An intelligent suppression method and system for random noise in magnetic resonance based on dynamic weighting

Abstract

This application belongs to the technical field of nuclear magnetic resonance sounding, and is a method and system for intelligently suppressing magnetic resonance random noise based on dynamic weighting, including: generating a pure signal in the pseudo-X domain from the noisy signal in the Y domain through a forward conversion path, or generating a pure signal in the reconstructed X domain from the noisy signal in the pseudo-Y domain; generating a reconstructed noisy signal in the Y domain from the pure signal in the pseudo-X domain through a reverse conversion path, or generating a noisy signal in the pseudo-Y domain from the pure signal in the X domain; performing a forward feature authenticity evaluation calculation on the pure signal in the pseudo-X domain; performing a reverse feature authenticity evaluation calculation on the noisy signal in the pseudo-Y domain; alternately optimizing the two conversion paths; and performing noise reduction processing on the measured noisy data using the optimized forward conversion path. This application enhances the suppression effect on random interference in actual field detection signals with more complex noise conditions and lower signal-to-noise ratios.

Description

Technical Field

[0001] This application belongs to the technical field of Magnetic Resonance Sounding (MRS), and specifically relates to a method and system for intelligently suppressing magnetic resonance random noise based on dynamic weighting. Background Art

[0002] With the increasingly severe global fresh water crisis and frequent occurrence of water-source geological disasters, the development of efficient groundwater detection technologies has become an urgent need to ensure water resource security and prevent and control geological disasters. As a non-invasive geophysical method for directly detecting groundwater, surface nuclear magnetic resonance has significant advantages such as fast speed, high efficiency, and rich information content, and has received extensive attention and application in recent years.

[0003] However, the magnetic resonance signal during actual detection is very weak, usually at the nanovolt level, and is extremely vulnerable to various electromagnetic noise interferences in the surrounding environment, especially a large amount of random noise. For random noise, due to its randomness in statistical characteristics, there is no unified law, and its frequency distribution range is wide, often overlapping with the frequency band of the magnetic resonance signal, making it difficult to effectively extract the signal and being unfavorable for subsequent inversion and interpretation of hydrogeological parameters. Therefore, effectively suppressing random noise is of great significance for improving the efficiency of magnetic resonance groundwater detection.

[0004] Patent CN116148935A discloses "a method for suppressing magnetic resonance random noise based on an adaptive autoencoder". This method constructs a noise suppression model to suppress the random noise in magnetic resonance groundwater detection through the noise suppression model, including: adding random noise to multiple groups of ideal magnetic resonance signals to obtain a data set; building a model, the model includes an encoder and a decoder, and initializing network parameters; training the model using the training set S, extracting features of the data through the encoder to obtain a latent variable z, and reconstructing an effective signal from the latent variable z through the decoder; enabling the autoencoder to establish a probability distribution model of the training samples to learn the distribution law of the signals, determining a loss function based on the deviation between the input signal and the reconstructed signal, and introducing a parameter estimation error to constrain the loss function, updating the network model parameters until the trend of the loss function is stable to obtain an adaptive autoencoder denoising model; using the test set T to test the denoising effect of the model. It solves the problem of limited denoising effect and improves the denoising efficiency. However, this method depends on the quality of the training data and is difficult to have strong generalization ability. When dealing with field detection signals with more complex actual random noise interference and lower signal-to-noise ratio, the denoising effect of the model is poor. Summary of the Invention

[0005] The embodiments of this application provide a method for intelligently suppressing magnetic resonance random noise based on dynamic weighting, which solves the problem of poor denoising effect when dealing with actual field detection signals with more complex random noise interference and lower signal-to-noise ratio.

[0006] On the other hand, an embodiment of the present application also provides an intelligent suppression system for magnetic resonance random noise based on dynamic weighting.

