Reconstruction method for breaking the quality limit of low-field magnetic resonance imaging

Abstract

The application discloses a reconstruction method for breaking through the quality limit of low-field magnetic resonance imaging, and comprises the following steps: acquisition and construction of a low-field data set; acquisition and construction of a high-field data set; training of a basic super-resolution model from a low-field, low-resolution image of 1~Nth excitation to a high-resolution image of Nth excitation of the low-field; comparison of the influence of different excitation times on the performance of the basic super-resolution model; training of a cross-field super-resolution model from the high-resolution image of Nth excitation of the low-field to a super-high-resolution image of the high-field; cascading of the basic super-resolution model and the cross-field super-resolution model to construct an enhanced two-stage cross-field super-resolution model; and reconstruction of the high-field, super-high-resolution image from the low-field, low-excitation-time, low-resolution image through the enhanced two-stage cross-field super-resolution model. The application enables the low-field magnetic resonance scanner to realize high-quality imaging breaking through the magnetic field strength constraint in a relatively short scanning time.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of medical imaging, and particularly relates to a reconstruction method for breaking through the quality limit of low-field magnetic resonance imaging. BACKGROUND

[0002] Low-field (<0.3 T) magnetic resonance scanners have the advantages of low cost, high flexibility and the ability to provide point-of-care imaging, and have attracted widespread attention in the field of medical imaging. However, due to physical limitations, low-field magnetic resonance scanners cannot obtain high-quality images within an acceptable scan time, which seriously affects the realization of advanced functions and accurate diagnosis.

[0003] Existing basic super-resolution methods use low-resolution images to reconstruct high-resolution images, thereby improving the acquisition speed of low-field imaging. However, the images reconstructed by this method do not exceed the constraints of the magnetic field strength, i.e., they cannot obtain images with high-field effects. In addition, some cross-field super-resolution methods use high-field images to construct a data set and train a super-resolution model to reconstruct high-resolution images with high signal-to-noise ratio from low-resolution low-field images. The signal-to-noise ratio of the images reconstructed by this method is significantly improved, but the improvement in image resolution is not obvious, and the spatial resolution cannot reach the level of high-field.

[0004] Therefore, it is necessary to propose a reconstruction method for breaking through the quality limit of low-field magnetic resonance imaging, which uses low-field, low-resolution images obtained in a short scan time to reconstruct images with a resolution close to that of high-field, super-high resolution. SUMMARY

[0005] In view of the above-mentioned defects or shortcomings in the prior art, the present application provides a reconstruction method for breaking through the quality limit of low-field magnetic resonance imaging, which trains a basic super-resolution model and a cross-field super-resolution model using a low-field data set and a high-field super-high resolution data set, respectively, then cascades the two models to construct an enhanced two-stage cross-field super-resolution model, and finally reconstructs high-quality images with a resolution exceeding the limit of the low-field magnetic field strength from low-field, low-resolution images through the enhanced two-stage cross-field super-resolution model.

[0006] To achieve the above-mentioned purpose, the present application adopts the following technical solutions:

[0007] A reconstruction method for breaking through the quality limit of low-field magnetic resonance imaging, comprising the following steps:

[0008] Step 1: acquisition and construction of low-field dataset, comprising,

[0009] Step 1.1: acquiring low-resolution images of a target object under low-field conditions, respectively, by 1~N excitations, N≥4, N being a natural number;

[0010] Step 1.2: acquiring a high-resolution image of the target object under low-field conditions by N excitations;

[0011] Step 1.3: constructing data pairs of the low-resolution images under low-field conditions by 1~N excitations and the high-resolution image under low-field conditions by N excitations, respectively;

[0012] Step 2: acquisition and construction of high-field dataset, comprising,

[0013] Step 2.1: acquiring a super-high-resolution image of the target object under high-field conditions by 1 excitation;

[0014] Step 2.2: performing degradation processing on the super-high-resolution image to obtain a high-field degraded image;

[0015] Step 2.3: constructing a data pair of the high-field degraded image and the super-high-resolution image;

[0016] Step 3: training a basic super-resolution model from the low-resolution images under low-field conditions by 1~N excitations to the high-resolution image under low-field conditions by N excitations, respectively;

[0017] Step 4: comparing the influence of different excitation times on the performance of the basic super-resolution model, comprising,

[0018] Step 4.1: comparing the image quality difference between the basic super-resolution results of the low-resolution images of the target object under low-field conditions by 1~N excitations and the high-resolution image of the target object under low-field conditions by N excitations;

