Time-frequency domain joint reconstruction of nuclear magnetic resonance spectrum and method for evaluating reconstruction quality

Abstract

The present disclosure provides a time-frequency domain joint reconstruction and reconstruction quality evaluation method of nuclear magnetic resonance spectrum, comprising the following steps: reconstructing according to undersampling data of simulation data; obtaining the uncertainty and Pearson correlation coefficient of the first reconstructed data; constructing a space rectangular coordinate system to form a point array of scattered points; reconstructing according to undersampling data of nuclear magnetic resonance data; obtaining the uncertainty of the second reconstructed data; using the approximate density to obtain the probability that the Pearson correlation coefficient is a higher value under the condition that the uncertainty is known; and evaluating the second reconstructed data. The present disclosure also provides a nuclear magnetic resonance spectrum reconstruction quality evaluation device, an electronic device and a readable storage medium. Based on the deep learning reconstruction algorithm of the time-frequency domain joint reconstruction, the present disclosure realizes the complementation and interaction of time-frequency domain information in the reconstruction process, completes the high-quality reconstruction of NMR spectrum under flexible sampling rate, sampling scheme and sampling size, and realizes the no-reference quality evaluation of the reconstruction result by constructing the reconstruction quality evaluation point array.

Description

Technical Field

[0001] This disclosure relates to a method, apparatus, electronic device, and storage medium for time-frequency domain joint reconstruction of nuclear magnetic resonance spectra and evaluation of reconstruction quality. Background Technology

[0002] Nuclear Magnetic Resonance (NMR) is a powerful spectroscopic analysis technique widely used in fields such as chemistry. NMR spectra can be divided into one-dimensional and multi-dimensional spectra. Compared to one-dimensional spectra, multi-dimensional spectra distribute information across multiple dimensions, offering higher resolution and better displaying overlapping signals, facilitating more accurate identification and analysis of substances. Multi-dimensional spectra also provide more information, such as the connectivity of atoms in molecules, their spatial orientation, and the interactions between different nuclei. While multi-dimensional spectra offer significant advantages, the time required to acquire them increases exponentially with the increase in indirect dimensions. The sampling times for different dimensions vary considerably. Typically, one-dimensional spectra have shorter sampling times, two-dimensional spectra usually take between minutes and hours, three-dimensional spectra typically take tens of hours, four-dimensional spectra take hundreds of hours, and five-dimensional spectra can even take thousands of hours. These excessively long experimental times limit the application of multi-dimensional spectra.

[0003] Sparse sampling, through undersampling, collects only a subset of data points, significantly reducing sampling time. However, signals obtained through sparse sampling cannot yield high-quality spectra using conventional Fourier transform techniques; sparse sampling reconstruction algorithms are required to reconstruct the spectra. Developing high-performance sparse sampling reconstruction algorithms is crucial for rapidly acquiring high-quality NMR multidimensional spectra and expanding the applications of NMR technology in fields such as chemistry. Currently, many NMR spectrum reconstruction algorithms exist, with traditional algorithms such as SMILE and hmsIST achieving good results in time-domain reconstruction. With the rapid development of deep learning, deep learning-based NMR spectrum reconstruction algorithms have also demonstrated their speed advantage. Deep learning-based algorithms include FID-Net and EDHRN, but these reconstruction algorithms are often based on a single domain (pure time domain or pure frequency domain). In general, while these single-domain-based reconstruction algorithms have some effectiveness, they cannot perform time-frequency domain information interaction during reconstruction, making it difficult to comprehensively consider information from both domains and achieve optimal results. Furthermore, full-sample data is unavailable, and the data in actual tests is often non-uniformly sampled, making it impossible to determine the reliability of the reconstruction results. Therefore, it is essential to propose a reference-free evaluation method for NMR spectrum reconstruction quality based on uncertainty. Summary of the Invention

[0004] To address at least one of the aforementioned technical problems, this disclosure provides a method, apparatus, electronic device, and storage medium for evaluating the quality of nuclear magnetic resonance spectrum reconstruction.

[0005] According to one aspect of this disclosure, a method for evaluating the quality of nuclear magnetic resonance spectrum reconstruction is provided, comprising the following steps:

[0006] Non-uniform sampling is performed on the simulated data to obtain undersampled data of the simulated data;

[0007] Reconstruction is performed based on the undersampled data of the simulated data to obtain the first reconstructed data;

[0008] Based on the first reconstructed data, the cognitive uncertainty of the first reconstructed data, the accidental uncertainty of the first reconstructed data, and the Pearson correlation coefficient of the peak intensity of the tag spectrum of the first reconstructed data and the simulated data are obtained.

[0009] A spatial rectangular coordinate system is constructed, with the cognitive uncertainty as the first direction, the accidental uncertainty as the second direction, and the Pearson correlation coefficient as the third direction, to obtain the scatter points corresponding to the simulated data in the spatial rectangular coordinate system; based on multiple sets of the simulated data, a point matrix of the scatter points is formed in the spatial rectangular coordinate system.

[0010] The sample was non-uniformly sampled using a nuclear magnetic resonance spectrometer to obtain undersampled nuclear magnetic resonance data.