[0007] The technical solution of the embodiment of the present application includes:

[0008] An intelligent suppression method for magnetic resonance random noise based on dynamic weighting, the method includes:

[0009] Using a forward conversion path to generate a pure signal in the pseudo-X domain from a noisy signal in the Y domain, or generating a pure signal in the reconstructed X domain from a noisy signal in the pseudo-Y domain;

[0010] Using a reverse conversion path to generate a noisy signal in the reconstructed Y domain from the pure signal in the pseudo-X domain, or generating a noisy signal in the pseudo-Y domain from the pure signal in the X domain;

[0011] Performing forward feature authenticity evaluation calculation on the pure signal in the pseudo-X domain;

[0012] Performing reverse feature authenticity evaluation calculation on the noisy signal in the pseudo-Y domain;

[0013] Alternately optimizing the two conversion paths until the signal generated by the conversion path approaches the signal in the real target domain, and when the feature difference between the reconstructed signal and the original input signal reaches stable convergence, the optimization ends;

[0014] Using the optimized forward conversion path to perform noise reduction processing on the measured noisy data.

[0015] Further, both the forward conversion path and the reverse conversion path perform the following processing on the input signal:

[0016] Extracting the feature detail information of the signal through multiple downsampling convolution operations;

[0017] Using multiple residual transformation operations to gradually map the feature detail information to the target domain;

[0018] Parallelly performing first weighting processing and second weighting processing to dynamically weight the mapped feature detail data;

[0019] Gradually performing upsampling convolution operations to restore to the input signal dimension.

[0020] Further, the first weighted processing includes: performing a maximum pooling operation and an average pooling operation on each channel of the input feature detail information to obtain maximum pooling channel feature data and average pooling channel feature data; adding the maximum pooling channel feature data and the average pooling channel feature data after parallel processing through a shared weight transformation, and then obtaining channel weight coefficients by using a Sigmoid activation function; multiplying the weight channel coefficients element-wise with the input feature detail information to obtain the output after the first weighted processing.

[0021] Further, the second weighted processing includes:

[0022] Performing a maximum pooling operation and an average pooling operation on the input feature detail information along the channel dimension to obtain maximum pooling spatial feature data and average pooling spatial feature data; concatenating the maximum pooling spatial feature data and the average pooling spatial feature data along the channel dimension and then fusing them through a convolution operation, and then obtaining spatial weight coefficients by using a Sigmoid activation function; multiplying the spatial weight coefficients element-wise with the input feature detail information to obtain the output after the second weighted processing.

[0023] Further, the alternating optimization of the forward conversion path and the reverse conversion path includes:

[0024] Inputting the noisy signal in the Y domain, processing it through the forward conversion path to obtain a clean signal in the pseudo X domain, performing a forward feature authenticity evaluation calculation on the clean signal in the pseudo X domain and the clean signal in the real X domain, and updating the parameters of the forward conversion path according to the evaluation result, so that the noisy signal in the Y domain can obtain a clean signal closer to the real X domain after being processed through the forward conversion path;

[0025] Inputting the clean signal in the X domain, processing it through the reverse conversion path to generate a noisy signal in the pseudo Y domain, performing a reverse feature authenticity evaluation calculation on the noisy signal in the pseudo Y domain and the original noisy signal in the Y domain, and updating the parameters of the reverse conversion path according to the evaluation result;

[0026] Inputting the clean signal in the pseudo X domain obtained by processing through the forward conversion path into the reverse conversion path to obtain a reconstructed noisy signal in the Y domain, comparing the reconstructed noisy signal in the Y domain with the original noisy signal in the Y domain, and updating the parameters of the two conversion paths according to the feature difference; inputting the noisy signal in the pseudo Y domain obtained by processing through the reverse conversion path into the forward conversion path to obtain a reconstructed clean signal in the X domain, and comparing it with the clean signal in the real X domain, and updating the parameters of the two conversion paths according to the feature difference.

[0027] Further, both the forward feature authenticity evaluation calculation and the reverse feature authenticity evaluation calculation include: gradually extracting feature information through multiple downsampling convolution operations, performing feature normalization processing and non-linear activation processing after each convolution operation, obtaining a probability matrix through the last convolution operation, and taking the mean of the probability matrix as the feature authenticity evaluation result.

[0028] Further, the calculation of the feature difference includes:

[0029] Calculating the norm of the difference between the pure signal of each real X domain and the corresponding reconstructed pure signal of the X domain, and taking the average of all norms to obtain the first mean value;

[0030] Calculating the norm of the difference between the noise signal of each original Y domain and the corresponding reconstructed noisy signal of the Y domain, and taking the mean of all norms to obtain the second mean value;

[0031] Summing the first mean value and the second mean value to obtain the feature difference value.

