Training method of thermal ablation related image generation model

Abstract

The invention provides a training method for a thermal ablation related image generation model, and the method comprises the steps: collecting a rapid magnetic resonance sequence image and a thermal ablation related image in the same thermal ablation process through a reproducible thermal ablation simulation test; forming a training sample by taking the fast magnetic resonance sequence image as a sample and taking a corresponding thermal ablation related image as annotation data; and generating a training sample set by referring to the previous steps, and training a deep learning model to obtain a thermal ablation related image generation model. According to the method, the rapid magnetic resonance sequence image and the corresponding thermal ablation related image data are acquired through the reproducible thermal ablation simulation test, and the defects that real thermal ablation data are insufficient and training data of the model are difficult to acquire are effectively overcome. Moreover, the user can acquire the image through the rapid magnetic resonance sequence in the clinical thermal ablation process and generate the thermal ablation related image by using the trained thermal ablation related image generation model, and the real-time performance and accuracy of ablation monitoring are both considered.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image processing, and in particular to a training method for a thermal ablation-related image generation model. Background Art

[0002] Thermal ablation is a technology that uses thermal effects to cause coagulation, necrosis, vaporization, and carbonization of diseased tissue, thereby achieving the purpose of ablation and inactivation treatment. It includes laser, high-frequency electric knife, plasma coagulation, microwave therapy, and radiofrequency therapy. Magnetic resonance imaging provides information support for the thermal ablation process. Magnetic resonance imaging can monitor the temperature state of the tissue, allowing doctors to understand the current ablation status of the target area, adjust the ablation position more accurately, and control the ablation power, thereby achieving the goal of "ablating lesions as much as possible and protecting normal tissue."

[0003] The drawback of magnetic resonance imaging is that it is time-consuming, while the temperature field changes rapidly during thermal ablation, which places high demands on the real-time performance of temperature imaging. Currently, several measures have been proposed to improve the real-time performance of temperature imaging: 1. Narrowing the imaging range and acquiring fewer layers (e.g., three layers) has the drawback of limited improvement in imaging efficiency. Furthermore, since the lesion itself occupies a certain space, too few layers may not cover the lesion, resulting in a temperature monitoring blind spot. 2. Using faster sequences (e.g., EPI sequences) for image acquisition during thermal ablation significantly reduces imaging quality, resulting in large errors in the generated temperature data and making it difficult to accurately guide the thermal ablation process.

[0004] In order to overcome or at least partially overcome the above-mentioned drawbacks, the present invention provides a training method for a thermal ablation-related image generation model. Summary of the Invention

[0005] The present invention provides a training method for a thermal ablation-related image generation model, which is used to solve the defects of insufficient real-time performance or accuracy of ablation monitoring in the prior art.

[0006] In a first aspect, the present invention provides a method for training a thermal ablation-related image generation model, comprising:

[0007] Through reproducible thermal ablation simulation experiments, rapid magnetic resonance imaging sequences and thermal ablation-related images were collected during the same thermal ablation process.

[0008] Using the rapid magnetic resonance sequence image as a sample and the corresponding thermal ablation-related image as annotation data to form a training sample;

[0009] Refer to the above steps to generate a training sample set for training the deep learning model and obtain a thermal ablation-related image generation model.

[0010] Optionally, the reproducible thermal ablation simulation test refers to repeating the same thermal ablation operation on the same tissue model in the same external environment, so that the same ablation process can be reproduced multiple times.

[0011] Optionally, the rapid magnetic resonance sequence image and the corresponding thermal ablation-related image in a training sample are acquired at the same position and the same time point in a thermal ablation simulation test.

[0012] Optionally, the rapid magnetic resonance sequence image is a magnetic resonance K-space image or an amplitude map, a phase map, or a phase difference map, a temperature difference map, a temperature map or an ablation map generated based on the phase map.

[0013] Furthermore, the fast magnetic resonance sequence adopts one of the following forms: EPI, SSFSE, FRFSE.

[0014] Optionally, the thermal ablation-related image includes a magnetic resonance image with higher image quality than the fast magnetic resonance sequence image.

[0015] Furthermore, the magnetic resonance image with higher image quality than the fast magnetic resonance sequence image adopts one of the following forms: SE, GRE, and FSE.

[0016] Optionally, the fast magnetic resonance sequence is an EPI sequence, and the magnetic resonance image with higher image quality than the fast magnetic resonance sequence image is an EPI sequence with a lower EPI factor.

