydrogen-induced (for details please refer to the website http: / / www.gpxygpfx.com / article / 2020 / 1000-0593-40-3-665.html), etc., which respectively transfer the energy of high polarization such as electrons and lasers to the atomic nucleus, thereby significantly improving the sensitivity of nuclear magnetic resonance signal. (2) Optimize the pulse sequence or data acquisition method, for example through 1 H, 13 C polarization transfer, 1 The polarization shift of the H nucleus 13 C core up; (3) Methods to improve the performance of instrument electronic devices. For example: improve the low noise amplifier and increase the device performance of ADC.

[0004] However, all of the above methods have some significant drawbacks: 1. For the first type of method, these methods have specific requirements for samples and application conditions, making it difficult to generalize them into universal nuclear magnetic resonance measurement methods; 2. For the second type of method, these methods depend on the molecular structure of the sample, especially the chemical bond relationships between atoms, and are only applicable to specific sample types; 3. For the third type of method, the improvement of electronic devices depends on the development level of electronics and semiconductors, but the development cycle of devices is relatively long, and manufacturers often need to redesign the overall architecture and interface of the instrument, making it difficult to achieve rapid promotion and iteration. Summary of the Invention

[0005] To address the aforementioned deficiencies or improvement needs of existing technologies, this invention provides a method and system for improving the sensitivity of multidimensional spectra in nuclear magnetic resonance (NMR). The aim is to solve the technical problems of existing methods for enhancing nuclear polarization, which have specific requirements for samples and application conditions, making them difficult to generalize into universal NMR measurement methods; existing methods for optimizing pulse sequences or data acquisition, which depend on the molecular structure of the sample, limiting their applicability to specific sample types; and existing methods for improving the performance of instrument electronic components, which have relatively long development cycles and often require manufacturers to redesign the overall instrument architecture and interfaces, hindering rapid promotion and iteration.

[0006] To achieve the above objectives, according to one aspect of the present invention, a method for improving the sensitivity of multidimensional spectra in nuclear magnetic resonance is provided, comprising the following steps:

[0007] (1) Based on the preset sampling parameters, the time domain data s1 of the multidimensional experiment was collected using a nuclear magnetic resonance spectrometer;

[0008] (2) Perform Fourier transform on the direct dimension of the time-domain data s1 collected in step (1), and perform phase correction on the Fourier transform result to obtain the corrected data s2.

[0009] (3) For the data s2 after correction in step (2), a non-uniform reconstruction algorithm is used to reconstruct the data column by column of its indirect dimension to obtain the reconstructed data s3, where the direct dimension of data s3 is the frequency domain and the indirect dimension is the time domain data.

[0010] (4) For the data s3 after data reconstruction obtained in step (3), perform finite impulse response digital filtering on its indirect dimension column by column to obtain the filtered data s4.

[0011] (5) Perform Fourier transform on the indirect dimension of the filtered data s4 in step (4) to obtain the transformed data s5, where both the direct and indirect dimensions of the data s5 are frequency domain data.

[0012] (6) Extract the middle 1 / β part of the indirect dimension of the data s5 after transformation in step (5) to obtain the spectral data s6, which is the multidimensional spectral data after sensitivity enhancement.

[0013] Preferably, the sampling parameters in step (1) are as follows: the sampling mode is non-uniform sampling, the non-uniform sampling rate is 1 / β, and non-uniform sampling index points are generated according to the non-uniform sampling rate. The sampling bandwidth of the indirect dimension is β*sw1, and the number of sampling points of the indirect dimension is β*np1. Wherein, sw1 is the estimated value of the sampling bandwidth required by the indirect dimension under normal sampling, np1 is the estimated value of the number of sampling points of the indirect dimension under normal sampling, and β represents the oversampling factor.

[0014] Preferably, the phase correction process in step (2) is to adjust the zero-order phase and the first-order phase of the Fourier transform result.

[0015] Preferably, the non-uniform reconstruction algorithm in step (3) can be an iterative soft thresholding algorithm (IST), a deep learning algorithm (DL), or a multidimensional decomposition algorithm (MDD), etc.

[0016] Preferably, the interception process in step (6) is as follows:

[0017] range=((0.5-0.5 / β)*np1+1):((0.5+0.5 / β)*np1)

[0018] s6 = s5(:, range)

[0019] Where range represents the range of the indirect dimension of the data s5 after transformation in step (5).

[0020] According to another aspect of the present invention, a system for improving the multidimensional spectral sensitivity in nuclear magnetic resonance is provided, comprising:

[0021] The first module is used to acquire time-domain data s1 of multidimensional experiments based on preset sampling parameters and using a nuclear magnetic resonance spectrometer;

[0022] The second module is used to perform Fourier transform on the direct dimension of the time-domain data s1 acquired by the first module, and to perform phase correction on the Fourier transform result to obtain the corrected data s2.

[0023] The third module is used to reconstruct the indirect dimension column by column for the data s2 after correction by the second module using a non-uniform reconstruction algorithm to obtain the reconstructed data s3, where the direct dimension of data s3 is in the frequency domain and the indirect dimension is in the time domain.