[0032] On the other hand, an embodiment of the present application provides a magnetic resonance random noise intelligent suppression system based on dynamic weighting. The system includes:

[0033] A forward conversion module for generating a pure signal of the pseudo X domain from a noisy signal of the Y domain, or generating a reconstructed pure signal of the X domain from a noisy signal of the pseudo Y domain;

[0034] A reverse conversion module for generating a reconstructed noisy signal of the Y domain from the pure signal of the pseudo X domain, or generating a noisy signal of the pseudo Y domain from the pure signal of the X domain;

[0035] A forward evaluation module for performing forward feature authenticity evaluation calculation on the pure signal of the pseudo X domain;

[0036] A reverse evaluation module for performing reverse feature authenticity evaluation calculation on the noisy signal of the pseudo Y domain;

[0037] An optimization module for alternately optimizing the two conversion paths until the signals generated by the conversion paths approach the signals of the real target domain, and when the feature difference between the reconstructed signal and the original input signal reaches stable convergence, the optimization ends;

[0038] The optimized forward conversion module performs denoising processing on the measured noisy data.

[0039] Compared with the prior art, the present application has at least the following beneficial effects:

[0040] For multi-scenario field detection data with complex noise characteristics, this application can dynamically focus on key features and adversarial training, automatically adjust the feature extraction strategy for the actual noise environment, generate a more accurate noise model for blind denoising, and enhance the suppression effect on random interference in actual field detection signals with more complex noise conditions and lower signal-to-noise ratios. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 FIG. is a schematic flow chart of an intelligent magnetic resonance random noise suppression method based on dynamic weighting provided by an embodiment of this application;

[0042] Figure 2 FIG. is a structural block diagram of an intelligent magnetic resonance random noise suppression system based on dynamic weighting provided by an embodiment of this application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0043] In order to make the objectives, technical solutions and advantages of this application clearer, the following further details this application in conjunction with embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not used to limit this application.

[0044] See Figure 1 As shown, an intelligent magnetic resonance random noise suppression method based on dynamic weighting includes:

[0045] Using a forward conversion path to generate a pure signal in the pseudo-X domain from a noisy signal in the Y domain, or generating a pure signal in the reconstructed X domain from a noisy signal in the pseudo-Y domain;

[0046] Here, the space where the noisy signal is located is regarded as the Y domain, the space where the pure signal is located is regarded as the X domain, and the noisy signal in the Y domain is converted to the X domain. Since it is not the pure signal in the real X domain, the pure signal in the pseudo-X domain is used to represent it here. The noisy signal in the pseudo-Y domain is not the original noisy signal in the Y domain, but the noisy signal converted to the Y domain through other means.

[0047] S102 Using a reverse conversion path to generate a reconstructed noisy signal in the Y domain from the pure signal in the pseudo-X domain, or generating a noisy signal in the pseudo-Y domain from the pure signal in the X domain;

[0048] It can be understood that the forward conversion path is used to implement the mapping conversion process from the Y domain to the X domain, compare the generated pure signal in the pseudo-X domain with the pure signal in the real X domain, and perform feature authenticity evaluation calculation; the reverse conversion path is used to implement the reverse mapping conversion process from the X domain to the Y domain, compare the generated reconstructed noisy signal in the Y domain with the original noisy signal in the Y domain, and perform feature authenticity evaluation calculation.

[0049] Performing forward feature authenticity evaluation calculation on the pure signal in the pseudo-X domain;

[0050] By using the forward conversion path to generate a clean signal in the pseudo-X domain from the noisy signal in the Y domain, it is necessary to verify whether the clean signal in the pseudo-X domain is close to the real clean signal in the X domain. Here, the clean signal in the pseudo-X domain is evaluated, and the forward conversion path is adjusted according to the evaluation result.

[0051] Perform reverse feature authenticity evaluation calculation on the noisy signal in the pseudo-Y domain;

[0052] It can be understood that when using the reverse conversion path to generate a noisy signal in the pseudo-Y domain from the clean signal in the X domain, it is necessary to verify whether the noisy signal in the pseudo-Y domain is close to the original noisy signal in the Y domain. Here, the noisy signal in the pseudo-Y domain is evaluated, and the noisy signal in the pseudo-Y domain is adjusted according to the evaluation result.