[0017] Optionally, the thermal ablation-related image includes a temperature difference map, a temperature map, or an ablation map generated based on a magnetic resonance image with higher image quality than the fast magnetic resonance sequence image.

[0018] Optionally, the deep learning model is selected from any one of the following: convolutional neural network, recurrent neural network, and generative adversarial network.

[0019] Optionally, after S13, the method further includes:

[0020] The thermal ablation-related image generation model is trained twice using historical clinical thermal ablation data to optimize the performance of the thermal ablation-related image generation model.

[0021] In a second aspect, the present invention further provides a method for generating thermal ablation-related images, comprising:

[0022] Acquire rapid magnetic resonance sequence images of the current patient;

[0023] Inputting the rapid magnetic resonance sequence images into a thermal ablation-related image generation model to obtain thermal ablation-related images of the current patient to provide information support for the thermal ablation process;

[0024] The thermal ablation-related image generation model is pre-trained according to any of the aforementioned training methods for the thermal ablation-related image generation model.

[0025] In a third aspect, the present invention also provides a magnetic resonance-guided thermal ablation monitoring system, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein when the processor executes the program, the steps of any of the aforementioned methods for generating thermal ablation-related images are implemented.

[0026] The present invention provides a method for training a thermal ablation-related image generation model, a method for generating thermal ablation-related images, and a magnetic resonance-guided thermal ablation monitoring system, which have at least the following beneficial effects:

[0027] 1. Rapid magnetic resonance imaging sequences and corresponding thermal ablation-related image data were collected through reproducible thermal ablation simulation experiments, effectively making up for the lack of access to three-dimensional real thermal ablation data and the difficulty in obtaining training data for constructing generative models.

[0028] 2. In some implementations, through thermal ablation simulation experiments, magnetic resonance images of sequences such as SE, GRE, and FS are collected as annotation data for fast magnetic resonance sequence images, which are used to train deep learning models. In some implementations, magnetic resonance images with less error are collected by adjusting the parameters of the fast magnetic resonance sequence, which are used as annotation data for training deep learning models, providing users with more options.

[0029] 3. In some implementations, it supports restoring magnetic resonance K-space images, in some implementations, it supports restoring magnetic resonance amplitude images, in some implementations, it supports restoring magnetic resonance phase images, in some implementations, it supports restoring temperature images, and in some implementations, it supports restoring ablation images.

[0030] 4. The model was pre-trained using thermal ablation simulation test data, providing the model with a foundation for transfer learning. The model was re-trained using historical clinical thermal ablation data, enabling the model to better supplement the missing data in the thermal ablation scenario and improve the model performance.

[0031] 5. By training a thermal ablation-related image generation model, rapid MRI sequence images can be efficiently and accurately restored and reconstructed, improving image quality. This allows for faster image acquisition during thermal ablation using rapid MRI sequences. Combined with the thermal ablation-related image generation model, high-quality thermal ablation-related images can be generated in real time, ensuring both real-time and accurate MRI temperature monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction is given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0033] Figure 1 It is a flowchart of a method for training a thermal ablation-related image generation model provided by the present invention;

[0034] Figure 2 It is a flow chart of a method for generating thermal ablation-related images provided by the present invention. DETAILED DESCRIPTION

[0035] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0036] The following combination Figure 1-Figure 2 Describe the image restoration method and model training method of a fast magnetic resonance sequence of the present invention, Figure 1 : is a flow chart of a training method for a magnetic resonance image restoration model provided by the present invention, such as Figure 1 As shown, the method includes:

[0037] S11. Through a reproducible thermal ablation simulation test, collect rapid magnetic resonance imaging sequences and thermal ablation-related images during the same thermal ablation process;

[0038] S12, using a rapid magnetic resonance image sequence as a sample and corresponding thermal ablation-related images as labeled data to form a training sample;

[0039] S13. Referring to S11 to S12, a training sample set is generated for training a deep learning model to obtain a thermal ablation-related image generation model.

[0040] First, it should be noted that, for the benefit of patients, it is impossible to intentionally interrupt MRI monitoring during clinical thermal ablation to collect training data specifically for model training. Furthermore, the thermal ablation process is continuous, and images collected at different time points have poor correlation, making them difficult to use as joint training samples. This makes training data for generating models related to thermal ablation images scarce. This method replicates the thermal ablation process through thermal ablation simulations, making training data collection more feasible and convenient. Furthermore, the acquired rapid MRI sequence images show good correlation with thermal ablation-related images.