[0011] Reconstruction is performed based on the undersampled data of the nuclear magnetic resonance data to obtain the second reconstructed data;

[0012] Based on the second reconstructed data, we obtain the cognitive uncertainty of the second reconstructed data and the accidental uncertainty of the second reconstructed data;

[0013] Based on the cognitive uncertainty and the accidental uncertainty of the second reconstructed data, a straight line parallel to the third direction is determined on the spatial rectangular coordinate system. The straight line is extended to the first direction and the second direction respectively to form a cuboid region, such that the number of scattered points contained in the cuboid region reaches Q. Based on the threshold of the third direction, the cuboid region is divided into planes parallel to the first direction and the second direction. The proportion of the number of scattered points contained in the region above the threshold obtained by the division is calculated in the total number of scattered points contained in the cuboid region, where Q is a positive integer.

[0014] The second reconstructed data is evaluated based on the stated proportions.

[0015] According to at least one embodiment of the present disclosure, the nuclear magnetic resonance spectrum reconstruction quality evaluation method includes N reconstructions, each yielding reconstruction data from the first to the Nth reconstruction.

[0016] The cognitive uncertainty is the variance of the output values ​​of the reconstructed data from the first to the Nth time. The output values ​​are obtained by taking the mean of the matrix of the reconstructed data. The random uncertainty is the mean of the random uncertainty of the reconstructed data from the first to the Nth time. The Pearson correlation coefficient is the Pearson correlation coefficient between the mean of the reconstructed data from the first to the Nth time and the peak intensity of the tag spectrum corresponding to the reconstructed data from the first to the Nth time.

[0017] According to at least one embodiment of the nuclear magnetic resonance spectrum reconstruction quality evaluation method of the present disclosure, the threshold is 0.97 and Q is 100.

[0018] According to at least one embodiment of the nuclear magnetic resonance spectrum reconstruction quality evaluation method of the present disclosure, the step size of the straight line extending in the first direction is determined based on the magnitude of the cognitive uncertainty of the second reconstructed data; the step size of the straight line extending in the second direction is determined based on the magnitude of the random uncertainty of the second reconstructed data.

[0019] According to the nuclear magnetic resonance spectrum reconstruction quality evaluation method of at least one embodiment of the present disclosure, the sampling rate for non-uniform sampling of the simulated data is [12.5%, 25%], and the sampling scheme for non-uniform sampling of the simulated data follows a Poisson distribution or a random distribution.

[0020] A method for evaluating the quality of nuclear magnetic resonance spectrum reconstruction according to at least one embodiment of the present disclosure, wherein the reconstruction includes the following steps:

[0021] Time-domain reconstruction is performed on the time-domain data in the undersampled data to obtain the time-domain reconstructed data;

[0022] The time-domain reconstructed data is subjected to Fourier transform to obtain frequency-domain data;

[0023] The frequency domain data is reconstructed in the frequency domain to obtain the reconstructed data.

[0024] According to at least one embodiment of the nuclear magnetic resonance spectrum reconstruction quality evaluation method of the present disclosure, both the time-domain reconstruction and the frequency-domain reconstruction include data consistency operations.

[0025] The nuclear magnetic resonance spectrum reconstruction quality evaluation method according to at least one embodiment of the present disclosure uses a time-frequency domain joint Bayesian deep learning reconstruction network.

[0026] According to at least one embodiment of the nuclear magnetic resonance spectrum reconstruction quality evaluation method of the present disclosure, the Bayesian deep learning reconstruction network includes a dropout layer.

[0027] According to one aspect of this disclosure, a device for evaluating the quality of nuclear magnetic resonance spectrum reconstruction is provided, comprising:

[0028] A sampling module is used to perform non-uniform sampling on the simulation data to obtain undersampled data of the simulation data;

[0029] Additionally, non-uniform sampling of the sample was performed using a nuclear magnetic resonance spectrometer to obtain undersampled nuclear magnetic resonance data;

[0030] The reconstruction module is used to reconstruct the simulated data based on the undersampled data to obtain the first reconstructed data; based on the first reconstructed data, the cognitive uncertainty of the first reconstructed data, the accidental uncertainty of the first reconstructed data, and the Pearson correlation coefficient of the peak intensity of the tag spectrum of the first reconstructed data and the simulated data are obtained.

[0031] Furthermore, based on the undersampled data of the nuclear magnetic resonance data, a second reconstructed data is obtained; based on the second reconstructed data, the cognitive uncertainty of the second reconstructed data and the accidental uncertainty of the second reconstructed data are obtained;

[0032] A construction module is used to construct a spatial rectangular coordinate system, with the cognitive uncertainty as the first direction, the accidental uncertainty as the second direction, and the Pearson correlation coefficient as the third direction, to obtain the scatter points corresponding to the simulated data in the spatial rectangular coordinate system; and to form a point matrix of the scatter points in the spatial rectangular coordinate system based on multiple sets of the simulated data.