[0053] Alternately optimize the two conversion paths until the signals generated by the conversion paths approach the signals in the real target domain, and when the feature difference between the reconstructed signal and the original input signal reaches stable convergence, the optimization ends;

[0054] The target domain here refers to the domain obtained through the conversion path, which can be the Y domain or the X domain. For example: when using the forward conversion path to generate a clean signal in the pseudo-X domain from the noisy signal in the Y domain, the domain where the clean signal in the pseudo-X domain is located is the target domain. In this example, the signal in the real target domain refers to the original clean signal or the real clean signal in the X domain. Another example is when using the reverse conversion path to generate a noisy signal in the pseudo-Y domain from the clean signal in the X domain, the domain where the noisy signal in the pseudo-Y domain is located is the target domain. In this example, the signal in the real target domain refers to the real noisy signal or the original noisy signal in the Y domain.

[0055] The reconstructed signal includes the reconstructed clean signal in the X domain obtained by reconstructing the noisy signal in the pseudo-Y domain into a clean signal in the X domain, and the reconstructed noisy signal in the Y domain obtained by reconstructing the clean signal in the pseudo-X domain into a noisy signal in the Y domain; the original input signal refers to the original (real) noisy signal in the Y domain and the real (original) clean signal in the X domain.

[0056] Alternately optimizing the two conversion paths includes: using the original (real) noisy signal in the Y domain as the input, generating a clean signal in the pseudo-X domain from the noisy signal in the Y domain through the forward conversion path, and then generating a reconstructed noisy signal in the Y domain from the clean signal in the pseudo-X domain through the reverse conversion path;

[0057] Using the real (original) clean signal in the X domain as the input to generate a noisy signal in the pseudo-Y domain from the clean signal in the X domain, and then generating a reconstructed clean signal in the X domain from the noisy signal in the pseudo-Y domain through the forward conversion path.

[0058] And perform a forward feature authenticity evaluation calculation on the pure signal in the pseudo-X domain;

[0059] Perform a reverse feature authenticity evaluation calculation on the noisy signal in the pseudo-Y domain.

[0060] Continuously repeat the above process until two conditions are met. The first condition is that the signal generated by the conversion path approximates the signal in the real target domain, that is, the signal in the real target domain and the signal generated by the conversion path cannot be distinguished. For example, the signal in the real target domain is the pure signal in the real X domain, and the signal generated by the conversion path is the pure signal in the pseudo-X domain; The second condition is that when the feature difference between the reconstructed signal and the original input signal reaches stable convergence. For example, the reconstructed signal is the noisy signal in the reconstructed Y domain, and the original input signal is the noisy signal in the original Y domain.

[0061] Use the optimized forward conversion path to denoise the measured noisy data.

[0062] In the embodiment of the present application, for multi-scenario field detection data with complex noise characteristics, unsupervised training is achieved through the forward conversion path and reverse conversion path of dynamic weighted bidirectional mapping, as well as forward feature authenticity evaluation calculation and reverse feature authenticity evaluation calculation. It can dynamically focus on key features and adversarial training, automatically adjust the feature extraction strategy for the actual noise environment, generate a more accurate forward conversion path for blind denoising, and enhance the suppression effect on random interference in actual field detection signals with more complex noise conditions and lower signal-to-noise ratios.

[0063] In one embodiment, the noisy signal in the Y domain is a pure signal containing noise, and the dataset of the noisy signal in the Y domain and the dataset of the pure signal in the X domain , meaning, , the noisy signal in the Y domain is from the actual field magnetic resonance noisy signal collected by a nuclear magnetic resonance groundwater detector, and the pure signal in the X domain is from the magnetic resonance signal simulated according to the expression . Select parameters according to the actual exploration situation. For example, the initial amplitude is 200 nV, the relaxation time is 150 ms, the Larmor frequency is 2320 Hz, the initial phase is 0, t = (0:N - 1) / , the sampling rate is 50 kHz, and the data length is 10000.

[0064] In one embodiment, both the forward conversion path and the reverse conversion path perform the following processing on the input signal:

[0065] Extract the feature details of the signal through multiple downsampling convolution operations;

[0066] Use multiple residual transformation operations to gradually map the feature details to the target domain;

[0067] Parallelly perform the first weighted processing and the second weighted processing to dynamically weight the mapped feature detail data;

[0068] Restore to the input signal dimension through successive upsampling convolution operations.

