Method, device and product for synthesizing CT image into MRI image based on deep learning
Through a deep learning-based method, using CT images to generate MRI images, solving the problem of simple CT image inspection and cumbersome MRI image inspection, improving the flexibility and experience of use, and reducing costs.
Patent Information
- Application Number
- CN202411782173.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-05
- Publication Date
- 2025-03-28
AI Technical Summary
In the prior art, CT image inspection is relatively simple, while MRI image inspection is cumbersome and time-consuming, resulting in poor use flexibility, high cost and poor user experience.
Using a deep learning-based method, an MRI image synthesis model is constructed by acquiring and processing the training data sets of CT images and MRI images, thereby converting the CT images into MRI images.
It is realized that MRI images are generated through CT images when MRI devices are lacking or MRI scans are not possible, thereby improving usage flexibility, reducing costs, and improving patient experience.
Smart Images
Figure CN119722483A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of image processing technology, and specifically relates to a method, device and product for synthesizing MRI images from CT images based on deep learning. Background Art
[0002] With the rapid iterative development of medical imaging technology, the continuous breakthroughs in computer-aided diagnosis technology, and the in-depth application of artificial intelligence technology in the medical field, multimodal medical imaging technology is playing an increasingly important role in clinical diagnosis, treatment planning, and prognosis assessment. Among them, Computed Tomography (CT) and Magnetic Resonance Imaging (MRI) are the two most important medical imaging examination methods, each with its own characteristics and complementing each other.
[0003] However, the equipment required for MRI is relatively complex and expensive. Therefore, it not only limits the application of MRI in medical institutions that lack MRI equipment or cannot perform MRI scans, resulting in poor flexibility in use, but also increases the cost of use. At the same time, the examination process of MRI is relatively cumbersome compared to CT imaging, and each time it takes a long time (CT imaging generally does not exceed 5 minutes, while the time required for MRI is often between 15 and 30 minutes). In this way, patients need to wait for a long time, which reduces the patient's user experience. Therefore, based on the aforementioned deficiencies, how to provide a method for synthesizing MRI images from CT images to improve the flexibility and user experience, as well as reduce costs, has become a problem that needs to be solved urgently. Summary of the invention
[0004] The purpose of the present invention is to provide a method, device and product for synthesizing MRI images from CT images based on deep learning, so as to solve the problems of poor flexibility, high cost and poor user experience in the prior art.
[0005] In order to achieve the above object, the present invention adopts the following technical solutions:
[0006] In a first aspect, a method for synthesizing an MRI image from a CT image based on deep learning is provided, comprising:
[0007] Acquire a training data set, wherein the training data set includes a plurality of sample CT images and a label image of each sample CT image, and the label image of any sample CT image is an MRI image that matches the any sample CT image;
[0008] Constructing an initial MRI image synthesis model, and using each sample CT image in the training data set and a label image corresponding to each sample CT image to train the initial MRI image synthesis model, so as to obtain an MRI image synthesis model after the training is completed;
[0009] A target CT image is acquired, and the target CT image is input into the MRI image synthesis model to obtain a synthetic MRI image corresponding to the target CT image.
[0010] Based on the above disclosed content, the present invention first obtains several sample CT images and MRI images matching each sample CT image, and then uses the aforementioned data to form a training data set; then, an initial MRI image synthesis model is constructed, and the aforementioned training data is used to train the initial MRI image synthesis model, so that after the training is completed, the final MRI image synthesis model is obtained; in this way, the present invention is equivalent to constructing a neural network model for generating MRI images based on CT images; based on this, in actual use, it is only necessary to input the target CT image into the aforementioned MRI image synthesis model to obtain a synthetic MRI image corresponding to the target CT image.
[0011] Through the above design, the present invention utilizes CT images and adopts deep learning technology to generate MRI images; in this way, patients only need to undergo a simple CT examination to obtain nuclear magnetic resonance images without using nuclear magnetic resonance equipment; based on this, not only can the application of nuclear magnetic resonance be realized when there is a lack of MRI equipment or MRI scanning cannot be performed, thereby greatly improving the flexibility of use and reducing the cost of use; at the same time, there is no need for patients to wait for a long time, thereby improving the patient's user experience; therefore, the present invention is very suitable for large-scale application and promotion.