[0033] The calculation module is used to determine a straight line parallel to the third direction on the spatial rectangular coordinate system based on the cognitive uncertainty and the accidental uncertainty of the second reconstructed data, extend the straight line to the first direction and the second direction respectively to form a cuboid region, so that the number of scattered points contained in the cuboid region reaches Q, divide the cuboid region into planes parallel to the first direction and the second direction according to the threshold of the third direction, and calculate the proportion of the number of scattered points contained in the region above the threshold in the total number of scattered points contained in the cuboid region, where Q is a positive integer;

[0034] An evaluation module is used to evaluate the second reconstructed data based on the stated proportion.

[0035] According to one aspect of this disclosure, an electronic device is provided, comprising:

[0036] Memory, the memory storing execution instructions; and

[0037] The processor executes the execution instructions stored in the memory, causing the processor to perform the above-described nuclear magnetic resonance spectrum reconstruction quality evaluation method.

[0038] According to one aspect of this disclosure, a readable storage medium is provided that stores executable instructions, which, when executed by a processor, are used to implement the above-described nuclear magnetic resonance spectrum reconstruction quality evaluation method. Attached Figure Description

[0039] The accompanying drawings illustrate exemplary embodiments of the present disclosure and, together with the description thereof, serve to explain the principles of the present disclosure. These drawings are included to provide a further understanding of the present disclosure and are incorporated in and constitute a part of this specification.

[0040] Figure 1 This is a flowchart illustrating the steps of an embodiment of the nuclear magnetic resonance spectrum reconstruction quality assessment method disclosed herein.

[0041] Figure 2 This is a structural diagram of a Bayesian deep learning reconstruction network based on joint time-frequency domain reconstruction, which is an embodiment of the nuclear magnetic resonance spectrum reconstruction quality evaluation method disclosed herein.

[0042] Figure 3 This is a flowchart illustrating the steps of constructing a simulation lattice and evaluating the reconstruction results in a reference-free evaluation method according to an embodiment of the nuclear magnetic resonance spectrum reconstruction quality evaluation method disclosed herein.

[0043] Figure 4 This is a schematic diagram of a simulation lattice representing an embodiment of the nuclear magnetic resonance spectrum reconstruction quality assessment method disclosed herein;

[0044] Figure 5 The reconstruction result and quality evaluation diagram of GB1 protein are shown in one embodiment of the nuclear magnetic resonance spectrum reconstruction quality evaluation method of this disclosure.

[0045] Figure 6 The reconstruction result and quality evaluation diagram of T4L L99A protein are shown in one embodiment of the nuclear magnetic resonance spectrum reconstruction quality evaluation method of this disclosure.

[0046] Figure 7 This is a schematic block diagram of a nuclear magnetic resonance spectrum reconstruction quality evaluation device that employs a hardware implementation of a processing system, as one embodiment of the present disclosure. Detailed Implementation

[0047] The present disclosure will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the disclosure. Furthermore, it should be noted that, for ease of description, only the parts relevant to the present disclosure are shown in the accompanying drawings.

[0048] It should be noted that, where there is no conflict, the embodiments and features described in this disclosure can be combined with each other. The technical solutions of this disclosure will now be described in detail with reference to the accompanying drawings and embodiments.

[0049] Unless otherwise stated, the exemplary implementations / embodiments shown are to be understood as providing exemplary features of various details that provide ways in which the technical concepts of this disclosure can be implemented in practice. Therefore, unless otherwise stated, the features of various implementations / embodiments may be additionally combined, separated, interchanged and / or rearranged without departing from the technical concepts of this disclosure.

[0050] The use of crosshairs and / or shading in the accompanying drawings is generally used to clarify the boundaries between adjacent components. Thus, unless otherwise stated, the presence or absence of crosshairs or shading does not convey or indicate any preference or requirement for the specific material, material properties, dimensions, proportions, commonalities between the illustrated components, or any other characteristics, properties, etc., of the components. Furthermore, in the accompanying drawings, the dimensions and relative dimensions of components may be exaggerated for clarity and / or descriptive purposes. When exemplary embodiments can be implemented differently, a specific process sequence may be performed in a different order than that described. For example, two consecutively described processes may be performed substantially simultaneously or in the reverse order of their description. Furthermore, the same reference numerals denote the same components.

[0051] When a component is referred to as being "on" or "above" another component, "connected to," or "joined to" another component, the component may be directly on, directly connected to, or directly joined to the other component, or there may be intermediate components. However, when a component is referred to as being "directly on" another component, "directly connected to," or "directly joined to" another component, there are no intermediate components. Therefore, the term "connection" can refer to a physical connection, an electrical connection, etc., and may or may not have intermediate components.

[0052] The terminology used herein is for the purpose of describing particular embodiments and is not intended to be limiting. As used herein, unless the context clearly indicates otherwise, the singular forms “a” and “the” are intended to include the plural forms as well. Furthermore, when the terms “comprising” and / or “including” and variations thereof are used in this specification, it indicates the presence of the stated features, integrals, steps, operations, parts, components, and / or groups thereof, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, parts, components, and / or groups thereof. It should also be noted that, as used herein, the terms “substantially,” “about,” and other similar terms are used as approximate terms rather than as terms of degree, thus explaining the inherent biases in measurements, calculated values, and / or provided values ​​that would be recognized by one of ordinary skill in the art.