[0069] Specifically: The input data first undergoes multiple downsampling operations. In one embodiment, convolution operations, data normalization processing, and non-linear activation processing can be sequentially performed to obtain feature detail information, which can be deep feature data; then through multiple residual transformation operations, for example, convolution operations, non-linear activation, adding the result after convolution operations to the input, and then through non-linear activation processing, the feature detail information is gradually mapped to the target domain; then it passes through the first weighted processing and the second weighted processing in parallel to dynamically weight the transformed feature detail data; finally, through multiple upsampling operations, for example, inverse convolution operations, data normalization, and non-linear activation processing can be sequentially performed to restore the input signal dimension or the original signal dimension, and the data after the transformation processing is output.

[0070] In one embodiment, the first weighted processing includes: performing max pooling operations and average pooling operations on each channel of the input feature detail information respectively to obtain max pooling channel feature data and average pooling channel feature data; adding the max pooling channel feature data and the average pooling channel feature data after parallel processing through a shared weight transformation, and then using the Sigmoid activation function to obtain channel weight coefficients; multiplying the weight channel coefficients element-wise with the input feature detail information to obtain the output after the first weighted processing.

[0071] The second weighted processing includes: performing max pooling operations and average pooling operations on the input feature detail information along the channel dimension respectively to obtain max pooling spatial feature data and average pooling spatial feature data; concatenating the max pooling spatial feature data and the average pooling spatial feature data along the channel dimension and then fusing them through convolution operations, and then using the Sigmoid activation function to obtain spatial weight coefficients; multiplying the spatial weight coefficients element-wise with the input feature detail information to obtain the output after the second weighted processing.

[0072] In one embodiment, the first weighted processing and the second weighted processing perform different dynamic weightings on the converted feature data through two parallel independent paths, that is, according to the importance differences of each part in the converted feature data, the corresponding weight coefficients are assigned and then weighted and fused to obtain the output after focusing on the key features. The two obtained weighted outputs are concatenated along the channel dimension, and then the number of channels after concatenation is compressed back to the original number of channels through a 1x1 convolution operation. Thus, the key information of the input features in the spatial and channel dimensions can be captured independently and comprehensively, avoiding the loss of sequential concatenated feature information, and the calculation efficiency is higher through two independent paths; and the gradients can be backpropagated to the two independent paths simultaneously, avoiding the problem of gradient attenuation caused by deep concatenation.

[0073] In one embodiment, the forward conversion path and the reverse conversion path are alternately optimized, including:

[0074] Input the noisy signal in the Y domain, and obtain the pure signal in the pseudo-X domain through the forward conversion path. Perform forward feature authenticity evaluation calculation on the pure signal in the pseudo-X domain and the real pure signal in the X domain, and update the parameters of the forward conversion path according to the evaluation result, so that the noisy signal in the Y domain can obtain a pure signal closer to the real X domain after being processed by the forward conversion path;

[0075] Input the pure signal in the X domain, generate a noisy signal in the pseudo-Y domain through the reverse conversion path, perform reverse feature authenticity evaluation calculation on the noisy signal in the pseudo-Y domain and the original noisy signal in the Y domain, and update the parameters of the reverse conversion path according to the evaluation result;

[0076] Input the pure signal in the pseudo-X domain obtained by the forward conversion path into the reverse conversion path to obtain the reconstructed noisy signal in the Y domain. Compare the reconstructed noisy signal in the Y domain with the original noisy signal in the Y domain, and update the parameters of the two conversion paths according to the feature differences; Input the noisy signal in the pseudo-Y domain obtained by the reverse conversion path into the forward conversion path to obtain the reconstructed pure signal in the X domain, and compare it with the real pure signal in the X domain, and update the parameters of the two conversion paths according to the feature differences.

[0077] In one embodiment, the feature difference is used to measure whether the parameters of the two conversion paths are optimized. The calculation of the feature difference includes: calculating the norm of the difference between each real pure signal in the X domain and the corresponding reconstructed pure signal in the X domain, and taking the average of all norms to obtain the first mean value;

[0078] Calculate the norm of the difference between each original noisy signal in the Y domain and the corresponding reconstructed noisy signal in the Y domain, and take the average of all norms to obtain the second mean value;

[0079] Sum the first mean value and the second mean value to obtain a feature difference value.

