Medical image processing method and device, computer device and storage medium

Abstract

The application relates to a medical image processing method and device, computer equipment, a storage medium and a computer program product. The method comprises the following steps: acquiring a sample artifact image set; each sample artifact image in the sample artifact image set is obtained by performing magnetic resonance scanning on each scanning sample by using an echo imaging sequence; a template magnetic resonance image corresponding to each scanning sample is determined by using a reference echo scanning method; a neural network to be trained is trained according to the sample artifact image corresponding to each scanning sample and the corresponding template magnetic resonance image, so as to obtain a trained image artifact elimination model; and the image artifact elimination model is used for performing peak correction on a magnetic resonance image to be eliminated. The method can improve the artifact removal efficiency of the magnetic resonance image.

Description

Technical Field

[0001] This application relates to the field of image processing technology, and in particular to a medical image method, apparatus, computer equipment, storage medium, and computer program product. Background Technology

[0002] Magnetic resonance imaging (MRI) plays a vital role in clinical diagnosis and scientific research due to its lack of ionizing radiation and its ability to capture multiple imaging parameters. Echoplanar spectroscopic imaging (EPSI) is a rapid spectral imaging technique within MRI.

[0003] Compared to traditional spectral imaging, EPSI significantly reduces sampling time, primarily due to the addition of a readout gradient with continuously changing polarity unique to Echo Planar Imaging (EPI). However, the introduction of this EPI readout gradient also introduces the Nyquist artifact, a characteristic of EPI. Related techniques typically employ a reference echo scan method to remove artifacts, which involves performing a second reference echo sequence scan without phase gradient encoding after the normal sequence scan. However, the double scan increases the scan time, making it inefficient at removing artifacts from magnetic resonance images.

[0004] Therefore, the related technologies suffer from the problem of low artifact removal efficiency in magnetic resonance images. Summary of the Invention

[0005] Therefore, it is necessary to provide a medical image processing method, apparatus, computer equipment, computer-readable storage medium, and computer program product that can improve the efficiency of artifact removal in magnetic resonance images, in order to address the above-mentioned technical problems.

[0006] Firstly, this application provides a medical image processing method. The method includes:

[0007] A sample artifact image set is obtained; each sample artifact image in the sample artifact image set is obtained by performing magnetic resonance scanning on each scanned sample using an echo imaging sequence;

[0008] The template magnetic resonance image corresponding to each scanned sample is determined by the reference echo scanning method;

[0009] The neural network to be trained is trained based on the sample artifact images and the corresponding template magnetic resonance images of each scanned sample to obtain a trained image artifact elimination model; the image artifact elimination model is used to perform peak correction on the magnetic resonance images with artifacts to be eliminated.

[0010] In one embodiment, the template magnetic resonance image corresponding to each scanned sample is determined by a reference echo scanning method, including:

[0011] For any of the scanned samples, obtain the k-space data to be corrected and the target reference echo data corresponding to the scanned sample under the reference scan sequence;

[0012] A phase compensation matrix is ​​established based on the phase difference between each echo and the first echo in the target reference echo data;

[0013] Based on the phase compensation matrix, peak correction is performed on the k-space data to be corrected corresponding to any scan sample to obtain peak-corrected k-space data;

[0014] The peak-corrected k-space data is reconstructed to obtain the template magnetic resonance image corresponding to any scanned sample.

[0015] In one embodiment, the step of performing peak correction on the k-space data to be corrected corresponding to any scan sample according to the phase compensation matrix to obtain peak-corrected k-space data includes:

[0016] Based on the phase compensation matrix, the peak values ​​corresponding to odd echoes and even echoes in the k-space data to be corrected are corrected to obtain the peak-corrected k-space data.

[0017] In one embodiment, the step of performing peak correction on the k-space data to be corrected corresponding to any scan sample according to the phase compensation matrix to obtain peak-corrected k-space data includes:

[0018] Based on the phase compensation matrix, the half-width at half maximum (WHM) corresponding to odd echoes and the half-width at half maximum (WHM) corresponding to even echoes in the k-space data to be corrected are corrected to obtain the peak-corrected k-space data.

[0019] In one embodiment, the step of performing peak correction on the k-space data to be corrected corresponding to any scan sample according to the phase compensation matrix to obtain peak-corrected k-space data includes:

[0020] Based on the phase compensation matrix, the spectral positions corresponding to odd echoes and even echoes in the k-space data to be corrected are corrected to obtain the peak-corrected k-space data.

[0021] In one embodiment, the neural network is a convolutional neural network; the convolutional neural network includes a feature extraction unit and a fully connected unit.

[0022] Secondly, this application provides another medical image processing method. The method includes:

[0023] The magnetic resonance image of the artifact to be removed and the trained image artifact removal model are obtained; the magnetic resonance image of the artifact to be removed is obtained by magnetic resonance scanning of the target object using an echo imaging sequence;

[0024] The magnetic resonance image of the artifact to be eliminated is input into the trained image artifact elimination model to obtain the artifact-eliminated magnetic resonance image; the artifact-eliminated magnetic resonance image is obtained by peak correction of the magnetic resonance image of the artifact to be eliminated by the trained image artifact elimination model.