[0019] Step 4.2: obtaining the minimum excitation times of the low-resolution images that can obtain the quality close to the high-resolution image quality through the basic super-resolution model by visual effect evaluation;

[0020] Step 5: training a cross-field super-resolution model from the high-resolution image under low-field conditions by N excitations to the super-high-resolution image under high-field conditions;

[0021] Step 6: cascading the basic super-resolution model and the cross-field super-resolution model to construct an enhanced two-stage cross-field super-resolution model;

[0022] Step 7: reconstructing the high-field, super-high-resolution image from the low-field, low-excitation, low-resolution image through the enhanced two-stage cross-field super-resolution model, the low excitation times ≤N-1;

[0023] Step 8: End.

[0024] Further, the low-resolution image and the high-resolution image in step 1 are from multiple scans of the same target object at the same position; the data of each data pair are from different target objects.

[0025] Further, in step 2, the ultrahigh resolution refers to the resolution that cannot be obtained by a low-field magnetic resonance scanner within an acceptable scanning time, including acquisition resolution and spatial resolution; the high-field degraded image is obtained by down-sampling, adding Gaussian noise, and adding Rician noise from the ultrahigh resolution image.

[0026] Further, the basic super-resolution model and the cross-field super-resolution model are both 2-fold super-resolution models.

[0027] Further, in step 4, the minimum number of excitations refers to the number of excitations that can obtain a low-resolution image with the same quality as the high-resolution image through a basic super-resolution model.

[0028] Further, in step 6, the first stage of the enhanced two-stage cross-field super-resolution model is a basic super-resolution model, and the second stage is a cross-field super-resolution model; the input of the basic super-resolution model is a low-field, minimum number of excitations low-resolution image, and the output is an intermediate feature of the enhanced two-stage cross-field super-resolution model; the input of the cross-field super-resolution model is the intermediate feature, and the output is the output of the enhanced two-stage cross-field super-resolution model.

[0029] Further, in step 7, the image reconstructed by the enhanced two-stage cross-field super-resolution model has a quality exceeding the constraint of the low-field magnetic field strength, and has an effect close to the high-field, ultrahigh resolution image, including contrast, signal-to-noise ratio, and spatial resolution.

[0030] The beneficial effects of the present application are:

[0031] First, in the present application, during low-field data acquisition, each data pair is a multiple scan of the same subject at the same position, so that the basic super-resolution model can learn the mapping relationship between the low-field, low-resolution image with low number of excitations and the low-field, high-resolution image with multiple number of excitations under real scanning conditions.

[0032] Second, while ensuring the performance of the low-field basic super-resolution, the present application reduces the number of excitations of the low-field, low-resolution image, and further improves the imaging speed of obtaining high-quality low-field high-resolution images.

[0033] Thirdly, the present application uses high-field super-high-resolution data to construct cross-field super-resolution data sets, so that the cross-field super-resolution model can reconstruct high-field super-high-resolution effect images from low-field high-resolution images.

[0034] Fourthly, the present application cascades the basic super-resolution model and the cross-field super-resolution model and constructs an enhanced two-stage cross-field super-resolution model, so that the reconstructed image can first reach the limit level of low-field imaging quality, and then further reach the level of high-field high-quality imaging. BRIEF DESCRIPTION OF DRAWINGS

[0035] Figure 1 The flow chart of the reconstruction method of the present application for breaking through the limit of low-field magnetic resonance imaging quality;

[0036] Figure 2 The schematic diagram of the present application for training the basic super-resolution model and the cross-field super-resolution model;

[0037] Figure 3 The comparison diagram of the basic super-resolution image of low-field, 1~4 excitation and the high-resolution image of low-field, 4 excitation of the present application;

[0038] Figure 4 The schematic diagram of the enhanced two-stage cross-field super-resolution model of the present application;

[0039] Figure 5 The comparison diagram of the effect of the enhanced two-stage cross-field super-resolution model of the present application at each stage;

[0040] Figure 6 The comparison diagram of the application effect of the embodiment of the present application in the 0.2 T low-field scanner and the high-field super-high-resolution image. DETAILED DESCRIPTION

[0041] In order to make the purpose and advantages of the present application more clear and obvious, the present application is further described below in combination with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present application, and are not used to limit the protection scope of the present application.

[0042] The preferred embodiments of the present application are described below with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are only used to explain the technical principles of the present application, and are not used to limit the protection scope of the present application.