[0012] In one possible design, a training data set is obtained, including:
[0013] Acquire an original data set, wherein the original data set includes a plurality of original sample CT images and a plurality of original MRI images;
[0014] Performing image registration processing on a plurality of original sample CT images and a plurality of original MRI images in the original data set, so as to obtain an original MRI image matching each original sample CT image after the image registration processing;
[0015] Performing image enhancement processing on each original sample CT image and each original MRI image matching the original sample CT image, so as to obtain a plurality of enhanced original sample CT images and an enhanced original MRI image matching each enhanced original sample CT image after the image enhancement processing;
[0016] Performing denoising processing on each enhanced original sample CT image and each enhanced original MRI image matched with the enhanced original sample CT image, so as to obtain a plurality of sample CT images and a plurality of MRI images after the denoising processing;
[0017] The MRI image matched with each sample CT image is used as the label image of each sample CT image;
[0018] The training data set is composed of each sample CT image and a label image corresponding to each sample CT image.
[0019] In a possible design, a feature-based image registration algorithm or a pixel-based image registration algorithm is used to perform image registration processing on a plurality of original sample CT images and a plurality of original MRI images.
[0020] In a possible design, each sample CT image in the training data set and a label image corresponding to each sample CT image are used to train the initial MRI image synthesis model, including:
[0021] The initial MRI image synthesis model is trained by taking each sample CT image in the training data set as input and the synthetic MRI image of each sample CT image as output, and during the training process, the synthetic MRI image of each sample CT image and the label image of each sample CT image are used to adjust the model parameters of the initial MRI image synthesis model, so as to obtain the MRI image synthesis model after the training is completed.
[0022] In a possible design, the model parameters of the initial MRI image synthesis model are adjusted by using the synthetic MRI images of each sample CT image and the label images of each sample CT image, including:
[0023] Performing consistency comparison processing on the synthetic MRI image of each sample CT image and the label image corresponding to each sample CT image, so as to obtain a consistency comparison result after the consistency comparison processing;
[0024] The model parameters of the initial MRI image synthesis model are adjusted according to the consistency comparison result.
[0025] In one possible design, the initial MRI image synthesis model includes a DRU-Net neural network model.
[0026] In a possible design, the MRI image matched by any sample CT image is a T1WI image.
[0027] In a second aspect, a device for synthesizing MRI images from CT images based on deep learning is provided, comprising:
[0028] An acquisition unit, configured to acquire a training data set, wherein the training data set includes a plurality of sample CT images and a label image of each sample CT image, and the label image of any sample CT image is an MRI image that matches the any sample CT image;
[0029] A model building unit, used to build an initial MRI image synthesis model, and use each sample CT image in the training data set and a label image corresponding to each sample CT image to train the initial MRI image synthesis model, so as to obtain the MRI image synthesis model after the training is completed;
[0030] The image synthesis unit is used to obtain a target CT image and input the target CT image into the MRI image synthesis model to obtain a synthesized MRI image corresponding to the target CT image.
[0031] In a third aspect, another device for synthesizing MRI images from CT images based on deep learning is provided. Taking the device as an electronic device as an example, the device includes a memory, a processor and a transceiver that are communicatively connected in sequence, wherein the memory is used to store computer programs, the transceiver is used to send and receive messages, and the processor is used to read the computer program to execute the method for synthesizing MRI images from CT images based on deep learning as described in the first aspect or any possible design of the first aspect.
[0032] In a fourth aspect, a storage medium is provided, on which instructions are stored. When the instructions are executed on a computer, the method for synthesizing MRI images from CT images based on deep learning as described in the first aspect or any possible design of the first aspect is executed.
[0033] In a fifth aspect, a computer program product comprising instructions is provided, which, when executed on a computer, causes the computer to execute the method for synthesizing MRI images from CT images based on deep learning as described in the first aspect or any possible design of the first aspect.