[0024] The fourth module is used to perform finite impulse response digital filtering on the indirect dimension of the reconstructed data s3 obtained from the third module to obtain the filtered data s4.

[0025] The fifth module is used to perform a Fourier transform on the indirect dimension of the data s4 filtered by the fourth module to obtain the transformed data s5, where both the direct and indirect dimensions of the data s5 are frequency domain data.

[0026] The sixth module is used to extract the middle 1 / β portion of the indirect dimension of the data s5 after the transformation by the fifth module to obtain the spectral data s6, which is the multidimensional spectral data after sensitivity enhancement.

[0027] In summary, compared with the prior art, the above-described technical solutions conceived by this invention can achieve the following beneficial effects:

[0028] (1) Since steps (1) and (2) of this invention employ a data acquisition method that combines oversampling with non-uniform sampling, it does not have special requirements for the molecular structure of the sample or the application conditions, and has good universality. Therefore, it can solve the technical problem that the existing first and second type methods have specific requirements for the sample and application conditions, and are difficult to promote into a universal nuclear magnetic resonance measurement method;

[0029] (2) Since the present invention employs steps (4) and (5), it uses a data post-processing algorithm to obtain a spectrum with higher sensitivity, thus solving the technical problem of existing third-class methods relying on the improvement of electronic device performance. Attached Figure Description

[0030] Figure 1 This is a flowchart of the method for improving the sensitivity of multidimensional spectra in nuclear magnetic resonance according to the present invention.

[0031] Figure 2 These are the index points generated by the non-uniform sampling described in this invention.

[0032] Figure 3 The spectrum s5 is obtained by non-uniform data acquisition and reconstruction according to the present invention.

[0033] Figure 4 The spectrum s6 is obtained by extracting the middle 1 / 4 portion of the spectrum s5 in this invention.

[0034] Figure 5 The results are a comparison of the spectra obtained by the conventional acquisition method and the method of this invention, with each spectrum taken from column 1396. Detailed Implementation

[0035] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0036] The basic idea of ​​this invention is to use non-uniform data acquisition to achieve oversampling of the indirect dimension within the same time period, then use a non-uniform reconstruction method to obtain the oversampled spectrum, and finally use digital filtering to downsample the indirect dimension to obtain the final spectrum, thereby improving the sensitivity of the multidimensional spectrum.

[0037] like Figure 1 As shown, the present invention provides a method for improving the sensitivity of multidimensional spectra in nuclear magnetic resonance, comprising the following steps:

[0038] (1) Based on the preset sampling parameters, the time domain data s1 of the multidimensional experiment was collected using a nuclear magnetic resonance spectrometer;

[0039] Specifically, the sampling parameters in this step are as follows: the sampling mode is non-uniform sampling, the non-uniform sampling rate is 1 / β, and non-uniform sampling index points are generated based on this non-uniform sampling rate. The sampling bandwidth of the indirect dimension is β*sw1, and the number of sampling points of the indirect dimension is β*np1. Here, sw1 is the estimated value of the sampling bandwidth required for the indirect dimension under normal sampling conditions, np1 is the estimated value of the number of sampling points of the indirect dimension under normal sampling conditions, and β represents the oversampling factor, which ranges from 2 to 16.

[0040] In this embodiment, sw1 = 164.98ppm, np1 = 512, β = 4, and the generated non-uniform sampling index points are as follows: Figure 2 As shown.

[0041] (2) Perform Fourier transform on the direct dimension of the time-domain data s1 collected in step (1), and perform phase correction on the Fourier transform result to obtain the corrected data s2.

[0042] Specifically, the phase correction process in this step involves adjusting the zeroth-order and first-order phases of the Fourier transform result.

[0043] (3) For the data s2 after correction in step (2), a non-uniform reconstruction algorithm is used to reconstruct the data column by column of its indirect dimension to obtain the reconstructed data s3, where the direct dimension of data s3 is the frequency domain and the indirect dimension is the time domain data.

[0044] Specifically, the non-uniform reconstruction algorithms in this step include, but are not limited to, iterative soft thresholding (IST), deep learning (DL), and multi-dimensional decomposition (MDD).

[0045] (4) For the data s3 after reconstruction obtained in step (3), finite impulse response (FIR) digital filtering is performed column by column on its indirect dimension to obtain the filtered data s4.

[0046] The purpose of this step is to suppress noise outside the indirect dimension bandwidth sw1.

[0047] (5) Perform Fourier transform on the indirect dimension of the filtered data s4 in step (4) to obtain the transformed data s5, where both the direct and indirect dimensions of the data s5 are frequency domain data.

[0048] In this embodiment, the reconstructed data s5 is as follows: Figure 3 As shown.

[0049] (6) Extract the middle 1 / β part of the indirect dimension of the data s5 after transformation in step (5) to obtain the spectral data s6, which is the multidimensional spectral data after sensitivity enhancement.