[0043] As Figure 1The flow chart shows a reconstruction method for breaking the quality limit of low-field magnetic resonance imaging. In the following, high excitation number refers to the excitation number that can obtain better image quality, such as 4 times. In the basic super-resolution of the present application, the image of high excitation number is set as the reference image. As mentioned before, the high excitation number N is not limited to a specific number, and in the following, N = 4 is taken as an example for specific description. Specifically, the method comprises the following steps:

[0044] Step 1: Acquisition and construction of low-field data set, comprising,

[0045] Step 1.1: Acquisition of low-field, low-resolution images of 1-4 excitations;

[0046] Step 1.2: Acquisition of low-field, high-resolution images of 4 excitations;

[0047] Step 1.3: Construction of data pairs of low-field, low-resolution images of 1-4 excitations and low-field, high-resolution images of 4 excitations. Respectively, data pairs of low-field, low-resolution images of 1 excitation and low-field, high-resolution images of 4 excitations, data pairs of low-field, low-resolution images of 2 excitations and low-field, high-resolution images of 4 excitations, data pairs of low-field, low-resolution images of 3 excitations and low-field, high-resolution images of 4 excitations, data pairs of low-field, low-resolution images of 4 excitations and low-field, high-resolution images of 4 excitations.

[0048] Acquisition of low-field data set of multiple subjects, for each subject, continuously acquire low-resolution images of 1-4 excitations and high-resolution images of 4 excitations at the same imaging orientation. Among them, the low-field low-resolution is 128x128 acquisition resolution, 1.87 mm 2 spatial resolution; the low-field high-resolution is 256x256 acquisition resolution, 0.93 mm 2 spatial resolution.

[0049] Step 2: Acquisition and construction of high-field data set, comprising,

[0050] Step 2.1: Under high-field conditions, acquire super-high-resolution images of the target object excited once;

[0051] Acquisition of high-field data set of multiple subjects, for each subject, acquire super-high-resolution images at the same imaging orientation as the low-field data set, with an excitation number of 1. Among them, super-high-resolution refers to the resolution that the low-field magnetic resonance scanner cannot obtain within an acceptable scan time, including acquisition resolution and spatial resolution; the high-field super-high-resolution is 512x512 acquisition resolution, 0.45 mm 2 spatial resolution.

[0052] The data pairs in the entire low-field dataset are divided into a training set, a validation set and a test set in a ratio of 8:1:1.

[0053] Step 2.2: Degenerate the super-high-resolution image to obtain a high-field degenerate image.

[0054] First, the resolution of the high-field super-resolution image is adjusted multiple times by bicubic interpolation downsampling method: 512x512→170x170→512x512→256x256, so that the image is visually close to the resolution of the low-field high-resolution image; then, the adjusted image is normalized; finally, the normalized image is added with Gaussian noise with a standard deviation of 0.005 and Rayleigh noise with a standard deviation of 0.002, and then normalized again to obtain the final high-field degenerate image.

[0055] Step 2.3: Construct the data pairs of the high-field degenerate image and the super-high-resolution image.

[0056] The data pairs in the entire high-field dataset are divided into a training set, a validation set and a test set in a ratio of 8:1:1.

[0057] Step 3: Train the basic super-resolution model from the low-field, 1~4 excitation low-resolution image to the low-field, 4 excitation high-resolution image, respectively.

[0058] As shown in FIG. 2, the upper graph is a schematic diagram of training the basic super-resolution model from the low-field, 1~4 excitation low-resolution image to the low-field, 4 excitation high-resolution image. The basic super-resolution model from the low-field, 1~4 excitation low-resolution image to the high-field, 4 excitation image is trained respectively, and then four basic super-resolution models with the best performance corresponding to different excitation times are obtained.

[0059] Step 4: Compare the influence of different excitation times on the performance of the basic super-resolution model, including,

[0060] Step 4.1: Compare the image quality difference between the basic super-resolution result of the target object being excited by 1~4 excitation low-resolution images under low-field conditions and the target object being excited by 4 excitation high-resolution images under low-field conditions.

[0061] Step 4.2: Obtain the minimum excitation times of the low-resolution image that can obtain the low-resolution image close to the high-resolution image quality through the basic super-resolution by visual effect evaluation.