[0025] In one embodiment, the trained image artifact removal model includes a feature extraction unit and a fully connected unit; the step of inputting the magnetic resonance image with artifacts to be removed into the trained image artifact removal model to obtain the artifact-removed magnetic resonance image includes:

[0026] The magnetic resonance image of the artifact to be removed is input into the feature extraction unit to obtain the image features corresponding to the magnetic resonance image of the artifact to be removed; the image features are obtained by the feature extraction unit from the magnetic resonance image of the artifact to be removed.

[0027] The image features are input into the fully connected unit to obtain the artifact-free magnetic resonance image; the artifact-free magnetic resonance image is obtained by the fully connected unit performing feature classification on the image features.

[0028] Thirdly, this application also provides a medical image processing apparatus. The apparatus includes:

[0029] The acquisition module is used to acquire a set of sample artifact images; each sample artifact image in the set is obtained by performing magnetic resonance scanning on each scanned sample using an echo imaging sequence.

[0030] The determination module is used to determine the template magnetic resonance image corresponding to each scanned sample by means of a reference echo scanning method;

[0031] The training module is used to train the neural network to be trained based on the sample artifact images and the corresponding template magnetic resonance images of each scanned sample, so as to obtain a trained image artifact removal model; the image artifact removal model is used to perform peak correction on the magnetic resonance image to be removed from the artifacts.

[0032] Fourthly, this application also provides a medical image processing apparatus. The apparatus includes:

[0033] The acquisition module is used to acquire the magnetic resonance image of the artifact to be eliminated and the trained image artifact elimination model; the magnetic resonance image of the artifact to be eliminated is obtained by performing magnetic resonance scanning on the target object using an echo imaging sequence;

[0034] The input module is used to input the magnetic resonance image of the artifact to be eliminated into the trained image artifact elimination model to obtain the artifact-eliminated magnetic resonance image; the artifact-eliminated magnetic resonance image is obtained by peak correction of the magnetic resonance image of the artifact to be eliminated by the trained image artifact elimination model.

[0035] Fifthly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the medical image processing method as described in the first aspect or any embodiment of the first aspect, or the medical image processing method as described in the second aspect or any embodiment of the second aspect.

[0036] Sixthly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the medical image processing method as described in the first aspect or any embodiment of the first aspect, or the medical image processing method as described in the second aspect or any embodiment of the second aspect.

[0037] In a seventh aspect, this application also provides a computer program product. The computer program product includes a computer program that, when executed by a processor, implements the medical image processing method as described in the first aspect or any embodiment of the first aspect, or the medical image processing method as described in the second aspect or any embodiment of the second aspect.

[0038] The aforementioned medical image processing method, apparatus, computer equipment, storage medium, and computer program product acquire a sample artifact image set; each sample artifact image in the sample artifact image set is obtained by performing magnetic resonance scanning on each scanned sample using an echo imaging sequence; a template magnetic resonance image corresponding to each scanned sample is determined by using a reference echo scanning method; a neural network to be trained is trained based on the sample artifact image and the corresponding template magnetic resonance image of each scanned sample to obtain a trained image artifact removal model; the image artifact removal model is used to perform peak correction on the magnetic resonance image of the artifact to be removed.

[0039] Thus, by training the neural network to be trained using the sample artifact images with artifacts and the corresponding template magnetic resonance images without artifacts for each scanned sample, a trained image artifact removal model can be obtained. This trained image artifact removal model can directly perform peak correction on the magnetic resonance image with artifacts to be removed. Since the influence of artifacts in magnetic resonance scanning is that a non-existent peak appears in the image, while reducing the peak values ​​of other compounds, affecting the judgment of the results, the peak correction processing of the trained image artifact removal model can directly obtain the magnetic resonance image with artifact removal corresponding to the magnetic resonance image with artifacts to be removed. There is no need to perform another reference scan to obtain the magnetic resonance image with artifact removal after obtaining the magnetic resonance image with artifacts to be removed, thereby reducing the scanning time and quickly obtaining the magnetic resonance image with artifact removal, effectively improving the artifact removal efficiency in magnetic resonance images. Attached Figure Description

[0040] Figure 1 This is a flowchart illustrating a medical image processing method in one embodiment;

[0041] Figure 2 This is a schematic diagram of the structure of a convolutional neural network in one embodiment;

[0042] Figure 3 This is a flowchart illustrating another medical image processing method in one embodiment;

[0043] Figure 4 This is a flowchart illustrating a medical image processing method in another embodiment;

[0044] Figure 5 This is a structural block diagram of a medical image processing device in one embodiment;

[0045] Figure 6 This is a structural block diagram of a medical image processing device in another embodiment;

[0046] Figure 7 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0048] It should be noted that the terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this disclosure are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this disclosure described herein can be implemented in orders other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure as detailed in the appended claims.