[0034] Beneficial effects:
[0035] (1) The present invention utilizes CT images and adopts deep learning technology to generate MRI images. In this way, patients only need to undergo a simple CT examination to obtain nuclear magnetic resonance images without using nuclear magnetic resonance equipment. Based on this, not only can the application of nuclear magnetic resonance be realized when there is a lack of MRI equipment or when MRI scanning cannot be performed, thereby greatly improving the flexibility of use and reducing the cost of use. At the same time, there is no need for patients to wait for a long time, thereby improving the patient's user experience. Therefore, the present invention is very suitable for large-scale application and promotion. BRIEF DESCRIPTION OF THE DRAWINGS
[0036] Figure 1 A schematic diagram of the steps of a method for synthesizing MRI images from CT images based on deep learning provided by an embodiment of the present invention;
[0037] Figure 2 A schematic diagram of the structure of a device for synthesizing MRI images from CT images based on deep learning provided by an embodiment of the present invention;
[0038] Figure 3 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0039] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the present invention will be briefly introduced below in combination with the drawings and the description of the embodiments or the prior art. Obviously, the following description of the structure of the drawings is only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. It should be noted that the description of these embodiments is used to help understand the present invention, but does not constitute a limitation of the present invention.
[0040] It should be understood that although the terms first, second, etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are only used to distinguish one unit from another unit. For example, a first unit can be referred to as a second unit, and similarly, a second unit can be referred to as a first unit without departing from the scope of the exemplary embodiments of the present invention.
[0041] It should be understood that the term "and / or" that may appear in this article is merely a description of the association relationship between associated objects, indicating that three relationships may exist. For example, A and / or B can represent three situations: A exists alone, B exists alone, and A and B exist at the same time. The term " / and" that may appear in this article describes another type of association object relationship, indicating that two relationships may exist. For example, A / and B can represent two situations: A exists alone, and A and B exist alone. In addition, the character " / " that may appear in this article generally indicates that the previous and next associated objects are in an "or" relationship.
[0042] Example:
[0043] See also Figure 1As shown, the method for synthesizing MRI images from CT images based on deep learning provided in this embodiment first constructs an initial MRI image synthesis model; then, a number of sample CT images and MRI images corresponding to each sample CT image are used to train the initial MRI image synthesis model, so that after the training, an MRI image synthesis model is obtained; finally, the target CT image is input into the aforementioned trained model, and a synthetic MRI image corresponding to the target CT image can be obtained; in this way, the method uses CT images and adopts deep learning technology to generate MRI images, so that patients only need to undergo simple CT examinations to obtain nuclear magnetic resonance images without using nuclear magnetic resonance equipment; based on this, not only can the application of nuclear magnetic resonance be realized when there is a lack of MRI equipment or MRI scanning cannot be performed, thereby greatly improving the flexibility of use and reducing the cost of use; at the same time, there is no need for patients to wait for a long time, thereby improving the patient's use experience; therefore, the method is very suitable for large-scale application and promotion; among them, for example, the method can be but not limited to running on the image synthesis end side, and optionally, the image synthesis end can be but not limited to a personal computer (personal computer, PC), tablet computer or smart phone. It can be understood that the aforementioned execution subject does not constitute a limitation on the embodiments of the present application. Accordingly, the operation steps of the method can be but are not limited to the following steps S1 to S3.
[0044] S1. Obtain a training data set, wherein the training data set includes a number of sample CT images and a label image of each sample CT image, and the label image of any sample CT image is an MRI image that matches the any sample CT image; in a specific implementation, the MRI image that matches the any sample CT image is taken as an example, which may be, but is not limited to, a T1WI image; of course, different types of nuclear magnetic resonance images may also be selected according to actual use, and are not limited to the aforementioned examples.
[0045] Meanwhile, in this embodiment, the following steps S11 to S16 may be used, for example but not limited to, to construct the aforementioned training data set to provide a data basis for subsequent model training.
[0046] S11. Acquire an original data set, wherein the original data set includes a plurality of original sample CT images and a plurality of original MRI images; in specific applications, the plurality of original sample CT images may be, but are not limited to, CT images of different parts of different human bodies, while the original MRI images are obtained by performing nuclear magnetic resonance scanning on the same part of the aforementioned human bodies as the object to be examined; specifically, if an original sample CT image is a CT image of the brain, then its corresponding original MRI image is a nuclear magnetic resonance image of the brain of the same human body; optionally, any original MRI image may be, but is not limited to, a T1WI image; of course, a T2WI image or a PDWI (proton weighted) image may also be used according to actual needs, and no specific limitation is made here.