[0062] The basic super-resolution result of the low-field, 1~4 excitation low-resolution image is evaluated by actual observation. As shown in FIG. 3, the upper graph is the basic super-resolution result of the low-field, 1 excitation low-resolution image, and the lower graph is the basic super-resolution result of the low-field, 4 excitation low-resolution image. Figure 3The white arrows in the figure point to the part of the details difference. It can be seen that the image quality of the third and fourth images is close to that of the fifth image, and there is no loss of detail information. There is a blurring effect in the first and second images, and there is a loss of detail information to varying degrees. Therefore, the minimum number of excitations of the low-field, low-resolution image that can obtain an image quality close to that of the low-field, 4-excitation high-resolution image through the basic super-resolution is 3.

[0063] Step 5: Training the cross-field super-resolution model from the low-field, 4-excitation high-resolution image to the high-field, super-high-resolution image.

[0064] As Figure 2 shown, it is a schematic diagram for training the cross-field super-resolution model from the high-field degraded image to the high-field, super-high-resolution image. In which, the low-field, 4-excitation high-resolution image is simulated by the high-field degraded image. By training the cross-field super-resolution model from the high-field degraded image to the high-field, super-high-resolution image, the cross-field super-resolution model with the best performance is obtained.

[0065] Step 6: Cascade the basic super-resolution model and the cross-field super-resolution model to construct an enhanced two-stage cross-field super-resolution model.

[0066] As shown in FIG. 4, the basic super-resolution model and the cross-field super-resolution model are cascaded by merging the input and output to construct an enhanced two-stage cross-field super-resolution model. In which, the first stage of the enhanced two-stage cross-field super-resolution model is the basic super-resolution model, and the second stage is the cross-field super-resolution model. The input of the basic super-resolution model is the low-field, low-resolution image with the minimum number of excitations, and the output is the intermediate feature of the enhanced two-stage cross-field super-resolution model; the input of the cross-field super-resolution model is the intermediate feature, and the output is the output of the enhanced two-stage cross-field super-resolution model.

[0067] Step 7: Reconstruct the high-field, super-high-resolution image from the low-field, low-excitation, low-resolution image through the enhanced two-stage cross-field super-resolution model.

[0068] The low-field, 3-excitation low-resolution image is input into the enhanced two-stage cross-field super-resolution model, and the intermediate feature close to the effect of the low-field, 4-excitation high-resolution image is obtained. The enhanced two-stage cross-field super-resolution model further processes the intermediate feature to output the final reconstructed image close to the effect of the high-field, super-high-resolution image.

[0069] As shown in FIG. 5, the first row shows the low-field, 3-shot low-resolution image and its detail image, the second row shows the intermediate feature output by the enhanced two-stage cross-field super-resolution model and its detail image, and the third row shows the final image output by the enhanced two-stage cross-field super-resolution model and its detail image. It can be observed that the image quality is significantly improved in resolution and signal-to-noise ratio from the first row to the third row.

[0070] Step 8: End the design.

[0071] In one embodiment, based on the present application, low-field data is collected and a dataset is constructed on a 0.2 T low-field mobile scanner, and high-field data is collected and a dataset is constructed on a 1.5 T high-field scanner. First, the enhanced two-stage cross-field super-resolution model is obtained. Then, a 0.2 T low-field, 3-shot, 128x128 resolution, 1.87 mm 2 spatial resolution imaging experiment is performed, and the single orientation scan time is 3.3 minutes. Finally, the image is input into the enhanced two-stage cross-field super-resolution model to obtain a reconstructed image close to the high-field, ultra-high resolution effect. As shown in FIG. 6, from left to right, the first image is a low-field, 3-shot, 128x128 resolution, 1.87 mm 2 spatial resolution image collected on a 0.2 T low-field scanner and its detail image, the second image is an intermediate feature output by the enhanced two-stage cross-field super-resolution model and its detail image, the third image is a final image output by the enhanced two-stage cross-field super-resolution model and its detail image, and the fourth image is a high-field, 1-shot, 512x512 resolution, 0.45 mm 2 spatial resolution image collected on a 1.5 T high-field scanner and its detail image. In the first image, the image quality is poor and lacks some tissue structure information. In the second image, the image quality is improved to a certain extent, and the tissue structure information is complete, and the image quality reaches the limit level that can be obtained by the 0.2 T scanner within an acceptable scan time. In the third image, the image quality is significantly improved, the tissue structure information is rich, the texture features are obvious, the spatial resolution is high, and the signal-to-noise ratio is high. Compared with the fourth image, the image quality of the third image reaches the level of the 1.5 T high-field ultra-high resolution image.