[0049] In one embodiment, such as Figure 1 As shown, a medical image processing method is provided. This embodiment illustrates the method applied to a terminal. It is understood that this method can also be applied to a server, and furthermore, to a system including both a terminal and a server, and is implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0050] Step S110: Obtain the sample artifact image set.

[0051] Among them, the artifact images of each sample in the sample artifact image set were obtained by performing magnetic resonance scanning on each scanned sample using an echo imaging sequence.

[0052] Among them, the echo imaging sequence can be an echo planar imaging (EPI) sequence or an echo planar spectroscopic imaging (EPSI) sequence.

[0053] The sample artifact images can be either EPSI images with N / 2 Ghost artifacts or EPI images with N / 2 Ghost artifacts.

[0054] In practice, magnetic resonance scanning equipment can use echo imaging sequences to perform magnetic resonance scanning on each scanned sample, so that the terminal can obtain the sample artifact images corresponding to each scanned sample and form a sample artifact image set.

[0055] Step S120: Determine the template magnetic resonance image corresponding to each scanned sample by using the reference echo scanning method.

[0056] The template magnetic resonance image is either the magnetic resonance image of the scanned sample that does not contain N / 2Ghost artifacts, or the magnetic resonance image that does not contain N / 2Ghost artifacts.

[0057] In practice, based on the reference echo scanning method, the terminal can acquire the template magnetic resonance image corresponding to each scanned sample.

[0058] Step S130: Train the neural network to be trained based on the sample artifact images and the corresponding template magnetic resonance images of each scanned sample to obtain a trained image artifact elimination model.

[0059] Among them, the image artifact removal model is used to perform peak correction on the magnetic resonance image with artifacts to be removed.

[0060] In practice, the terminal can train the neural network to be trained based on the sample artifact images and the corresponding template magnetic resonance images of each scanned sample to obtain a trained image artifact elimination model. This trained image artifact elimination model is used to perform peak correction on the magnetic resonance image with artifacts to be eliminated. Since the influence of N / 2Ghost artifacts in magnetic resonance scanning is that a non-existent peak appears in the image, while reducing the peak values ​​of other compounds, it affects the judgment of the results. Therefore, by performing peak correction on the magnetic resonance image with artifacts to be eliminated, the magnetic resonance image with artifact elimination can be obtained.

[0061] In practical applications, the sample artifact image and the corresponding template magnetic resonance image of each scan sample can form a set of sample data. The terminal can use a preset proportion of the sample data from multiple sets of sample data for training the neural network, and use the remaining sample data for testing the neural network (for example, 70% of the sample data is used for training and 30% of the sample data is used for testing).

[0062] In the aforementioned medical image processing method, a sample artifact image set is acquired; each sample artifact image in the sample artifact image set is obtained by performing magnetic resonance scanning on each scanned sample using an echo imaging sequence; the template magnetic resonance image corresponding to each scanned sample is determined by the reference echo scanning method; the neural network to be trained is trained based on the sample artifact image and the corresponding template magnetic resonance image of each scanned sample to obtain a trained image artifact elimination model; the image artifact elimination model is used to perform peak correction on the magnetic resonance image with artifacts to be eliminated.

[0063] Thus, by training the neural network to be trained using the sample artifact images with artifacts and the corresponding template magnetic resonance images without artifacts for each scanned sample, a trained image artifact removal model can be obtained. This trained image artifact removal model can directly perform peak correction on the magnetic resonance image with artifacts to be removed. Since the influence of artifacts in magnetic resonance scanning is that a non-existent peak appears in the image, while reducing the peak values ​​of other compounds, affecting the judgment of the results, the peak correction processing of the trained image artifact removal model can directly obtain the magnetic resonance image with artifact removal corresponding to the magnetic resonance image with artifacts to be removed. There is no need to perform another reference scan to obtain the magnetic resonance image with artifact removal after obtaining the magnetic resonance image with artifacts to be removed, thereby reducing the scanning time and quickly obtaining the magnetic resonance image with artifact removal, effectively improving the artifact removal efficiency in magnetic resonance images.

[0064] In one embodiment, determining the template magnetic resonance image corresponding to each scan sample using the reference echo scanning method includes: for any scan sample, acquiring the k-space data to be corrected and the target reference echo data corresponding to the scan sample under the reference scanning sequence; establishing a phase compensation matrix based on the phase difference between each echo and the first echo in the reference echo data; performing peak correction on the k-space data to be corrected corresponding to any scan sample based on the phase compensation matrix to obtain peak-corrected k-space data; and reconstructing the peak-corrected k-space data to obtain the template magnetic resonance image corresponding to any scan sample.

[0065] The reference scan sequence is a scan sequence without phase gradient encoding.