[0041] Specifically, the "thermal ablation simulation test" refers to simulating clinical thermal ablation scenarios through experiments. Specifically, it is necessary to simulate clinical thermal ablation objects and clinical thermal ablation operations. Traditional thermal ablation tests only perform a single thermal ablation to verify the ablation effect, and do not reproduce the thermal ablation process. Moreover, traditional thermal ablation tests use living animals or ex vivo tissues as test subjects. Due to possible differences among individual animals, their reproducibility is poor. Individual differences can only be controlled as much as possible by screening animal breeds, ages and other conditions. This method is specifically designed for thermal ablation operations, a thermal ablation simulation test with "specific external environment, tissue model to be ablated, and ablation operation". The thermal ablation simulation test has good reproducibility and can reproduce the same thermal ablation process with the same external environment, tissue model to be ablated, and ablation operation. By repeating the thermal ablation simulation test multiple times, we have enough time to acquire magnetic resonance images, and by selecting the image acquisition position and the image acquisition time point, the images of multiple tests have a corresponding relationship and can be used to generate training samples. By changing the external environment, the tissue model to be ablated, and the ablation operation, multiple groups of thermal ablation simulation tests are formed, and more training samples are collected, which makes up for the problem that clinical data is difficult to obtain and difficult to generate sufficient training samples.

[0042] In step S11, rapid magnetic resonance sequence images and thermal ablation-related images of the same thermal ablation process are collected by reproducing the thermal ablation simulation test. Among them, rapid magnetic resonance sequence images refer to magnetic resonance sequence images with higher acquisition efficiency than conventional magnetic resonance sequences. For example, in the prior art, magnetic resonance images of three parallel slices are collected per cycle, and the data resolution of each layer (i.e., the thickness of each layer) is 3mm. The acquisition time of each cycle of data is about 4s. The use of rapid magnetic resonance sequence scanning can support the acquisition of 20 layers of continuous magnetic resonance images per cycle, and the data resolution of each layer is 1mm. The acquisition time of each cycle of data is 5 to 6s, which has higher data resolution and can still meet real-time requirements. In the present invention, rapid magnetic resonance imaging refers to a magnetic resonance scan in which the acquisition time for a single-layer image does not exceed 0.3s.

[0043] Rapid MRI images have higher acquisition efficiency but slightly lower image quality (e.g., artifacts or slightly lower resolution). Thermal ablation-related images, on the other hand, are higher quality than rapid MRI sequence images, or are generated based on higher quality images. Furthermore, thermal ablation-related images correspond to the acquisition position and time of the rapid MRI sequence. For example, a rapid MRI image is acquired at the fifth layer at 1 minute and 30 seconds into the thermal ablation process. By replicating the thermal ablation simulation, a thermal ablation-related image is also acquired at the fifth layer at 1 minute and 30 seconds into the thermal ablation process.

[0044] In step S12, the rapid magnetic resonance sequence images are used as samples, and the higher quality thermal ablation-related images are used as annotated data to form a training sample. More training samples can be generated by referring to S11 to S12. Of course, the image acquisition positions and acquisition time points of different training samples can be different, and the thermal ablation simulation tests can also be different, such as simulating different thermal ablation objects and simulating different thermal ablation operations, so that the training samples are more diversified and the generalization ability of the model is improved. Step S13 trains the deep learning model with the training sample set to obtain a thermal ablation-related image generation model. The model can generate corresponding thermal ablation-related images based on the rapid magnetic resonance sequence images during the thermal ablation process to optimize image quality.

[0045] It can be seen that with the support of this method, users can collect images through fast magnetic resonance sequences during clinical thermal ablation, thereby improving the real-time performance of ablation monitoring. The thermal ablation-related image generation model trained by this method processes fast magnetic resonance sequence images to generate thermal ablation-related images, thereby improving the image quality of ablation monitoring, thus taking into account both real-time performance and accuracy.

[0046] Furthermore, with the support of this method, users can collect continuous tomographic images through a rapid magnetic resonance sequence during clinical thermal ablation, and then process the rapid magnetic resonance sequence images through the thermal ablation-related image generation model trained by this method to obtain higher-quality continuous tomographic magnetic resonance images, and then generate three-dimensional thermal ablation-related images, reducing monitoring blind spots, improving monitoring quality, and providing more comprehensive information support for the thermal ablation process.