[0053] The following text combines Figures 1 to 5 The method for evaluating the quality of nuclear magnetic resonance spectrum reconstruction disclosed herein is described in detail.

[0054] Example 1

[0055] refer to Figure 1 The evaluation method includes the following steps:

[0056] S100. Perform non-uniform sampling on the simulation data to obtain undersampled data of the simulation data;

[0057] S200: Reconstruct the data based on the undersampled data of the simulation data to obtain the first reconstructed data;

[0058] S300. Based on the first reconstructed data, obtain the cognitive uncertainty of the first reconstructed data, the accidental uncertainty of the first reconstructed data, and the Pearson correlation coefficient of the peak intensity of the tag spectrum of the first reconstructed data and the simulated data.

[0059] S400. Construct a spatial rectangular coordinate system, taking the cognitive uncertainty as the first direction, the accidental uncertainty as the second direction, and the Pearson correlation coefficient as the third direction, to obtain the scatter points corresponding to the simulated data on the spatial rectangular coordinate system; and form a point matrix of the scatter points on the spatial rectangular coordinate system based on multiple sets of the simulated data.

[0060] S500: Use a nuclear magnetic resonance spectrometer to perform non-uniform sampling on the sample to obtain undersampled nuclear magnetic resonance data.

[0061] S600: Reconstruct the data based on the undersampled data of the nuclear magnetic resonance data to obtain the second reconstructed data;

[0062] S700. Based on the second reconstructed data, obtain the cognitive uncertainty of the second reconstructed data and the accidental uncertainty of the second reconstructed data;

[0063] S800. Based on the cognitive uncertainty and the accidental uncertainty of the second reconstructed data, a straight line parallel to the third direction is determined on the spatial rectangular coordinate system. The straight line is extended to the first direction and the second direction respectively to form a cuboid region, such that the number of scattered points contained in the cuboid region reaches Q. Based on the threshold of the third direction, the cuboid region is divided into planes parallel to the first direction and the second direction. The proportion of the number of scattered points contained in the region above the threshold obtained by the division is calculated in the total number of scattered points contained in the cuboid region, where Q is a positive integer.

[0064] S900. Evaluate the second reconstructed data based on the stated proportion.

[0065] I. A deep learning reconstruction algorithm based on joint time-frequency domain reconstruction, realizing the interaction of time-frequency domain information and completing high-quality spectrum reconstruction.

[0066] 1. Deep learning network for joint reconstruction in the time and frequency domains: The network structure is as follows Figure 2 As shown, the network consists of two parts: time-domain reconstruction and frequency-domain reconstruction, which are connected by Fourier transform.

[0067] The temporal reconstruction section consists of a series of convolutional layers, used for preliminary reconstruction of undersampled temporal data. The intermediate results of the temporal reconstruction are then Fourier transformed to the frequency domain and input into the frequency domain reconstruction module.

[0068] The frequency domain reconstruction part consists of a series of upsampling and downsampling operations, combined with residual connections in the middle, to perform frequency domain reconstruction.

[0069] Both the time-domain reconstruction and frequency-domain reconstruction parts underwent data consistency operations during training to improve the network's universality for different sampling templates.

[0070] The intermediate results of time-domain reconstruction and the final results of frequency-domain reconstruction form a loss function with the mean square error (MSE) of the label spectrum calculation; combined with uncertainty modeling, the network can simultaneously output the reconstruction results, the accidental uncertainty of the reconstruction results, and the cognitive uncertainty.

[0071] 2. Simulated Dataset: A large amount of simulated data is generated using the theoretical formulas of NMR signals. Specifically, MATLAB is used to simulate the free induction decay (FID) signal.

[0072] The simulated data has a label spectrum.

[0073] The simulated data is used to construct a simulated dataset.

[0074] In this embodiment, a first simulated dataset for training the model, a second simulated dataset for verifying the model's performance, and a third simulated dataset for constructing a simulated dot matrix were established.

[0075] In this embodiment, the undersampled data of the simulation data in the first and second simulation datasets is 1-dimensional (D) data, which is a time-domain signal.

[0076] The same model is used to reconstruct the simulated data in all simulated datasets, eliminating the need to train multiple models.

[0077] Non-uniform sampling is performed on the simulated data in each simulated dataset to obtain undersampled data. The sampling template is randomly generated, i.e., the sampling rate and sampling scheme are random.

[0078] In this embodiment, the sampling rate for undersampling the simulated data is [12.5%, 25%], and the sampling scheme follows a Poisson distribution or a random distribution.

[0079] 3. Model Training: In this embodiment, a Bayesian deep learning network model combining time and frequency domains is used, trained using undersampled data from the first simulated dataset. The real and imaginary parts of the time-domain signal are placed in separate channels during input. Randomly generated simulated data with different sampling rates and schemes are used during training to ensure the trained model is applicable to NMR spectra with different sampling sizes, schemes, and rates.