[0066] In specific implementation, during the process of determining the template magnetic resonance image corresponding to each scan sample using the reference echo scanning method, for any given scan sample, the terminal can acquire the k-space data to be corrected and the reference echo data corresponding to that scan sample under the reference scanning sequence. Then, the terminal can perform a one-dimensional Fourier transform on the reference echo data along the readout direction to obtain the target reference echo data corresponding to that scan sample under the reference scanning sequence. Next, the terminal can use the first echo in the target reference echo data corresponding to that scan sample as a reference to determine the phase difference between each echo in the target reference echo data and the first echo, and establish a phase compensation matrix based on the phase difference. Then, the terminal can perform peak correction on the k-space data to be corrected corresponding to that scan sample based on the phase compensation matrix to obtain the peak-corrected k-space data corresponding to that scan sample. Finally, the terminal can perform a two-dimensional inverse Fourier transform on the peak-corrected k-space data corresponding to that scan sample to reconstruct the template magnetic resonance image corresponding to that scan sample. In this way, the terminal can acquire the template magnetic resonance image corresponding to each scan sample.

[0067] The technical solution of this embodiment involves acquiring, for any scanned sample, the k-space data to be corrected and the target reference echo data corresponding to that scanned sample under a reference scan sequence; establishing a phase compensation matrix based on the phase difference between each echo and the first echo in the reference echo data; performing peak correction on the k-space data to be corrected corresponding to any scanned sample based on the phase compensation matrix to obtain peak-corrected k-space data; and reconstructing the peak-corrected k-space data to obtain the template magnetic resonance image corresponding to any scanned sample. Thus, by using the phase difference in the target reference echo data corresponding to the scanned sample under a reference scan sequence to perform peak correction on the k-space data to be corrected corresponding to the scanned sample and reconstructing the template magnetic resonance image, the phase error caused by the k-space offset between odd and even echoes of the readout gradient in the image domain can be eliminated, accurately obtaining the template magnetic resonance image corresponding to the scanned sample without artifacts. Furthermore, based on the sample artifact image and the corresponding template magnetic resonance image corresponding to the scanned sample, an image artifact removal model that can accurately remove artifacts from the magnetic resonance image can be obtained.

[0068] In one embodiment, peak correction is performed on the k-space data to be corrected corresponding to any scan sample according to the phase compensation matrix to obtain peak-corrected k-space data, including: correcting the peak values ​​corresponding to odd echoes and even echoes in the k-space data to be corrected according to the phase compensation matrix to obtain peak-corrected k-space data.

[0069] In one embodiment, peak correction is performed on the k-space data to be corrected corresponding to any scan sample according to the phase compensation matrix to obtain peak-corrected k-space data, including: correcting the half-width at half maximum (WHM) corresponding to odd echoes and the half-width at half maximum (WHM) corresponding to even echoes in the k-space data to be corrected according to the phase compensation matrix to obtain peak-corrected k-space data.

[0070] In one embodiment, peak correction is performed on the k-space data to be corrected corresponding to any scan sample according to the phase compensation matrix to obtain peak-corrected k-space data, including: correcting the spectral positions corresponding to odd echoes and even echoes in the k-space data to be corrected according to the phase compensation matrix to obtain peak-corrected k-space data.

[0071] In practice, the Ghost artifact in EPSI or EPI images is caused by the phase error in the image domain due to the K-space offset between odd and even echoes of the readout gradient. This results in a non-existent peak and a reduction in the peak values ​​of other compounds, affecting the judgment of the results. Therefore, to eliminate the Ghost artifact, the terminal performs peak correction on the k-space data to be corrected for any scan sample according to the phase compensation matrix. In the process of obtaining peak-corrected k-space data, the terminal can directly correct the peak values ​​corresponding to odd echoes and even echoes in the k-space data to be corrected according to the phase compensation matrix, thus obtaining peak-corrected k-space data. Based on the peak-corrected k-space data, the terminal can reconstruct a template magnetic resonance image without the Ghost artifact.

[0072] In addition, the terminal can also correct the half-width at half maximum (WHM) of odd echoes and the half-width at half maximum (WHM) of even echoes in the k-space data to be corrected based on the phase compensation matrix. Based on the odd echoes after WHM correction and the even echoes after WHM correction, the peak-corrected k-space data can be obtained.

[0073] Simultaneously, the terminal can also correct the spectral positions corresponding to odd echoes and even echoes in the k-space data to be corrected based on the phase compensation matrix. Based on the odd echoes after spectral position correction and the even echoes after spectral position correction, the peak-corrected k-space data can be obtained.

[0074] Since the Ghost artifact is caused by the phase error in the image domain due to the K-space offset between the readout gradient odd and even echoes, the Ghost artifact results in a non-existent peak in the image and reduces the peak values ​​of other compounds, affecting the judgment of the results. The technical solution of this embodiment corrects the peak values ​​corresponding to odd echoes and even echoes in the k-space data to be corrected by using a phase compensation matrix, thus obtaining peak-corrected k-space data; or, it corrects the half-width at half-maximum (WHM) of odd echoes and even echoes in the k-space data to be corrected by using a phase compensation matrix, thus obtaining peak-corrected k-space data; or, it corrects the spectral positions corresponding to odd echoes and even echoes in the k-space data to be corrected by using a phase compensation matrix, thus obtaining peak-corrected k-space data.