[0050] Specifically, the extraction process in this step is as follows:

[0051] range=((0.5-0.5 / β)*np1+1):((0.5+0.5 / β)*np1)

[0052] s6 = s5(:, range)

[0053] Where range represents the range of the indirect dimension of the data s5 after transformation in step (5).

[0054] In this embodiment, range = 769:1280, and the reconstructed spectral data s6 is as follows: Figure 4 As shown. The comparison results of the 1396th column of data from s6 with those from the conventional data acquisition method are as follows. Figure 5 As shown, the signal-to-noise ratio (SNR) of the spectrum obtained by the conventional acquisition method is 31.14, while the SNR of the spectrum reconstructed by this method is 70.69. This demonstrates that the sensitivity of the multidimensional spectrum is significantly improved by using this method.

[0055] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for improving the sensitivity of multidimensional spectra in nuclear magnetic resonance, characterized in that, Includes the following steps: (1) Collect time-domain data s1 of the multidimensional experiment using a nuclear magnetic resonance spectrometer according to the preset sampling parameters; The sampling parameters in step (1) are as follows: the sampling mode is non-uniform sampling, the non-uniform sampling rate is 1 / β, and non-uniform sampling index points are generated according to the non-uniform sampling rate. The sampling bandwidth of the indirect dimension is β*sw1, and the number of sampling points of the indirect dimension is β*np1. Wherein, sw1 is the estimated value of the sampling bandwidth required for the indirect dimension under normal sampling, np1 is the estimated value of the number of sampling points of the indirect dimension under normal sampling, and β represents the oversampling factor. (2) Perform Fourier transform on the direct dimension of the time-domain data s1 collected in step (1), and perform phase correction on the Fourier transform result to obtain the corrected data s2. (3) For the data s2 after correction in step (2), a non-uniform reconstruction algorithm is used to reconstruct the data column by column of its indirect dimension to obtain the reconstructed data s3, where the direct dimension of data s3 is the frequency domain and the indirect dimension is the time domain data. (4) For the data s3 after data reconstruction obtained in step (3), perform finite impulse response digital filtering on its indirect dimension column by column to obtain the filtered data s4. (5) Perform Fourier transform on the indirect dimension of the filtered data s4 in step (4) to obtain the transformed data s5, where both the direct and indirect dimensions of the data s5 are frequency domain data. (6) Extract the middle 1 / β part of the indirect dimension of the data s5 after transformation in step (5) to obtain the spectral data s6, which is the multidimensional spectral data after sensitivity enhancement.

2. The method for improving multidimensional spectral sensitivity in nuclear magnetic resonance according to claim 1, characterized in that, The phase correction process in step (2) is to adjust the zero-order phase and the first-order phase of the Fourier transform result.

3. The method for improving multidimensional spectral sensitivity in nuclear magnetic resonance according to claim 2, characterized in that, The non-uniform reconstruction algorithm in step (3) can be the iterative soft thresholding algorithm IST, the deep learning algorithm DL, or the multidimensional decomposition algorithm MDD.

4. The method for improving multidimensional spectral sensitivity in nuclear magnetic resonance according to claim 1, characterized in that, The extraction process in step (6) is as follows: range=((0.5-0.5 / β)*np1+1):((0.5+0.5 / β)*np1) s6 = s5(:, range) Where range represents the range of the indirect dimension of the data s5 after transformation in step (5).

5. A system for improving the sensitivity of multidimensional spectra in nuclear magnetic resonance, characterized in that, include: The first module is used to collect time-domain data s1 of the multidimensional experiment using a nuclear magnetic resonance spectrometer according to preset sampling parameters; the sampling parameters in step (1) are as follows: the sampling mode is non-uniform sampling, the non-uniform sampling rate is 1 / β, and non-uniform sampling index points are generated according to the non-uniform sampling rate, the sampling bandwidth of the indirect dimension is β*sw1, and the number of sampling points of the indirect dimension is β*np1, where sw1 is the estimated value of the sampling bandwidth required for the indirect dimension under normal sampling, np1 is the estimated value of the number of sampling points of the indirect dimension under normal sampling, and β represents the oversampling factor; The second module is used to perform Fourier transform on the direct dimension of the time-domain data s1 acquired by the first module, and to perform phase correction on the Fourier transform result to obtain the corrected data s2. The third module is used to reconstruct the indirect dimension column by column for the data s2 after correction by the second module using a non-uniform reconstruction algorithm to obtain the reconstructed data s3, where the direct dimension of data s3 is in the frequency domain and the indirect dimension is in the time domain. The fourth module is used to perform finite impulse response digital filtering on the indirect dimension of the reconstructed data s3 obtained from the third module to obtain the filtered data s4. The fifth module is used to perform a Fourier transform on the indirect dimension of the data s4 filtered by the fourth module to obtain the transformed data s5, where both the direct and indirect dimensions of the data s5 are frequency domain data. The sixth module is used to extract the middle 1 / β portion of the indirect dimension of the data s5 after the transformation by the fifth module to obtain the spectral data s6, which is the multidimensional spectral data after sensitivity enhancement.

Images