[0047] In this way, after the original data set is obtained, the aforementioned training data set can be constructed based on the original data set, and its construction can be but is not limited to as shown in the following steps S12 to S16.
[0048] S12. Perform image registration processing on several original sample CT images and several original MRI images in the original data set, so as to obtain an original MRI image that matches each original sample CT image after the image registration processing; in the specific implementation, since the collected original sample CT images and the corresponding original MRI images may not be completely matched, therefore, in order to ensure the accuracy of subsequent deep learning image synthesis, this embodiment needs to perform image registration processing on the aforementioned original sample CT images and original MRI images.
[0049] Among them, the main steps of the example image registration are: feature detection, feature matching, estimation of transformation model (i.e., calculating the parameters of the transformation function according to the matched feature points, usually including translation, rotation, scaling, etc.) and image resampling and transformation (i.e., resampling and transforming the image according to the calculated transformation function to align it with the reference image); further, the example can be but not limited to using a feature-based image registration algorithm or a pixel-based image registration algorithm to perform image registration processing on several original sample CT images and several original MRI images, so as to obtain an original MRI image that matches each original sample CT image; of course, the image registration algorithms listed above are all commonly used techniques for image registration, and their principles will not be repeated.
[0050] Thus, after the registration of the original sample CT image and the original MRI image is achieved through the aforementioned step S12, image enhancement processing can be performed to improve the image quality; wherein, the image enhancement processing process can be but is not limited to the following step S13.
[0051] S13. Perform image enhancement processing on each original sample CT image and each original MRI image that matches the original sample CT image, so as to obtain a plurality of enhanced original sample CT images and an enhanced original MRI image that matches each enhanced original sample CT image after the image enhancement processing; in specific implementation, image enhancement can improve the resolution of the image, thereby highlighting the details and making full use of the useful information in the image; at the same time, it can also enhance the contrast of the image and make the image easier to identify; among them, image enhancement is a commonly used technology for image processing, and its principle will not be repeated.
[0052] After completing the image enhancement process, denoising processing is required to reduce the interference of noise in the image on subsequent model training; wherein, the image denoising process can be but is not limited to the following step S14.
[0053] S14. Perform denoising processing on each enhanced original sample CT image and each enhanced original MRI image that matches the enhanced original sample CT image, so as to obtain several sample CT images and several MRI images after the denoising processing; in specific implementation, various noises may exist in the image in actual application, and these noises may be generated during transmission or during processing such as quantization; therefore, in order to improve the image quality and enhance the interpretability of the image, this embodiment needs to filter out the noise from the aforementioned enhanced original sample CT image and the enhanced original MRI image before model training; among them, for example, it can be but not limited to using feature analysis based on wavelet domain or feature analysis based on spatial domain to perform image denoising; of course, other image denoising methods can also be used, and this is not specifically limited.
[0054] In this way, after completing the preprocessing of the aforementioned images and obtaining several sample CT images and several MRI images, the MRI image that matches each sample CT image can be used as a label image for each sample CT image; then, based on this, the aforementioned training data set is constructed; wherein the aforementioned process can be but is not limited to as shown in the following steps S15 and S16.
[0055] S15. Use the MRI image that matches each sample CT image as the label image of each sample CT image.
[0056] S16. Use each sample CT image and the label image corresponding to each sample CT image to form a training data set.
[0057] Therefore, through the aforementioned steps S11 to S16, a data set for subsequent model training can be constructed, and then an initial MRI image synthesis model can be constructed; then, the aforementioned training data set is used to train the model, thereby obtaining a neural network model for generating MRI images based on CT images.
[0058] The model training process may be, but is not limited to, as shown in the following step S2.