[0075] Thus, during image reconstruction based on peak-corrected k-space data, the phase error caused by the K-space offset between readout gradient odd and even echoes in the image domain can be eliminated, resulting in a template MRI image free of Ghost artifacts after peak correction. Consequently, during the training of the neural network using the template MRI image and sample artifact images, the training features become the peak, full width at half maximum (FWHM), or spectral position in the image. This solves the problem of excessive training features caused by directly correcting Ghost artifacts in the image in related techniques, significantly reducing training difficulty, computational load, and computation time, while also greatly improving Ghost artifact removal efficiency. The resulting image artifact removal model can accurately perform peak correction on the MRI image with artifacts to be removed, achieving accurate removal of Ghost artifacts. For corrections of image deformation or other artifacts, machine learning or other artifact removal methods can be applied after this method.

[0076] In one embodiment, during the process of training the neural network to be trained based on the sample artifact images and the corresponding template magnetic resonance images of each scanned sample to obtain a trained image artifact elimination model, the terminal can input the sample artifact images corresponding to each scanned sample into the neural network to be trained to obtain the optimized images corresponding to each scanned sample. Then, the terminal can adjust the network parameters of the neural network to be trained based on the differences between the optimized images and the corresponding template magnetic resonance images of each scanned sample. Finally, the terminal can retrain the neural network with adjusted network parameters until the trained neural network meets the preset training termination conditions to obtain the trained image artifact elimination model.

[0077] The neural network is a convolutional neural network (CNN). The CNN includes a feature extraction unit and a fully connected unit. During the process of inputting the sample artifact images corresponding to each scanned sample into the neural network to be trained, and obtaining the optimized images corresponding to each scanned sample, the terminal can input any sample artifact image into the feature extraction unit to obtain the target feature corresponding to that sample artifact image. The target feature is obtained by the feature extraction unit extracting features from any sample artifact image. Then, the terminal can input the target feature into the fully connected unit to obtain the optimized image corresponding to that scanned sample. The optimized image is obtained by the fully connected layer in the fully connected unit performing feature classification on the target feature.

[0078] The feature extraction unit includes a first feature extraction unit and a second feature extraction unit. During the process of inputting any sample artifact image into the feature extraction unit to obtain the target feature corresponding to the sample artifact image, the terminal can input the any sample artifact image into the first feature extraction unit to obtain the feature map to be optimized corresponding to the sample artifact image. The feature map to be optimized is obtained by the first feature extraction unit through filtering the features of the extracted sample artifact image. Then, the terminal can input the feature map to be optimized corresponding to the sample artifact image into the second feature extraction unit to obtain the target feature corresponding to the sample artifact image. The target feature is obtained by the second feature extraction unit through filtering the features of the extracted feature map to be optimized.

[0079] The first feature extraction unit includes a first convolutional layer and a first downsampling layer. During the process of inputting any sample artifact image into the first feature extraction unit to obtain the corresponding feature map to be optimized, the terminal can input the any sample artifact image into the first convolutional layer to obtain the corresponding first feature map. The first feature map is obtained by the first convolutional layer performing feature extraction processing on the any sample artifact image. Then, the terminal can input the first feature map into the first downsampling layer to obtain the corresponding feature map to be optimized. The feature map to be optimized is obtained by the first downsampling layer performing feature filtering on the first feature map through downsampling processing.

[0080] The second feature extraction unit includes a second convolutional layer and a second downsampling layer. During the process of inputting the feature map to be optimized corresponding to any sample artifact image into the second feature extraction unit to obtain the target feature corresponding to the sample artifact image, the terminal can input the feature map to be optimized corresponding to any sample artifact image into the second convolutional layer to obtain a second feature map. The second feature map is obtained by the second convolutional layer performing feature extraction processing on the feature map to be optimized. Then, the terminal can input the second feature map into the second downsampling layer to obtain the target feature corresponding to the sample artifact image. The target feature is obtained by the second downsampling layer performing feature filtering on the second feature map through downsampling processing.

[0081] For the ease of understanding of those skilled in the art, Figure 2 A schematic diagram of a convolutional neural network is provided, such as... Figure 2 As shown, the convolutional neural network includes a first convolutional layer, a first downsampling layer, a second convolutional layer, a second downsampling layer in the feature extraction unit, and a fully connected layer in the fully connected unit.

[0082] In the technical solution of this embodiment, during the training of the neural network to be trained based on the sample artifact images and the corresponding template magnetic resonance images of each scanned sample, a convolutional neural network including feature extraction units and fully connected units is used for training. By using a relatively mature convolutional neural network to train and obtain an image artifact elimination model, the reliability can be effectively improved while removing artifacts in the magnetic resonance image through the image artifact elimination model.