[0080] The loss function consists of three parts: random uncertainty loss, time-domain reconstruction loss, and frequency-domain reconstruction loss. Random uncertainty, to some extent, measures the difficulty of predicting the data to be tested, and can be modeled using the following loss function:

[0081]

[0082] Where σ(x) k (x) represents the data to be tested. k Cognitive uncertainty, f θ (x k ) represents the output value when the network parameters are θ. This output value is obtained by taking the mean of the reconstructed data matrix. K is the batch size during training. k This is the ground truth. The random uncertainty is unrelated to the network itself, but depends only on the data being tested.

[0083] Cognitive uncertainty is represented by the variance of the output values ​​as follows:

[0084]

[0085] in, The mean of the output values ​​predicted by the network in P iterations is the reconstruction result.

[0086] Add convolutional layers to a deep learning model, using L... ale The random uncertainty is obtained by performing unsupervised learning as a loss function during training. The mean of the random uncertainties from multiple reconstructions is taken as the final random uncertainty.

[0087] The temporal loss function is calculated separately for the reconstruction results of each encoding / decoding module in the temporal reconstruction section, and the final temporal loss function is as follows:

[0088]

[0089] Where β is the weight of the temporal loss function, and L is the number of temporal reconstruction modules. For time-domain label values, Ψ l (x k ) represents the output value of the l-th time-domain reconstruction module.

[0090] The frequency domain loss function is calculated separately for the reconstruction results of each encoding / decoding module in the frequency domain reconstruction section, and the final frequency domain loss function is as follows:

[0091]

[0092] Where M represents the number of encoding / decoding / reconstruction modules. The frequency domain label value, i.e., the ground truth, Φ m (x k ) represents the m-th frequency domain reconstruction module Φ m (x k The output value of ).

[0093] The loss function used for training is shown below:

[0094] L(θ)=L t +L f +L ale

[0095] The optimizer was Adam, the learning rate was set to 0.001, and the model was trained for 50 epochs. The best model on the second simulated dataset was selected as the final reconstruction model.

[0096] 4. Simulation lattice construction without a reference evaluation method:

[0097] refer to Figure 3In this embodiment, the third simulated dataset is generated using theoretical formulas for 2D NMR signals, and then each 2D spectrum is reconstructed row by row using a trained 1D model.

[0098] The specific number of simulated data points in the third simulation dataset is denoted by S, and S is on the order of 10. 4 The above ensures that the construction of the simulated dot matrix can include the features of most of the data. In this embodiment, S = 10000.

[0099] The trained model is used to reconstruct the undersampled data of the simulated data in the third simulated dataset N times. In this embodiment, N=10.

[0100] During reconstruction, dropout is enabled, and a certain number of neurons are randomly discarded in each prediction, thereby achieving the purpose of sampling the model weight parameters multiple times.

[0101] The average of the 10 predicted output values ​​is the reconstruction result, and cognitive uncertainty is calculated based on the reconstruction result. The model outputs the random uncertainty for each prediction along with the reconstruction result; the average of these values ​​is the final random uncertainty. Since the simulated data has a label spectrum, the Pearson correlation coefficient (PCC) between the peak intensity of the reconstruction result and the label spectrum is calculated using the label spectrum plot of the simulated data in the third simulated dataset.

[0102] A spatial rectangular coordinate system is constructed, with PCC value, cognitive uncertainty, and random uncertainty corresponding to the Z, X, and Y axes, respectively. Each simulated data point corresponds to a scatter plot on this system. The S simulated data points form a lattice on the spatial rectangular coordinate system, representing the simulated lattice without a reference evaluation method. The scatter plot is shown below. Figure 4 As shown.

[0103] The data range of the Z-axis is limited by the PCC value, which ranges from [0,1].

[0104] 5. Reference-Free Evaluation Method for Reconstruction Quality: A new undersampled spectrum is processed. This undersampled spectrum is obtained by non-uniformly sampling the sample using an NMR spectrometer, yielding true NMR data. The same model is used to reconstruct the undersampled NMR data N times. This yields a new reconstructed spectrum, along with its cognitive and accidental uncertainties. These uncertainties are used to establish a point in the XOY plane of a Cartesian coordinate system. This point is extended in the ±Z direction to form a straight line l. This line is then extended in the ±X and ±Y directions to form a cuboid region, containing Q scattered points. In this embodiment, Q = 100.

[0105] The expansion step sizes in the ±X and ±Y directions are determined based on the magnitudes of cognitive uncertainty and random uncertainty, respectively. In this embodiment, the magnitude of cognitive uncertainty is 10. -7 Take the expansion step size as 10 -10 The magnitude of the random uncertainty is 10. -4 Take the expansion step size as 10 -7 .

[0106] Since the reconstruction results are excellent when the PCC value is in the range [0.97, 1], this range is selected as the evaluation range, i.e., the threshold is 0.97.

[0107] Based on Z∈[0.97, 1], a cuboid region is divided parallel to the XOY plane. The proportion of the number of scattered points in the divided region to the total number of scattered points contained in the cuboid region is calculated as the approximate density of PCC in this interval under this uncertainty, and is expressed as a percentage.

[0108] To make the final evaluation more intuitive, the approximation density is converted into a percentage. Based on the approximation density, the evaluation results are divided into three levels: credible, reasonably credible, and unreliable. When the approximation density of the reconstruction result is greater than 70%, it is considered credible; when the approximation density is between 50% and 70%, it is considered reasonably credible; and when it is less than 50%, it is considered unreliable.