[0072] So far, the technical solutions of the present application have been described in combination with the preferred embodiments shown in the drawings, but those skilled in the art can easily understand that the protection scope of the present application is obviously not limited to these specific embodiments. Those skilled in the art can make equivalent changes or replacements to the related technical features without departing from the principles of the present application, and the technical solutions after the changes or replacements will all fall within the protection scope of the present application.

[0073] The above-described specific embodiments further illustrate the purpose, technical solutions and beneficial effects of the present application, and it should be understood that the above-described is only a specific embodiment of the present application and is not intended to limit the present application, and any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. A reconstruction method to break the low-field magnetic resonance imaging quality limit, characterized in that, The method comprises the following steps: Step 1: acquisition and construction of low-field data set, comprising, Step 1.1: acquiring low-resolution images of a target object under low-field conditions, respectively, by 1~N excitations, wherein N≥4, and N is a natural number; Step 1.2: acquiring a high-resolution image of the target object under low-field conditions by N excitations; Step 1.3: constructing data pairs of low-resolution images under low-field conditions by 1~N excitations and the high-resolution image under low-field conditions by N excitations, respectively; Step 2: acquisition and construction of high-field data set, comprising, Step 2.1: acquiring a super-high-resolution image of the target object under high-field conditions by 1 excitation; Step 2.2: performing degradation processing on the super-high-resolution image to obtain a high-field degraded image; Step 2.3: constructing a data pair of the high-field degraded image and the super-high-resolution image; Step 3: training a basic super-resolution model from low-field, low-resolution images by 1~N excitations to low-field, high-resolution images by N excitations, respectively; Step 4: comparing the influence of different excitation times on the performance of the basic super-resolution model, comprising, Step 4.1: comparing the image quality difference between the basic super-resolution results of low-resolution images of the target object under low-field conditions by 1~N excitations and the high-resolution image of the target object under low-field conditions by N excitations; Step 4.2: obtaining the minimum excitation times of the low-resolution images that can obtain the quality close to the high-resolution image quality through the basic super-resolution model by visual effect evaluation; Step 5: training a cross-field super-resolution model from low-field, high-resolution images by N excitations to high-field, super-high-resolution images; Step 6: cascading the basic super-resolution model and the cross-field super-resolution model to construct an enhanced two-stage cross-field super-resolution model; Step 7: reconstructing a high-field, super-high-resolution image from a low-field, low-resolution image with low excitation times through the enhanced two-stage cross-field super-resolution model, wherein the low excitation times≤N-1; Step 8: ending.

2. The reconstruction method of breaking the low-field magnetic resonance imaging quality limit according to claim 1, characterized in that, The low-resolution images and the high-resolution images in the step 1 come from multiple scans of the same target object at the same position; the data of each data pair come from different target objects.

3. The reconstruction method of breaking the low-field magnetic resonance imaging quality limit according to claim 1, characterized in that, In the step 2, the super-high resolution refers to the resolution that cannot be obtained by a low-field magnetic resonance scanner within an acceptable scan time, including acquisition resolution and spatial resolution; the high-field degraded image is obtained from the super-high-resolution image by downsampling, adding Gaussian noise, and adding Rician noise.

4. The reconstruction method of breaking the low-field magnetic resonance imaging quality limit according to claim 1, characterized in that, The basic super-resolution model and the cross-field super-resolution model are both 2-fold super-resolution models.

5. The reconstruction method of breaking the low-field magnetic resonance imaging quality limit according to claim 1, characterized in that, The minimum excitation times in the step 4 refer to the excitation times of the low-resolution images that can obtain the same quality as the high-resolution image through the basic super-resolution model.

6. The reconstruction method of breaking the low-field magnetic resonance imaging quality limit according to claim 1, characterized in that, The first stage of the enhanced two-stage cross-field super-resolution model in step 6 is a basic super-resolution model, and the second stage is a cross-field super-resolution model; the input of the basic super-resolution model is a low-field, low-resolution image with the minimum number of excitations, and the output is an intermediate feature of the enhanced two-stage cross-field super-resolution model; the input of the cross-field super-resolution model is the intermediate feature, and the output is the output of the enhanced two-stage cross-field super-resolution model.

7. The reconstruction method of breaking the low-field magnetic resonance imaging quality limit according to claim 1, characterized in that, The image reconstructed by the enhanced two-stage cross-field super-resolution model in step 7 has a quality exceeding the low-field magnetic field strength constraint and has an effect close to the high-field, super-high-resolution image, including contrast, signal-to-noise ratio, and spatial resolution.

Images