[0083] In one embodiment, such as Figure 3 As shown, another medical image processing method is provided. This embodiment illustrates the application of this method to a terminal. In this embodiment, the method includes the following steps:

[0084] Step S310: Obtain the magnetic resonance image of the artifact to be removed and the trained image artifact removal model.

[0085] The magnetic resonance image to be removed is obtained by performing magnetic resonance scanning on the target object using an echo imaging sequence.

[0086] The magnetic resonance image to be removed can be an EPSI image with N / 2Ghost artifacts or an EPI image with N / 2Ghost artifacts.

[0087] In practice, the terminal can acquire the magnetic resonance image of the artifact to be eliminated and the trained image artifact elimination model.

[0088] Step S320: Input the magnetic resonance image with artifacts to be removed into the trained image artifact removal model to obtain the magnetic resonance image after artifact removal.

[0089] Among them, the artifact-removed magnetic resonance image is obtained by peak correction of the magnetic resonance image to be removed by the trained image artifact removal model.

[0090] In practice, the terminal can input the magnetic resonance image with artifacts to be removed into the trained image artifact removal model. The trained image artifact removal model performs peak correction on the magnetic resonance image with artifacts to be removed, and obtains the magnetic resonance image after artifact removal.

[0091] The technical solution of this embodiment involves acquiring a magnetic resonance image (MRI) of the object to be artifact-free and a trained image artifact removal model. The MRI image is obtained by performing an MRI scan on the target object using an echo imaging sequence. The MRI image is then input into the trained image artifact removal model to obtain an MRI image with artifact removal removed. This MRI image is obtained by peak correction of the MRI image with artifact removal removed using the trained image artifact removal model. Thus, by performing peak correction on the MRI image with artifact removal removed using the trained image artifact removal model, the MRI image with artifact removal removed can be directly obtained without performing a reference scan on the target object after acquiring the MRI image with artifact removal removed. This shortens the scanning time, saves scanning costs, improves the efficiency of artifact removal in the MRI image, and avoids errors caused by inconsistent positioning between the two scans, thereby accurately removing artifacts from the MRI image.

[0092] In one embodiment, the trained image artifact removal model includes a feature extraction unit and a fully connected unit. Inputting the magnetic resonance image with artifacts to be removed into the trained image artifact removal model to obtain the artifact-removed magnetic resonance image includes: inputting the magnetic resonance image with artifacts to be removed into the feature extraction unit to obtain image features corresponding to the magnetic resonance image with artifacts to be removed; the image features are obtained by the feature extraction unit performing feature extraction on the magnetic resonance image with artifacts to be removed; inputting the image features into the fully connected unit to obtain the artifact-removed magnetic resonance image; the artifact-removed magnetic resonance image is obtained by the fully connected unit performing feature classification on the image features.

[0093] In specific implementation, the trained image artifact removal model includes a feature extraction unit and a fully connected unit. During the process of inputting the magnetic resonance image (MRA) containing the artifacts to be removed into the trained model to obtain the artifact-removed MRA image, the terminal can input the MRA image containing the artifacts to be removed into the feature extraction unit. The feature extraction unit extracts features from the MRA image containing the artifacts to be removed, obtaining the image features corresponding to the MRA image containing the artifacts to be removed. Then, the terminal can input the image features corresponding to the MRA image containing the artifacts to be removed into the fully connected unit. The fully connected unit performs feature classification on these image features, obtaining the artifact-removed MRA image corresponding to the MRA image containing the artifacts to be removed.

[0094] The technical solution of this embodiment involves inputting the magnetic resonance image of the artifact to be removed into a feature extraction unit to obtain the image features corresponding to the artifact to be removed magnetic resonance image. The image features are obtained by the feature extraction unit through feature extraction of the magnetic resonance image of the artifact to be removed. The image features are then input into a fully connected unit to obtain the artifact-removed magnetic resonance image. The artifact-removed magnetic resonance image is obtained by the fully connected unit through feature classification of the image features. In this way, by using a trained image artifact removal model to extract and classify features of the magnetic resonance image of the artifact to be removed, the method of combining deep learning is realized to solve the artifact problem in magnetic resonance images. The artifact-removed magnetic resonance image can be obtained directly and accurately without the need for an additional reference scan of the target object to obtain the artifact-removed magnetic resonance image, reducing scanning time and efficiently acquiring the artifact-removed magnetic resonance image.

[0095] In another embodiment, such as Figure 4 As shown, a medical image processing method is provided. Taking the application of this method to a terminal as an example, it includes the following steps:

[0096] Step S410: Obtain the sample artifact image set.

[0097] Step S420: Determine the template magnetic resonance image corresponding to each scanned sample by using the reference echo scanning method.

[0098] Step S430: Train the neural network to be trained based on the sample artifact images and the corresponding template magnetic resonance images of each scanned sample to obtain a trained image artifact elimination model.