[0109] The innovative points of this disclosure

[0110] 1. This disclosure constructs a deep learning reconstruction network for joint reconstruction of NMR spectra in the time and frequency domains. During the reconstruction process, it combines information from the time and frequency domains and their complementary features to achieve high-quality spectrum reconstruction.

[0111] 2. This disclosure is the first in the NMR field to propose a no-reference quality assessment of reconstruction results. It uses uncertainty index to build a lattice for no-reference quality assessment based on simulation data, and uses this lattice in combination with uncertainty to assess the quality of reconstruction results.

[0112] Compared with existing technologies, the present invention can achieve better results in non-uniform sampling spectrum reconstruction and realize reference-free quality evaluation of reconstruction quality.

[0113] Example 2

[0114] 2D of GB1 protein was selected. 15 N- 1 The H HSQC spectrum was reconstructed and evaluated using the evaluation method described in Example 1 after non-uniform sampling with randomly generated sampling templates following a random distribution and sampling rates of 15%, 12.5%, and 8%, respectively.

[0115] refer to Figure 5 Figures (a), (c), and (e) show the reconstruction results and their evaluations under a Poisson distribution with sampling rates of 15%, 12.5%, and 8%, respectively; figures (b), (d), and (f) show the reconstruction results and their evaluations under a random distribution with sampling rates of 15%, 12.5%, and 8%, respectively. The actual PCC values ​​of the data in the figures are (a) 0.997, (b) 0.992, (c) 0.991, (d) 0.991, (e) 0.950, and (f) 0.958, demonstrating the high reconstruction quality and accurate no-reference quality evaluation of this disclosure.

[0116] Example 3

[0117] 2D of T4L L99A protein was selected. 15 N- 1 The H HSQC spectrum was reconstructed and evaluated using the evaluation method described in Example 1 after non-uniform sampling with randomly generated sampling templates following a random distribution and sampling rates of 15%, 12.5%, and 8%, respectively.

[0118] refer to Figure 6 Figures (a), (c), and (e) show the reconstruction results and their evaluations under a Poisson distribution with sampling rates of 15%, 12.5%, and 8%, respectively; figures (b), (d), and (f) show the reconstruction results and their evaluations under a random distribution with sampling rates of 15%, 12.5%, and 8%, respectively. The actual PCC values ​​of the data in the figures are (a) 0.996, (b) 0.993, (c) 0.993, (d) 0.963, (e) 0.924, and (f) 0.943, demonstrating the high reconstruction quality and accurate no-reference quality evaluation of this disclosure.

[0119] Experiments on GB1 and T4L L99A proteins demonstrate that this disclosure can achieve high-quality spectral reconstruction using the same model with flexible sampling templates, sampling rates, and sizes, and accurately evaluate the quality of the reconstructed spectra.

[0120] Example 4

[0121] Figure 7 This is a schematic block diagram of a nuclear magnetic resonance spectrum reconstruction quality evaluation device that employs a hardware implementation of a processing system, as one embodiment of the present disclosure.

[0122] The evaluation device 1000 includes:

[0123] Sampling module 1002, a sampling module, is used to perform non-uniform sampling on the simulation data to obtain undersampled data of the simulation data;

[0124] Additionally, non-uniform sampling of the sample was performed using a nuclear magnetic resonance spectrometer to obtain undersampled nuclear magnetic resonance data;

[0125] The reconstruction module 1004 is used to reconstruct the simulated data based on the undersampled data to obtain the first reconstructed data; and based on the first reconstructed data, to obtain the cognitive uncertainty of the first reconstructed data, the accidental uncertainty of the first reconstructed data, and the Pearson correlation coefficient of the peak intensity of the tag spectrum of the first reconstructed data and the simulated data.

[0126] Furthermore, based on the undersampled data of the nuclear magnetic resonance data, a second reconstructed data is obtained; based on the second reconstructed data, the cognitive uncertainty of the second reconstructed data and the accidental uncertainty of the second reconstructed data are obtained;

[0127] The construction module 1006 is used to construct a spatial rectangular coordinate system, with the cognitive uncertainty as the first direction, the accidental uncertainty as the second direction, and the Pearson correlation coefficient as the third direction, to obtain the scatter points corresponding to the simulated data in the spatial rectangular coordinate system; and to form a point matrix of the scatter points in the spatial rectangular coordinate system based on multiple sets of the simulated data.

[0128] The calculation module 1008 is used to determine a straight line parallel to the third direction on the spatial rectangular coordinate system based on the cognitive uncertainty and the accidental uncertainty of the second reconstructed data, extend the straight line to the first direction and the second direction respectively to form a cuboid region, so that the number of scattered points contained in the cuboid region reaches Q, divide the cuboid region into planes parallel to the first direction and the second direction according to the threshold of the third direction, and calculate the proportion of the number of scattered points contained in the region above the threshold in the total number of scattered points contained in the cuboid region, where Q is a positive integer;

[0129] Evaluation module 1010 is used to evaluate the second reconstructed data based on the stated proportion.