[0080] It can be understood that the feature difference value is an overall difference value, that is, the feature difference of the data in the forward conversion path and the reverse conversion path. When the overall feature difference value is the smallest and the signal generated by the conversion path approximates the signal of the true target domain and it is impossible to distinguish the original input signal from the generated signal, the obtained forward conversion path is the forward conversion path required finally.

[0081] In an embodiment of the present application, refer to Figure 2 the structural block diagram of an intelligent magnetic resonance random noise suppression system based on dynamic weighting shown in, an intelligent magnetic resonance random noise suppression system based on dynamic weighting, which can be correspondingly explained with an intelligent magnetic resonance random noise suppression method based on dynamic weighting. It includes: a forward conversion module, configured to generate a pseudo pure signal in the X domain from a noisy signal in the Y domain, or generate a reconstructed pure signal in the X domain from a pseudo noisy signal in the Y domain;

[0082] a reverse conversion module, configured to generate a reconstructed noisy signal in the Y domain from the pseudo pure signal in the X domain, or generate a pseudo noisy signal in the Y domain from the pure signal in the X domain;

[0083] a forward evaluation module, configured to perform forward feature authenticity evaluation calculation on the pseudo pure signal in the X domain;

[0084] a reverse evaluation module, configured to perform reverse feature authenticity evaluation calculation on the pseudo noisy signal in the Y domain;

[0085] an optimization module, which alternately optimizes the two conversion paths until the signal generated by the conversion path approximates the signal of the true target domain and the feature difference between the reconstructed signal and the original input signal reaches stable convergence, and then ends the optimization;

[0086] The optimized forward conversion module performs denoising processing on the measured noisy data.

[0087] In the embodiment of the present application, for multi-scenario field detection data with complex noise characteristics, an unsupervised training system is implemented through the forward conversion module and the reverse conversion module of dynamic weighted bidirectional mapping and the forward evaluation module and the reverse evaluation module, which can dynamically focus on key features and adversarial training, automatically adjust the feature extraction strategy of the actual noise environment, generate a more accurate forward conversion module for blind denoising, and enhance the suppression effect on random interference in actual field detection signals with more complex noise conditions and lower signal-to-noise ratio.

[0088] The above are only the preferred embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present application shall be included within the protection scope of the present application.

Claims

1. An intelligent suppression method for random noise in magnetic resonance based on dynamic weighting, characterized in that, The method includes: Generating a clean signal in the pseudo-X domain from a noisy signal in the Y domain using a forward conversion path, or generating a clean signal in the reconstructed X domain from a noisy signal in the pseudo-Y domain; Generating a noisy signal in the reconstructed Y domain from the clean signal in the pseudo-X domain using a reverse conversion path, or generating a noisy signal in the pseudo-Y domain from the clean signal in the X domain; Performing a forward feature authenticity evaluation calculation on the clean signal in the pseudo-X domain; Performing a reverse feature authenticity evaluation calculation on the noisy signal in the pseudo-Y domain; Alternately optimizing the two conversion paths until the signals generated by the conversion paths approximate the signals in the true target domain and the feature difference between the reconstructed signal and the original input signal reaches stable convergence, and then ending the optimization; Performing noise reduction processing on the measured noisy data using the optimized forward conversion path.

2. The intelligent suppression method of magnetic resonance random noise based on dynamic weighting according to claim 1, characterized in that Both the forward conversion path and the reverse conversion path perform the following processing on the input signal: Extracting the feature detail information of the signal through multiple downsampling convolution operations; Gradually mapping the feature detail information to the target domain using multiple residual transformation operations; Dynamically weighting the mapped feature detail data by parallelly adopting the first weighting process and the second weighting process; Restoring to the input signal dimension through a gradual upsampling convolution operation.

3. The intelligent magnetic resonance random noise suppression method based on dynamic weighting according to claim 2, wherein The first weighting process includes: performing a max pooling operation and an average pooling operation on each channel of the input feature detail information respectively to obtain max pooling channel feature data and average pooling channel feature data; adding the max pooling channel feature data and the average pooling channel feature data after parallel processing through a shared weight transformation, and then obtaining a channel weight coefficient using a Sigmoid activation function; multiplying the weight channel coefficient element-wise with the input feature detail information to obtain the output after the first weighting process.