[0099] Step S440: Obtain the magnetic resonance image of the artifact to be eliminated.

[0100] Step S450: Input the magnetic resonance image with artifacts to be removed into the trained image artifact removal model to obtain the magnetic resonance image after artifact removal.

[0101] It should be noted that the specific limitations of the above steps can be found in the specific limitations of a medical image processing method described above.

[0102] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0103] Based on the same inventive concept, this application also provides a medical image processing apparatus for implementing the aforementioned medical image processing method. The solution provided by this apparatus is similar to the implementation described in the above method; therefore, the specific limitations in the one or more medical image processing apparatus embodiments provided below can be found in the above-described limitations of the medical image processing method, and will not be repeated here.

[0104] In one embodiment, such as Figure 5 As shown, a medical image processing device is provided, including: an acquisition module 510, a determination module 520, and a training module 530, wherein:

[0105] The acquisition module 510 is used to acquire a set of sample artifact images; each sample artifact image in the set of sample artifact images is obtained by magnetic resonance scanning of each scanned sample using an echo imaging sequence.

[0106] The determination module 520 is used to determine the template magnetic resonance image corresponding to each scanned sample by means of a reference echo scanning method.

[0107] The training module 530 is used to train the neural network to be trained based on the sample artifact images and the corresponding template magnetic resonance images of each scanned sample, so as to obtain a trained image artifact elimination model; the image artifact elimination model is used to perform peak correction on the magnetic resonance image to be eliminated.

[0108] In one embodiment, the determining module 520 is specifically configured to, for any scan sample, acquire the k-space data to be corrected and the target reference echo data corresponding to the scan sample under the reference scan sequence; establish a phase compensation matrix based on the phase difference between each echo in the target reference echo data and the first echo; perform peak correction on the k-space data to be corrected corresponding to the scan sample based on the phase compensation matrix to obtain peak-corrected k-space data; and reconstruct the peak-corrected k-space data to obtain the template magnetic resonance image corresponding to the scan sample.

[0109] In one embodiment, the determining module 520 is specifically used to correct the peak values ​​corresponding to odd echoes and even echoes in the k-space data to be corrected according to the phase compensation matrix, so as to obtain the peak-corrected k-space data.

[0110] In one embodiment, the determining module 520 is specifically used to correct the half-width at half maximum (WHM) of odd echoes and the half-width at half maximum (WHM) of even echoes in the k-space data to be corrected according to the phase compensation matrix, so as to obtain the peak-corrected k-space data.

[0111] In one embodiment, the determining module 520 is specifically used to correct the spectral positions corresponding to odd echoes and even echoes in the k-space data to be corrected according to the phase compensation matrix, so as to obtain the peak-corrected k-space data.

[0112] In one embodiment, the neural network is a convolutional neural network; the convolutional neural network includes a feature extraction unit and a fully connected unit.

[0113] In another embodiment, such as Figure 6 As shown, a medical image processing device is provided, including: an acquisition module 610 and an input module 620, wherein:

[0114] The acquisition module 610 is used to acquire the magnetic resonance image of the artifact to be eliminated and the trained image artifact elimination model; the magnetic resonance image of the artifact to be eliminated is obtained by performing magnetic resonance scanning on the target object using an echo imaging sequence.

[0115] The input module 620 is used to input the magnetic resonance image of the artifact to be eliminated into the trained image artifact elimination model to obtain the artifact-eliminated magnetic resonance image; the artifact-eliminated magnetic resonance image is obtained by peak correction of the magnetic resonance image of the artifact to be eliminated by the trained image artifact elimination model.

[0116] In one embodiment, the trained image artifact removal model includes a feature extraction unit and a fully connected unit; the input module 620 is specifically used to input the magnetic resonance image of the artifact to be removed into the feature extraction unit to obtain image features corresponding to the magnetic resonance image of the artifact to be removed; the image features are obtained by the feature extraction unit from the magnetic resonance image of the artifact to be removed; the image features are input into the fully connected unit to obtain the magnetic resonance image after artifact removal; the magnetic resonance image after artifact removal is obtained by the fully connected unit from the feature classification of the image features.

[0117] The various modules in the aforementioned medical image processing device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0118] In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 7 As shown, the computer device includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements a medical image processing method. The display unit is used to form a visually visible image and can be a display screen, projection device, or virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0119] Those skilled in the art will understand that Figure 7The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0120] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0121] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above method embodiments.

[0122] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0123] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data shall comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0124] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0125] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0126] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A medical image processing method, characterized in that, The method includes: A sample artifact image set is obtained; each sample artifact image in the sample artifact image set is obtained by performing magnetic resonance scanning on each scanned sample using an echo imaging sequence; The template magnetic resonance image corresponding to each scanned sample is determined by a reference echo scanning method, including: for any scanned sample, acquiring the k-space data to be corrected and the target reference echo data corresponding to the scanned sample under a reference scanning sequence; establishing a phase compensation matrix based on the phase difference between each echo in the target reference echo data and the first echo; performing peak correction on the k-space data to be corrected corresponding to the scanned sample based on the phase compensation matrix to obtain peak-corrected k-space data; and reconstructing the peak-corrected k-space data to obtain the template magnetic resonance image corresponding to the scanned sample. The neural network to be trained is trained based on the sample artifact images and the corresponding template magnetic resonance images of each scanned sample to obtain a trained image artifact elimination model; the image artifact elimination model is used to perform peak correction on the magnetic resonance images with artifacts to be eliminated.