[0047] Based on the aforementioned embodiments, in some embodiments, the thermal ablation simulation test simulates the external environment, tissue model and thermal ablation operation during the thermal ablation process, and the reproduced thermal ablation simulation test refers to reproducing the external environment, tissue model and thermal ablation operation during the same thermal ablation process.

[0048] Specifically, the external environment, such as ambient temperature and humidity, for example, refers to the standard temperature of 18° to 24° and the standard humidity of 40% to 60% in the magnetic resonance imaging room. In the repeated thermal ablation simulation test, the same ambient temperature of 22° and the same ambient humidity of 50% are used.

[0049] A tissue model is a tissue structure model of the object to be thermally ablated. For example, in laser ablation scenarios, an agar block is used to simulate brain tissue. The refractive index of the agar is adjusted by adding auxiliary materials such as sugar and acid. The laser is introduced into the agar block using an optical fiber, and the thermal effect of the laser is used to ablate a specific area of ​​the agar block. For another example, in radiofrequency ablation scenarios, gelatin and agar are used as the main materials, and sodium chloride and other materials are added to adjust the conductivity to simulate brain tissue. Silicone tubes are used to simulate blood vessels, and methacrylated gelatin is used to simulate tumors. A radiofrequency ablation needle is used to apply radiofrequency current to heat a specific area of ​​the "brain model." For another example, in ultrasonic ablation scenarios, an agar block is used to simulate brain tissue, and the absorption and reflectivity of ultrasound are adjusted by adjusting the porosity of the agar block.

[0050] For example, in laser ablation scenarios, optical fibers of the same specifications were used in multiple experiments, implanted in the same location within the simulated thermal ablation environment, and outputting laser light with the same time-power curve. Another example is ultrasound ablation, where an ultrasound transducer array was placed in the same location and outputting ultrasound energy with the same time-power curve at each time period.

[0051] In this embodiment, a thermal ablation simulation test simulates the external environment, tissue model, and thermal ablation procedure during clinical thermal ablation. Because thermal ablation simulation tests are highly reproducible, the repeated tests allow the MRI system to acquire image data of varying sequence formats (or varying time consumption) and quality over multiple trials, forming training samples and avoiding the difficulty in acquiring clinical thermal ablation data.

[0052] Based on the foregoing embodiments, in some embodiments, the rapid magnetic resonance sequence images and the corresponding thermal ablation-related images in a training sample are acquired at the same position and the same time point in a thermal ablation simulation test.

[0053] Specifically, each training sample has a corresponding relationship between the rapid MRI sequence image and the thermal ablation-related image. The two images were acquired at the same location and time during the same thermal ablation process. The thermal ablation-related images have higher image quality and can be used as annotation data for the rapid MRI sequence phase.

[0054] In this embodiment, rapid magnetic resonance sequence images and thermal ablation-related images are collected at the same position and time point in a reproducible thermal ablation simulation test, so that the two are correlated, facilitating the generation of training samples.

[0055] Based on the aforementioned embodiments, in some embodiments, the fast magnetic resonance sequence image is a magnetic resonance K-space image or an amplitude map, a phase map, or a phase difference map, a temperature difference map, a temperature map or an ablation map generated based on the phase map.

[0056] Specifically, the rapid magnetic resonance sequence images can be K-space images, and the trained thermal ablation-related images can directly generate corresponding thermal ablation-related images based on the K-space images acquired during the thermal ablation process. The rapid magnetic resonance sequence images can be magnetic resonance amplitude images or phase images obtained by performing an inverse Fourier transform of the K-space images by the magnetic resonance device, or they can be phase difference images, temperature difference images, temperature images, or ablation images generated based on the phase images. The trained thermal ablation-related images can generate corresponding thermal ablation-related images based on the magnetic resonance time domain images acquired during the thermal ablation process.

[0057] Based on the aforementioned embodiments, in some embodiments, the fast magnetic resonance sequence adopts one of the following forms: EPI, SSFSE, FRFSE.

[0058] Specifically, (echo planar imaging) EPI sequence: The EPI sequence uses continuous switching of positive and negative gradients to read the acquisition signal after a single RF excitation. This type of sequence has the fastest scanning speed but the worst signal-to-noise ratio. This sequence is often used for functional imaging.

[0059] (Single-Shot FSE Sequence) SSFSE Sequence: This sequence fills all phase-encoding lines in k-space within one TR. Typically, k-space lines from more than half of the phase-encoding matrix are acquired, and the conjugate properties of k-space are used to recover the remaining half. This sequence is commonly used in dynamic and rapid imaging.