[0059] S2. Construct an initial MRI image synthesis model, and use each sample CT image in the training data set and the label image corresponding to each sample CT image to train the initial MRI image synthesis model, so as to obtain the MRI image synthesis model after the training is completed; in the specific implementation, the initial MRI image synthesis model may include but is not limited to the DRU-Net neural network model; of course, other neural network models may also be used, and are not limited to the aforementioned examples.
[0060] After the initial MRI image synthesis model is constructed, the model can be trained, that is, each sample CT image in the training data set is used as input and the synthesized MRI image of each sample CT image is used as output to train the initial MRI image synthesis model, and during the training process, the synthesized MRI image of each sample CT image and the label image of each sample CT image are used to adjust the model parameters of the initial MRI image synthesis model, so as to obtain the MRI image synthesis model after the training is completed; wherein, the aforementioned DRU-Net neural network model is a deep learning network based on ResNet and U-Net structures, and in actual training, U-Net performs feature sparsification and reconstruction estimation on T1WI data (i.e., MRI images to be synthesized) based on the input sample CT images, while ResNet ensures the efficiency of gradient updating as the number of U-Net layers increases; at the same time, the model introduces a context loss function, so that the high-frequency information in the synthesized MRI image can be retained and the low-frequency background noise in the T1WI can be removed; in this way, using this model, the accuracy of the synthesized MRI image can be guaranteed.
[0061] Based on this, this embodiment designs a deep learning model architecture based on DRU-Net, which combines the advantages of ResNet and U-Net to improve the accuracy of image synthesis; at the same time, an innovative image preprocessing process is proposed, including steps such as registration, enhancement and denoising, which can reduce the interference of noise on model training; in addition, a multi-level feature extraction and fusion strategy is adopted to improve the detail performance of the synthesized image.
[0062] Furthermore, in the training process, for example, but not limited to, the synthetic MRI images of each sample CT image and the label images corresponding to each sample CT image are subjected to consistency comparison processing, so as to obtain consistency comparison results after the consistency comparison processing; and then, according to the consistency comparison results, the model parameters of the initial MRI image synthesis model are adjusted, so as to continuously optimize the model; wherein, the example consistency comparison processing may be, but not limited to, adopting a radiological measurement method, specifically, selecting several standardized radiological measurement parameters, and then calling ITK-SNAP software, between the synthetic MRI image and the corresponding real MRI image (i.e., the corresponding A double-blind crossover measurement is performed on a labeled image), and the measurement is repeated at fixed intervals to obtain the measurement results; finally, the measurement results are evaluated by statistical methods such as the intraclass correlation coefficient (ICC), standard error of measurement (SEM) and Bland-Altman analysis, and the consistency between the synthetic image and the real image is verified within the 5% equivalent interval; optionally, taking the MRI image of the brain as an example, examples of standardized radiological measurement parameters may include but are not limited to ventricular angle, ventricular distance, intracranial size, eyeball size, pons parameters, cortical thickness and thalamus size, etc. Of course, different standardized radiological measurement parameters can be selected according to actual use, and are not limited to the aforementioned examples.
[0063] In this way, when the consistency comparison result of the synthesized MRI image of the sample CT image is the same as that of the corresponding label image, it means that the accuracy of the synthesized MRI image is higher and the error with the real image is smaller; otherwise, it means that the error is larger. At this time, it is necessary to adjust the model parameters and continue training until the error is within the preset range. Then the training can be terminated to obtain an MRI image synthesis model for generating MRI images based on CT images.
[0064] After the MRI image synthesis model is trained based on the aforementioned step S2, the MRI image can be synthesized based on the model, and the process can be but is not limited to as shown in the following step S3.
[0065] S3. Acquire a target CT image, and input the target CT image into the MRI image synthesis model to obtain a synthetic MRI image corresponding to the target CT image; in this embodiment, the target CT image is a CT image to be converted, which can be obtained by a CT device. Therefore, by directly inputting it into the aforementioned MRI image synthesis model, a corresponding synthetic MRI image can be generated from the CT image, thereby performing nuclear magnetic resonance applications based on the synthetic MRI image.