2. The method according to claim 1, characterized in that, The step of performing peak correction on the k-space data to be corrected corresponding to any scan sample according to the phase compensation matrix to obtain peak-corrected k-space data includes: Based on the phase compensation matrix, the peak values ​​corresponding to odd echoes and even echoes in the k-space data to be corrected are corrected to obtain the peak-corrected k-space data.

3. The method according to claim 1, characterized in that, The step of performing peak correction on the k-space data to be corrected corresponding to any scan sample according to the phase compensation matrix to obtain peak-corrected k-space data includes: Based on the phase compensation matrix, the half-width at half maximum (WHM) corresponding to odd echoes and the half-width at half maximum (WHM) corresponding to even echoes in the k-space data to be corrected are corrected to obtain the peak-corrected k-space data.

4. The method according to claim 1, characterized in that, The step of performing peak correction on the k-space data to be corrected corresponding to any scan sample according to the phase compensation matrix to obtain peak-corrected k-space data includes: Based on the phase compensation matrix, the spectral positions corresponding to odd echoes and even echoes in the k-space data to be corrected are corrected to obtain the peak-corrected k-space data.

5. The method according to claim 1, characterized in that, The neural network is a convolutional neural network; the convolutional neural network includes a feature extraction unit and a fully connected unit.

6. A medical image processing method, characterized in that, The method includes: The method involves acquiring a magnetic resonance image of the object to be de-artifacted and a trained image artifact removal model; the magnetic resonance image of the object to be de-artifacted is obtained by performing magnetic resonance scanning on the target object using an echo imaging sequence; and the trained image artifact removal model is trained according to the medical image processing method as described in any one of claims 1 to 5. The magnetic resonance image of the artifact to be eliminated is input into the trained image artifact elimination model to obtain the artifact-eliminated magnetic resonance image; the artifact-eliminated magnetic resonance image is obtained by peak correction of the magnetic resonance image of the artifact to be eliminated by the trained image artifact elimination model.

7. The method according to claim 6, characterized in that, The trained image artifact removal model includes a feature extraction unit and a fully connected unit; the step of inputting the magnetic resonance image with artifacts to be removed into the trained image artifact removal model to obtain the artifact-removed magnetic resonance image includes: The magnetic resonance image of the artifact to be removed is input into the feature extraction unit to obtain the image features corresponding to the magnetic resonance image of the artifact to be removed; the image features are obtained by the feature extraction unit from the magnetic resonance image of the artifact to be removed. The image features are input into the fully connected unit to obtain the artifact-free magnetic resonance image; the artifact-free magnetic resonance image is obtained by the fully connected unit performing feature classification on the image features.

8. A medical image processing device, characterized in that, The device includes: The acquisition module is used to acquire a set of sample artifact images; each sample artifact image in the set is obtained by performing magnetic resonance scanning on each scanned sample using an echo imaging sequence. The determination module is used to determine the template magnetic resonance image corresponding to each scanned sample by means of a reference echo scanning method; The determining module is specifically configured to, for any one of the scanned samples, acquire the k-space data to be corrected and the target reference echo data corresponding to the scanned sample under the reference scan sequence; establish a phase compensation matrix based on the phase difference between each echo in the target reference echo data and the first echo; perform peak correction on the k-space data to be corrected corresponding to the scanned sample based on the phase compensation matrix to obtain peak-corrected k-space data; and reconstruct the peak-corrected k-space data to obtain the template magnetic resonance image corresponding to the scanned sample. The training module is used to train the neural network to be trained based on the sample artifact images and the corresponding template magnetic resonance images of each scanned sample, so as to obtain a trained image artifact removal model; the image artifact removal model is used to perform peak correction on the magnetic resonance image to be removed from the artifacts.

9. A medical image processing device, characterized in that, The device includes: An acquisition module is used to acquire the magnetic resonance image of the object to be de-artifacted and a trained image artifact removal model; the magnetic resonance image of the object to be de-artifacted is obtained by performing magnetic resonance scanning on the target object using an echo imaging sequence; the trained image artifact removal model is trained according to the medical image processing method as described in any one of claims 1 to 5. The input module is used to input the magnetic resonance image of the artifact to be eliminated into the trained image artifact elimination model to obtain the artifact-eliminated magnetic resonance image; the artifact-eliminated magnetic resonance image is obtained by peak correction of the magnetic resonance image of the artifact to be eliminated by the trained image artifact elimination model.

Images