[0060] (Fast Recovery Spin Echo Sequence) FRFSE sequence: uses a flip recovery pulse to quickly restore the tissue magnetization vector from the excited state to the equilibrium state, thereby speeding up the scanning speed and increasing the contrast. This sequence is often used for T2WI imaging, but this sequence cannot be used for T1WI imaging.

[0061] Based on the aforementioned embodiments, in some embodiments, the thermal ablation-related images include magnetic resonance images with higher image quality than fast magnetic resonance sequence images.

[0062] Specifically, thermal ablation-related images have higher image quality and can be used as phase annotation data for rapid magnetic resonance sequences, facilitating the generation of training samples.

[0063] Based on the previous embodiment, in some embodiments, the magnetic resonance image with higher image quality than the fast magnetic resonance sequence image adopts one of the following forms: SE, GRE, and FSE.

[0064] Specifically, a spin echo sequence (SE) fills one phase-encoding line in K-space with each TR. Each acquisition requires a TR-long wait before the next excitation. This sequence has the slowest scan time, but offers excellent tissue contrast, the highest signal-to-noise ratio, and minimal artifacts, with scan times of 2-5 minutes.

[0065] Gradient echo (GRE) sequence: Compared to FSE, GRE significantly shortens the TR (time difference) and produces images with varying contrast using different TE and FA ratios. While the GRE sequence's signal-to-noise ratio is inferior to that of FSE, its exceptionally fast scanning speed makes it the preferred sequence for rapid imaging and functional imaging.

[0066] (Fast Spin Echo Sequence) FSE sequence: introduces an echo chain (such as N), in which a TR fills N phase-encoding lines in K space, greatly shortening the scanning time. This type of sequence is currently the most widely used.

[0067] Based on the foregoing embodiments, in some embodiments, the fast magnetic resonance sequence is an EPI sequence, and a magnetic resonance image with higher image quality than a fast magnetic resonance sequence image is an EPI sequence with a lower EPI factor.

[0068] Specifically, the EPI sequence has a parameter, the EPI factor, which indicates the number of echoes generated after a single excitation. Because the EPI sequence generates signals through continuous alternating gradients, there are no reconstruction pulses to correct for main magnetic field inhomogeneities and phase errors. As the EPI factor increases, the accumulated phase error increases, leading to increasingly severe image distortion and artifacts. Lowering the EPI factor can reduce artifacts and improve image quality.

[0069] Based on the aforementioned embodiments, in some embodiments, the thermal ablation-related image includes a temperature difference map, a temperature map, or an ablation map generated based on a magnetic resonance image with higher image quality than a fast magnetic resonance sequence image.

[0070] Specifically, in some examples, the annotated data corresponding to the rapid magnetic resonance sequence images are directly temperature difference maps, and the trained thermal ablation-related image generation model can directly generate corresponding temperature difference maps based on the rapid magnetic resonance sequence images during the thermal ablation process.

[0071] In some examples, the annotated data corresponding to the rapid magnetic resonance sequence images are directly temperature maps, and the trained thermal ablation-related image generation model can directly generate corresponding temperature maps based on the rapid magnetic resonance sequence images during the thermal ablation process.

[0072] It can be understood that the rapid magnetic resonance sequence can be a magnetic resonance image at a single slice, and the thermal ablation-related image is a temperature difference map / temperature map at the corresponding position; the rapid magnetic resonance sequence can also be a rapid magnetic resonance image at continuous slices, and the thermal ablation-related image is a three-dimensional temperature difference map / temperature map at the corresponding position; the rapid magnetic resonance sequence can also be a rapid magnetic resonance image at several slices, and the thermal ablation-related image is a temperature difference map / temperature map at the corresponding slice; the rapid magnetic resonance sequence can also be a rapid magnetic resonance image at continuous slices, and the thermal ablation-related image is a three-dimensional temperature difference map / temperature map at the corresponding position.

[0073] In some other examples, the annotated data corresponding to the rapid magnetic resonance sequence images are directly ablation images, and the trained thermal ablation-related image generation model can directly generate ablation images of corresponding positions based on the rapid magnetic resonance sequence images of each cycle during the thermal ablation process.

[0074] Based on the foregoing embodiments, in some embodiments, the deep learning model is selected from any one of the following: convolutional neural network (CNN), recurrent neural network (RNN), and generative adversarial network (GAN).