[0130] The apparatus may include corresponding modules that perform one or more steps in the flowchart above. Therefore, each or more steps in the flowchart above can be performed by a corresponding module, and the apparatus may include one or more of these modules. A module may be one or more hardware modules specifically configured to perform a corresponding step, or implemented by a processor configured to perform a corresponding step, or stored in a computer-readable medium for implementation by a processor, or implemented through some combination thereof.

[0131] This hardware architecture can be implemented using a bus architecture. The bus architecture can include any number of interconnect buses and bridges, depending on the specific application and overall design constraints of the hardware. Bus 1100 connects various circuits, including one or more processors 1200, memory 1300, and / or hardware modules. Bus 1100 can also connect various other circuits 1400, such as peripherals, voltage regulators, power management circuits, external antennas, etc.

[0132] Bus 1100 can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Component (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of representation, only one connection line is used in this diagram, but this does not imply that there is only one bus or only one type of bus.

[0133] Any process or method description in the flowcharts or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing a particular logical function or process, and the scope of the preferred embodiments of this disclosure includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as will be understood by those skilled in the art to which embodiments of this disclosure pertain. The processor performs the various methods and processes described above. For example, the method embodiments of this disclosure may be implemented as software programs tangibly contained in a machine-readable medium, such as memory. In some embodiments, part or all of the software program may be loaded and / or installed via memory and / or a communication interface. When the software program is loaded into memory and executed by the processor, one or more steps of the methods described above may be performed. Alternatively, in other embodiments, the processor may be configured to perform one of the methods described above by any other suitable means (e.g., by means of firmware).

[0134] The logic and / or steps represented in the flowchart or otherwise described herein may be specifically implemented in any readable storage medium for use by, or in conjunction with, an instruction execution system, apparatus or device (such as a computer-based system, a processor-included system or other system that can fetch and execute instructions from, an instruction execution system, apparatus or device).

[0135] For the purposes of this specification, a "readable storage medium" can be any means capable of containing, storing, communicating, propagating, or transmitting a program for use by or in conjunction with an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of readable storage media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and programmable read-only memory (EPROM or flash memory), fiber optic devices, and portable read-only memory (CDROM). Furthermore, a readable storage medium can even be paper or other suitable media on which a program can be printed, since a program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in memory.

[0136] It should be understood that various parts of this disclosure can be implemented in hardware, software, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0137] Those skilled in the art will understand that all or part of the steps of the methods described above can be implemented by a program instructing related hardware. The program can be stored in a readable storage medium, and when executed, the program includes one or a combination of the steps of the method implementation.

[0138] Furthermore, the functional units in the various embodiments of this disclosure can be integrated into a single processing module, or each unit can exist physically separately, or two or more units can be integrated into a single module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a readable storage medium. The storage medium can be a read-only memory, a disk, or an optical disk, etc.

[0139] This disclosure also provides an electronic device, including: a memory storing execution instructions; and a processor or other hardware module executing the execution instructions stored in the memory, causing the processor or other hardware module to execute the above-described nuclear magnetic resonance spectrum reconstruction quality evaluation method.

[0140] This disclosure also provides a readable storage medium storing executable instructions, which, when executed by a processor, are used to implement the above-described nuclear magnetic resonance spectrum reconstruction quality evaluation method.

[0141] In the description of this specification, the references to terms such as "one embodiment / mode," "some embodiments / modes," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment / mode or example is included in at least one embodiment / mode or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment / mode or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments / modes or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments / modes or examples described in this specification, as well as the features of different embodiments / modes or examples.

[0142] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0143] Those skilled in the art should understand that the above embodiments are merely for illustrating the present disclosure and are not intended to limit the scope of the disclosure. Those skilled in the art can make other changes or modifications based on the above disclosure, and these changes or modifications still fall within the scope of the present disclosure.

Claims

1. A method for evaluating the quality of nuclear magnetic resonance spectrum reconstruction, characterized in that, Includes the following steps: Non-uniform sampling is performed on the simulated data to obtain undersampled data of the simulated data; Reconstruction is performed based on the undersampled data of the simulated data to obtain the first reconstructed data; Based on the first reconstructed data, the cognitive uncertainty of the first reconstructed data, the accidental uncertainty of the first reconstructed data, and the Pearson correlation coefficient of the peak intensity of the tag spectrum of the first reconstructed data and the simulated data are obtained. A spatial rectangular coordinate system is constructed, with the cognitive uncertainty as the first direction, the accidental uncertainty as the second direction, and the Pearson correlation coefficient as the third direction, to obtain the scatter points corresponding to the simulated data in the spatial rectangular coordinate system; based on multiple sets of the simulated data, a point matrix of the scatter points is formed in the spatial rectangular coordinate system. The sample was non-uniformly sampled using a nuclear magnetic resonance spectrometer to obtain undersampled nuclear magnetic resonance data. Reconstruction is performed based on the undersampled data of the nuclear magnetic resonance data to obtain the second reconstructed data; Based on the second reconstructed data, we obtain the cognitive uncertainty of the second reconstructed data and the accidental uncertainty of the second reconstructed data; Based on the cognitive uncertainty and the accidental uncertainty of the second reconstructed data, a straight line parallel to the third direction is determined on the spatial rectangular coordinate system. The straight line is extended to the first direction and the second direction respectively to form a cuboid region, such that the number of scattered points contained in the cuboid region reaches Q. Based on the threshold of the third direction, the cuboid region is divided into planes parallel to the first direction and the second direction. The proportion of the number of scattered points contained in the region above the threshold obtained by the division is calculated in the total number of scattered points contained in the cuboid region, where Q is a positive integer. The second reconstructed data is evaluated based on the stated proportions.