[0066] Therefore, through the method of synthesizing MRI images from CT images based on deep learning described in detail in the aforementioned steps S1 to S3, the present invention utilizes CT images and adopts deep learning technology to generate MRI images; in this way, patients only need to undergo a simple CT examination to obtain nuclear magnetic resonance images without using nuclear magnetic resonance equipment; based on this, not only can the application of nuclear magnetic resonance be realized when there is a lack of MRI equipment or MRI scanning is impossible, thereby greatly improving the flexibility of use and reducing the cost of use; at the same time, there is no need for patients to wait for a long time, thereby improving the patient's user experience; therefore, the present invention is very suitable for large-scale application and promotion.
[0067] like Figure 2 As shown, the second aspect of this embodiment provides a hardware device for implementing the method for synthesizing MRI images from CT images based on deep learning described in the first aspect of the embodiment, including:
[0068] The acquisition unit is used to acquire a training data set, wherein the training data set includes a plurality of sample CT images and a label image of each sample CT image, and the label image of any sample CT image is an MRI image that matches the any sample CT image.
[0069] The model building unit is used to build an initial MRI image synthesis model, and use each sample CT image in the training data set and the label image corresponding to each sample CT image to train the initial MRI image synthesis model, so as to obtain the MRI image synthesis model after the training is completed.
[0070] The image synthesis unit is used to obtain a target CT image and input the target CT image into the MRI image synthesis model to obtain a synthesized MRI image corresponding to the target CT image.
[0071] The working process, working details and technical effects of the device provided in this embodiment can be found in the first aspect of the embodiment and will not be described in detail here.
[0072] like Figure 3 As shown, the third aspect of this embodiment provides another device for synthesizing MRI images from CT images based on deep learning. Taking the device as an electronic device as an example, it includes: a memory, a processor and a transceiver that are communicatively connected in sequence, wherein the memory is used to store computer programs, the transceiver is used to send and receive messages, and the processor is used to read the computer program to execute the method for synthesizing MRI images from CT images based on deep learning as described in the first aspect of the embodiment.
[0073] For example, the memory may include, but is not limited to, random access memory (RAM), read only memory (ROM), flash memory, first input first output (FIFO) and / or first in last out (FILO), etc. Specifically, the processor may include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor may be implemented in at least one hardware form of DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), and PLA (Programmable Logic Array). At the same time, the processor may also include a main processor and a coprocessor. The main processor is a processor for processing data in the awake state, also known as a CPU (Central Processing Unit); the coprocessor is a low-power processor for processing data in the standby state.
[0074] In some embodiments, the processor may be integrated with a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content to be displayed on the display screen. For example, the processor may not be limited to a microprocessor of the STM32F105 series, a reduced instruction set computer (RISC) microprocessor, an X86 architecture processor, or a processor with an integrated embedded neural network processing unit (NPU); the transceiver may be, but is not limited to, a wireless fidelity (WIFI) wireless transceiver, a Bluetooth wireless transceiver, a general packet radio service technology (General Packet Radio Service, GPRS) wireless transceiver, a ZigBee protocol (a low-power local area network protocol based on the IEEE802.15.4 standard, ZigBee) wireless transceiver, a 3G transceiver, a 4G transceiver and / or a 5G transceiver, etc. In addition, the device may also include, but is not limited to, a power module, a display screen, and other necessary components.
[0075] The working process, working details and technical effects of the electronic device provided in this embodiment can be found in the first aspect of the embodiment and will not be described in detail here.
[0076] The fourth aspect of this embodiment provides a storage medium storing instructions for the method for synthesizing MRI images from CT images based on deep learning as described in the first aspect of the embodiment, that is, the storage medium stores instructions, and when the instructions are run on a computer, the method for synthesizing MRI images from CT images based on deep learning as described in the first aspect of the embodiment is executed.
[0077] The storage medium refers to a carrier for storing data, which may include but is not limited to a floppy disk, a CD, a hard disk, a flash memory, a USB flash drive and / or a memory stick, and the computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices.
[0078] The working process, working details and technical effects of the storage medium provided in this embodiment can be found in the first aspect of the embodiment and will not be described in detail here.
[0079] The fifth aspect of this embodiment provides a computer program product comprising instructions, which, when executed on a computer, enables the computer to execute the method for synthesizing MRI images from CT images based on deep learning as described in the first aspect of the embodiment, wherein the computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices.