[0075] Specifically, a convolutional neural network (CNN) is a type of feedforward neural network with a deep structure that includes convolutional calculations. It has representational learning capabilities and can process input information according to its hierarchical structure. A recurrent neural network (RNN) is a type of recursive neural network that takes sequence data as input, recursively evolves in the direction of the sequence, and all nodes are connected in a chain. A generative adversarial network (GAN) consists of a generative model and a discriminative model. The generative model is responsible for capturing the distribution of sample data, while the discriminative model is generally a binary classifier that determines whether the input is real data or a generated sample. During training, one of the models is fixed and the parameters of the other model are updated, alternating iterations. Ultimately, the generative model can estimate the distribution of the sample data.

[0076] Based on the foregoing embodiment, in some embodiments, after S13, the process further includes:

[0077] The thermal ablation-related image generation model was retrained using historical clinical thermal ablation data to optimize its performance.

[0078] It is understandable that since the thermal simulation test is reproducible, it is possible to collect magnetic resonance image data of more sections and shorter time intervals of the "same thermal ablation process" through repeated tests. These data can be easily obtained and used for preliminary training of the initial deep learning model and preliminary optimization and adjustment of the model parameters so that the model has a basis for transfer learning. After that, training samples can be generated based on the magnetic resonance images in the historical clinical thermal ablation data, and the model can be trained again to further improve the performance of the model.

[0079] This embodiment collects magnetic resonance image data through reproducible thermal ablation simulation tests, effectively making up for the problem of insufficient and difficult to obtain clinical thermal ablation data, especially solving the problem of difficulty in collecting large-scale and high-quality magnetic resonance image data within a limited time. Through pre-training based on thermal ablation simulation test data and secondary training based on clinical thermal ablation data, the performance of the model is effectively improved.

[0080] A method for generating a thermal ablation-related image provided by the present invention is described below. The method for generating a thermal ablation-related image described below and the training method for generating a thermal ablation-related image model described above can refer to each other.

[0081] Figure 2 FIG. 1 is a flow chart of a method for generating thermal ablation-related images provided by the present invention, such as Figure 2 As shown, the method includes:

[0082] S21, obtaining a rapid magnetic resonance imaging sequence image of the current patient;

[0083] S22. Inputting the rapid magnetic resonance sequence image into the thermal ablation-related image generation model to obtain the thermal ablation-related image of the current patient to provide information support for the thermal ablation process;

[0084] The thermal ablation-related image generation model is pre-trained according to any of the aforementioned training methods for the thermal ablation-related image generation model.

[0085] Specifically, by acquiring rapid magnetic resonance image sequences during the thermal ablation process, the acquisition efficiency can be accelerated and the real-time performance of magnetic resonance monitoring can be improved. The rapid magnetic resonance image sequences are input into a thermal ablation-related image generation model pre-trained using the aforementioned training method for the thermal ablation-related image generation model. The model can output thermal ablation-related images at the corresponding locations, providing information support for the thermal ablation process. It is understood that the acquisition method of the rapid magnetic resonance image sequences needs to be consistent with the acquisition method of the rapid magnetic resonance image sequences used in the model training phase.

[0086] In this embodiment, during the model training stage, reproducible thermal ablation simulation tests are used to generate training samples for training the deep learning model. The training data is used to obtain a deep learning model with a thermal ablation scenario. During the model application stage, the downsampled magnetic resonance image is input into the trained deep learning model to obtain a restored and supplemented magnetic resonance image, thereby reducing the amount of data collected by the magnetic resonance equipment during the thermal ablation process and improving the real-time monitoring while ensuring data quality.

[0087] Furthermore, the thermal ablation-related images generated in the above embodiments can be magnetic resonance K-space images, or magnetic resonance amplitude maps or phase maps, and the change in phase is linearly related to the change in temperature. Therefore, a temperature difference map can be obtained by transformation based on the phase difference, and the temperature difference map can be combined with the basic temperature to generate a temperature map. The temperature map can indicate the temperature state of the tissue at a specific fault at a certain moment. Furthermore, according to the cumulative effect of temperature over time, the ablation state of the tissue can be judged, that is, an ablation map is generated according to the temperature map. Based on the above description, it can be understood that the thermal ablation-related images generated in the above embodiments can also be directly phase difference maps or temperature difference maps or temperature maps or ablation maps.