2. The method for evaluating the quality of nuclear magnetic resonance spectrum reconstruction as described in claim 1, characterized in that, The reconstruction includes N iterations, obtaining reconstruction data from the 1st to the Nth iterations respectively; The cognitive uncertainty is the variance of the output values ​​of the reconstructed data from the first to the Nth time. The output values ​​are obtained by taking the mean of the matrix of the reconstructed data. The random uncertainty is the mean of the random uncertainty of the reconstructed data from the first to the Nth time. The Pearson correlation coefficient is the Pearson correlation coefficient between the mean of the reconstructed data from the first to the Nth time and the peak intensity of the tag spectrum corresponding to the reconstructed data from the first to the Nth time.

3. The method for evaluating the quality of nuclear magnetic resonance spectrum reconstruction as described in claim 1, characterized in that, The step size for extending the line in the first direction is determined based on the magnitude of the cognitive uncertainty of the second reconstructed data; the step size for extending the line in the second direction is determined based on the magnitude of the random uncertainty of the second reconstructed data.

4. The method for evaluating the quality of nuclear magnetic resonance spectrum reconstruction as described in claim 1, characterized in that, The sampling rate for non-uniform sampling of the simulated data is [12.5%, 25%], and the sampling scheme for non-uniform sampling of the simulated data follows a Poisson distribution or a random distribution.

5. The method for evaluating the quality of nuclear magnetic resonance spectrum reconstruction as described in claim 1, characterized in that, The reconstruction includes the following steps: Time-domain reconstruction is performed on the time-domain data in the undersampled data to obtain the time-domain reconstructed data; The time-domain reconstructed data is subjected to Fourier transform to obtain frequency-domain data; The frequency domain data is reconstructed in the frequency domain to obtain the reconstructed data.

6. The method for evaluating the quality of nuclear magnetic resonance spectrum reconstruction as described in claim 5, characterized in that, Both the time-domain reconstruction and the frequency-domain reconstruction include data consistency operations.

7. The method for evaluating the quality of nuclear magnetic resonance spectrum reconstruction as described in claim 5, characterized in that, Reconstructing the network using Bayesian deep learning that combines time and frequency domains.

8. A device for evaluating the quality of nuclear magnetic resonance spectrum reconstruction, characterized in that, include: A sampling module is used to perform non-uniform sampling on the simulation data to obtain undersampled data of the simulation data; Additionally, non-uniform sampling of the sample was performed using a nuclear magnetic resonance spectrometer to obtain undersampled nuclear magnetic resonance data; The reconstruction module is used to reconstruct the simulated data based on the undersampled data to obtain the first reconstructed data; based on the first reconstructed data, the cognitive uncertainty of the first reconstructed data, the accidental uncertainty of the first reconstructed data, and the Pearson correlation coefficient of the peak intensity of the tag spectrum of the first reconstructed data and the simulated data are obtained. Furthermore, based on the undersampled data of the nuclear magnetic resonance data, a second reconstructed data is obtained; based on the second reconstructed data, the cognitive uncertainty of the second reconstructed data and the accidental uncertainty of the second reconstructed data are obtained; A construction module is used to construct a spatial rectangular coordinate system, with the cognitive uncertainty as the first direction, the accidental uncertainty as the second direction, and the Pearson correlation coefficient as the third direction, to obtain the scatter points corresponding to the simulated data in the spatial rectangular coordinate system; and to form a point matrix of the scatter points in the spatial rectangular coordinate system based on multiple sets of the simulated data. The calculation module is used to determine a straight line parallel to the third direction on the spatial rectangular coordinate system based on the cognitive uncertainty and the accidental uncertainty of the second reconstructed data, extend the straight line to the first direction and the second direction respectively to form a cuboid region, so that the number of scattered points contained in the cuboid region reaches Q, divide the cuboid region into planes parallel to the first direction and the second direction according to the threshold of the third direction, and calculate the proportion of the number of scattered points contained in the region above the threshold in the total number of scattered points contained in the cuboid region, where Q is a positive integer; An evaluation module is used to evaluate the second reconstructed data based on the stated proportion.

9. An electronic device, characterized in that, include: The memory stores execution instructions; as well as A processor that executes the execution instructions stored in the memory, causing the processor to perform the nuclear magnetic resonance spectrum reconstruction quality evaluation method according to any one of claims 1 to 7.

10. A readable storage medium, characterized in that, The readable storage medium stores execution instructions, which, when executed by a processor, are used to implement the nuclear magnetic resonance spectrum reconstruction quality evaluation method according to any one of claims 1 to 7.

Images