[0080] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the protection scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.
Claims
1. A method for synthesizing MRI images from CT images based on deep learning, characterized in that: include: Acquire a training data set, wherein the training data set includes a plurality of sample CT images and a label image of each sample CT image, and the label image of any sample CT image is an MRI image that matches the any sample CT image; Constructing an initial MRI image synthesis model, and using each sample CT image in the training data set and a label image corresponding to each sample CT image to train the initial MRI image synthesis model, so as to obtain an MRI image synthesis model after the training is completed; A target CT image is acquired, and the target CT image is input into the MRI image synthesis model to obtain a synthetic MRI image corresponding to the target CT image.
2. The method according to claim 1, characterized in that Get the training dataset, including: Acquire an original data set, wherein the original data set includes a plurality of original sample CT images and a plurality of original MRI images; Performing image registration processing on a plurality of original sample CT images and a plurality of original MRI images in the original data set, so as to obtain an original MRI image matching each original sample CT image after the image registration processing; Performing image enhancement processing on each original sample CT image and each original MRI image matching the original sample CT image, so as to obtain a plurality of enhanced original sample CT images and an enhanced original MRI image matching each enhanced original sample CT image after the image enhancement processing; Performing denoising processing on each enhanced original sample CT image and each enhanced original MRI image matched with the enhanced original sample CT image, so as to obtain a plurality of sample CT images and a plurality of MRI images after the denoising processing; The MRI image matched with each sample CT image is used as the label image of each sample CT image; The training data set is composed of each sample CT image and a label image corresponding to each sample CT image.
3. The method according to claim 2, characterized in that A feature-based image registration algorithm or a pixel-based image registration algorithm is used to perform image registration processing on a number of original sample CT images and a number of original MRI images.
4. The method according to claim 1, characterized in that: The initial MRI image synthesis model is trained using each sample CT image in the training data set and the label image corresponding to each sample CT image, including: The initial MRI image synthesis model is trained by taking each sample CT image in the training data set as input and the synthetic MRI image of each sample CT image as output, and during the training process, the synthetic MRI image of each sample CT image and the label image of each sample CT image are used to adjust the model parameters of the initial MRI image synthesis model, so as to obtain the MRI image synthesis model after the training is completed.
5. The method according to claim 4, characterized in that Using the synthetic MRI images of each sample CT image and the label images of each sample CT image to adjust the model parameters of the initial MRI image synthesis model includes: Performing consistency comparison processing on the synthetic MRI image of each sample CT image and the label image corresponding to each sample CT image, so as to obtain a consistency comparison result after the consistency comparison processing; The model parameters of the initial MRI image synthesis model are adjusted according to the consistency comparison result.
6. The method according to claim 1, characterized in that The initial MRI image synthesis model includes a DRU-Net neural network model.
7. The method according to claim 1, characterized in that The MRI image matched by any sample CT image is a T1WI image.
8. A device for synthesizing MRI images from CT images based on deep learning, characterized in that: include: An acquisition unit, configured to acquire a training data set, wherein the training data set includes a plurality of sample CT images and a label image of each sample CT image, and the label image of any sample CT image is an MRI image that matches the any sample CT image; A model building unit, used to build an initial MRI image synthesis model, and use each sample CT image in the training data set and the label image corresponding to each sample CT image to train the initial MRI image synthesis model, so as to obtain the MRI image synthesis model after the training is completed; The image synthesis unit is used to obtain a target CT image and input the target CT image into the MRI image synthesis model to obtain a synthesized MRI image corresponding to the target CT image.
9. A device for synthesizing MRI images from CT images based on deep learning, characterized in that: include: A memory, a processor and a transceiver that are sequentially connected in communication, wherein the memory is used to store a computer program, the transceiver is used to send and receive messages, and the processor is used to read the computer program, and execute the method for synthesizing MRI images from CT images based on deep learning as described in any one of claims 1 to 7.
10. A computer program product comprising instructions, characterized in that When the instructions are executed on a computer, the computer is caused to execute the method for synthesizing MRI images from CT images based on deep learning as described in any one of claims 1 to 7.