[0088] More embodiments of the method for generating thermal ablation-related images can refer to the training method of the thermal ablation-related image generation model described above, which will not be repeated here.

[0089] The present invention also provides an MRI-guided thermal ablation monitoring system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps of any of the aforementioned methods for generating thermal ablation-related images are implemented.

[0090] Furthermore, the magnetic resonance-guided thermal ablation monitoring system further includes an ablation module and a display module.

[0091] The ablation module outputs energy under the control of the processor, such as outputting laser light to the target area through an optical fiber, or emitting ultrasonic energy to the target area through an ultrasonic generator array. The display module can display patient data, such as a three-dimensional model of the patient, and can also display ablation process data, such as an image data stream collected by an MRI device, and can also display continuously updated temperature maps and ablation maps to provide information support for the thermal ablation process. Of course, the display module can also be used to interact with medical staff for instructions, making it easier for medical staff to control the ablation process.

[0092] Furthermore, the magnetic resonance-guided thermal ablation monitoring system also includes a magnetic resonance device, which acquires rapid magnetic resonance sequence images. The processor processes the rapid magnetic resonance sequence images to generate thermal ablation-related images, providing information support for the thermal ablation process.

[0093] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0094] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, or of course, by hardware. Based on this understanding, the essence of the above technical solution or the part that contributes to the existing technology can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or certain parts of the embodiments.

[0095] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A training method for a thermal ablation-related image generation model, characterized in that: include: Through reproducible thermal ablation simulation experiments, rapid magnetic resonance imaging sequences and thermal ablation-related images were collected during the same thermal ablation process. Using the rapid magnetic resonance sequence image as a sample and the corresponding thermal ablation-related image as annotation data to form a training sample; Refer to the above steps to generate a training sample set for training the deep learning model to obtain a thermal ablation-related image generation model.

2. The method for training a thermal ablation-related image generation model according to claim 1, characterized in that: The reproducible thermal ablation simulation test refers to repeating the same thermal ablation operation on the same tissue model in the same external environment, so that the same ablation process can be reproduced multiple times.

3. The method for training a thermal ablation-related image generation model according to claim 1, characterized in that: The rapid magnetic resonance sequence images and the corresponding thermal ablation-related images in a training sample are acquired at the same position and time point in the thermal ablation simulation test.

4. The method for training a thermal ablation-related image generation model according to claim 1, characterized in that: The rapid magnetic resonance sequence image is a magnetic resonance K-space image or an amplitude map, a phase map, or a phase difference map, a temperature difference map, a temperature map or an ablation map generated based on the phase map.

5. The method for training a thermal ablation-related image generation model according to claim 4, characterized in that: The fast magnetic resonance sequence adopts one of the following forms: EPI, SSFSE, and FRFSE.

6. The method for training a thermal ablation-related image generation model according to claim 1, characterized in that: The thermal ablation-related images include magnetic resonance images having higher image quality than the fast magnetic resonance sequence images.

7. The method for training a thermal ablation-related image generation model according to claim 6, characterized in that: The magnetic resonance image with higher image quality than the fast magnetic resonance sequence image adopts one of the following forms: SE, GRE, and FSE.

8. The method for training a thermal ablation-related image generation model according to claim 6, characterized in that: The fast magnetic resonance sequence is an EPI sequence, and the magnetic resonance image with higher image quality than the fast magnetic resonance sequence image is an EPI sequence with a lower EPI factor.

9. The method for training a thermal ablation-related image generation model according to claim 1, characterized in that: The thermal ablation-related image includes a temperature difference map, a temperature map, or an ablation map generated based on a magnetic resonance image having higher image quality than the fast magnetic resonance sequence image.

10. The method for training a thermal ablation-related image generation model according to claim 1, characterized in that: The deep learning model is selected from any one of the following: convolutional neural network, recurrent neural network, and generative adversarial network.

11. A method for generating thermal ablation-related images, characterized in that: include: Acquire rapid magnetic resonance sequence images of the current patient; Inputting the rapid magnetic resonance sequence images into a thermal ablation-related image generation model to obtain thermal ablation-related images of the current patient to provide information support for the thermal ablation process; The thermal ablation-related image generation model is pre-trained according to the training method for the thermal ablation-related image generation model according to any one of claims 1-11.

12. A magnetic resonance-guided thermal ablation monitoring system, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps of the method for generating thermal ablation-related images as claimed in claim 12 are implemented.

Images