4. The intelligent magnetic resonance random noise suppression method based on dynamic weighting according to claim 2, wherein The second weighting process includes: Performing a max pooling operation and an average pooling operation on the input feature detail information along the channel dimension respectively to obtain max pooling spatial feature data and average pooling spatial feature data; splicing the max pooling spatial feature data and the average pooling spatial feature data along the channel dimension and then fusing them through a convolution operation, and then obtaining a spatial weight coefficient using a Sigmoid activation function; multiplying the spatial weight coefficient element-wise with the input feature detail information to obtain the output after the second weighting process.

5. The intelligent suppression method for magnetic resonance random noise based on dynamic weighting according to claim 1, wherein The alternately optimizing the forward conversion path and the reverse conversion path includes: Inputting a noisy signal in the Y domain, processing it through the forward conversion path to obtain a clean signal in the pseudo-X domain, performing a forward feature authenticity evaluation calculation on the clean signal in the pseudo-X domain and the clean signal in the true X domain, and updating the parameters of the forward conversion path according to the evaluation result, so that the noisy signal in the Y domain can obtain a clean signal closer to the true X domain after being processed by the forward conversion path; The pure signal in the input X domain is processed through the reverse conversion path to generate a noisy signal in the pseudo Y domain. The noisy signal in the pseudo Y domain and the original noisy signal in the Y domain are subjected to reverse feature authenticity evaluation calculation, and the parameters of the reverse conversion path are updated according to the evaluation result; The pure signal in the pseudo X domain obtained by processing through the forward conversion path is input into the reverse conversion path to obtain the reconstructed noisy signal in the Y domain. The reconstructed noisy signal in the Y domain is compared with the original noisy signal in the Y domain, and the parameters of the two conversion paths are updated according to the feature difference; The noisy signal in the pseudo Y domain obtained by processing through the reverse conversion path is input into the forward conversion path to obtain the reconstructed pure signal in the X domain, and it is compared with the real pure signal in the X domain, and the parameters of the two conversion paths are updated according to the feature difference.

6. The intelligent suppression method for random noise in magnetic resonance based on dynamic weighting according to claim 1, wherein Both the forward feature authenticity evaluation calculation and the reverse feature authenticity evaluation calculation include: gradually extracting feature information through multiple downsampling convolution operations, performing feature normalization processing and non-linear activation processing after each convolution operation, obtaining a probability matrix through the final convolution operation, and taking the mean value of the probability matrix as the feature authenticity evaluation result.

7. The intelligent suppression method for magnetic resonance random noise based on dynamic weighting according to claim 6, wherein The calculation of the feature difference includes: Calculating the norm of the difference between the pure signal of each true X domain and the corresponding reconstructed pure signal of the X domain, and averaging all norms to obtain the first mean value; ​ Calculating the norm of the difference between the noise signal of each original Y domain and the noisy signal of the corresponding reconstructed Y domain, and taking the mean of all norms to obtain a second mean value; ​ The sum of the first mean value and the second mean value is obtained as the feature difference value.

8. An intelligent suppression system for magnetic resonance random noise based on dynamic weighting, characterized in that, The system includes: A forward conversion module for generating a pure signal in the pseudo X domain from a noisy signal in the Y domain, or generating a reconstructed pure signal in the X domain from a noisy signal in the pseudo Y domain; A reverse conversion module for generating a reconstructed noisy signal in the Y domain from a pure signal in the pseudo X domain, or generating a noisy signal in the pseudo Y domain from a pure signal in the X domain; A forward evaluation module for performing forward feature authenticity evaluation calculation on the pure signal in the pseudo X domain; A reverse evaluation module for performing reverse feature authenticity evaluation calculation on the noisy signal in the pseudo Y domain; An optimization module alternately optimizes the two conversion paths until the signal generated by the conversion path approaches the signal in the real target domain, and when the feature difference between the reconstructed signal and the original input signal reaches stable convergence, the optimization ends; The optimized forward conversion module performs denoising processing on the measured noisy data.

Images