Medical image processing method, system, device, and medium
Patent Information
- Application Number
- CN202510331744.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-19
- Publication Date
- 2026-09-22
AI Technical Summary
然而，尽管CCTA有着不可忽视的价值，其应用仍面临着一些挑战
[0033]首先，将传统训练模型的焦点从不断去噪改变为对输入数据来源的调整，即训练模型的所有数据均由具有更高图像处理能力的第二设备生成，例如由当下新进入市场的光子计数CT生成，这保证了高分辨率DICOM图像和整个训练数据集的质量。
Smart Images
Figure CN122798684A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of medical image processing technology, and in particular to a medical image processing method, system, device and medium. Background Technology
[0002] Computed tomography (CT) is a widely used imaging technique in clinical diagnosis. For example, coronary computed tomography angiography (CCTA) uses CT scans to generate detailed images of the heart's blood vessels, helping doctors identify atherosclerosis, stenosis, and other problems that can lead to heart attacks. However, despite its undeniable value, CCTA faces several challenges. First, blooming artifacts are a key issue affecting CCTA image quality. This phenomenon typically occurs around high-density material, such as calcified plaques within the coronary arteries, making these areas appear larger or more blurred than they actually are, potentially misleading diagnostic results. Furthermore, limited spatial resolution is a significant factor limiting CCTA accuracy. Lower spatial resolution means that small lesions or early changes may not be clearly captured, impacting early disease detection and treatment planning.
[0003] While recent technological advancements, such as photon-counting CT, have provided ultra-high-resolution images, significantly improving image quality and reducing artifacts, their high cost currently hinders widespread global adoption. Most medical institutions still rely on traditional CT equipment for clinical examinations. Therefore, developing a technology that can significantly improve the quality of conventional CT images is crucial. Advances in this technology promise to leverage lower healthcare costs while significantly enhancing diagnostic accuracy and reliability. Summary of the Invention
[0004] In view of the shortcomings of the prior art described above, the purpose of this application is to provide a computed tomography image processing method, system, device and medium to solve one or more technical problems in the prior art.
[0005] To achieve the above and other related objectives, this application provides a medical image processing method, comprising:
[0006] Obtain raw data;
[0007] The raw data is used to reconstruct the image to generate the image to be processed;
[0008] The image to be processed is preprocessed;
[0009] The preprocessed image to be processed is input into the trained image optimization model to obtain a target image, such that the resolution of the target image is higher than the resolution of the image to be processed.
[0010] The image optimization model is obtained by training a diffusion model. The training process of the diffusion model starts with high-dose, high-resolution image data and ends with target high-resolution image data. The generated data comes from a first device, and the high-dose, high-resolution image data used to train the diffusion model comes from a second device. The second device has better image processing capabilities than the first device, and under the same external conditions, the image resolution generated by the second device is higher than that generated by the first device.
[0011] In some embodiments of this application, the second device is a photon counting CT, and / or the first device is a non-photon counting CT.
[0012] In some embodiments of this application, the equivalent external conditions include using the same radiation dose during the imaging process.
[0013] In some embodiments of this application, preprocessing the image to be processed includes:
[0014] Extract pixel data from the image to be processed;
[0015] The pixel data is processed into virtual high-resolution image data using interpolation.
[0016] The virtual high-resolution image data is then normalized.
[0017] In some embodiments of this application, the image optimization model is obtained by training a diffusion model, wherein training the diffusion model includes:
[0018] A multi-dose training dataset is constructed, which includes multiple paired high-dose high-resolution data and low-dose normal-resolution data, both of which are derived from the second device.
[0019] In some embodiments of this application, the low-dose normal resolution data is processed into low-dose simulated high-resolution image data through interpolation, and paired with the high-dose high-resolution data as input data and target images for training. In some embodiments of this application, the training process of the diffusion model includes:
[0020] Noise is gradually added to the high-dose, high-resolution image data through a diffusion process using a mean-preserving degradation operator, causing it to degrade into the corresponding low-dose, low-resolution image data. Then, the noise in the low-dose, low-resolution image data is gradually eliminated through a reverse diffusion process to generate high-resolution image data. Through continuous training, the high-resolution image data gradually approaches the high-dose, high-resolution image data and is used as the target high-resolution image data.
[0021] This application also provides a medical image processing system, the system comprising:
[0022] A medical image acquisition device for generating data, the medical image acquisition device being defined as a first device;
[0023] Processor, the processor being configured to:
[0024] The raw data is used to reconstruct the image to generate the image to be processed;
[0025] The image to be processed is preprocessed;
[0026] The preprocessed image to be processed is input into the trained image optimization model to obtain a target image, such that the resolution of the target image is higher than the resolution of the image to be processed.
[0027] The image optimization model is obtained by training a diffusion model. The training process of the diffusion model starts with high-dose high-resolution image data and ends with target high-resolution image data. The generated data comes from a first device, and the high-dose high-resolution image data used to train the diffusion model comes from a second device. The second device has better image processing capabilities than the first device. Under the same external conditions, the image resolution generated by the second device is higher than that generated by the first device.
[0028] This application also provides an electronic device, which includes:
[0029] One or more processors;
[0030] A storage device for storing one or more programs that, when executed by one or more processors, enable the electronic device to perform the medical image processing method as described above.
[0031] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a computer's processor, causes the computer to perform the medical image processing method as described above.
[0032] The medical image processing methods, systems, equipment, and media provided in this application have at least the following beneficial effects:
[0033] First, the focus of traditional training models is shifted from continuous denoising to adjusting the source of input data. That is, all data for training models is generated by a second device with higher image processing capabilities, such as photon counting CT, which is now entering the market. This ensures the quality of high-resolution DICOM images and the entire training dataset.
[0034] Secondly, this application introduces a diffusion model as a deblurring and denoising network, which can effectively distinguish noise and meaningful structural information in an image. This allows for the removal of noise while preserving important edges and details as much as possible, and can better reduce or eliminate artifacts caused by the imaging process, such as blurring due to partial volume effects, thereby improving the overall clarity and readability of the image.
[0035] Furthermore, although emerging imaging devices such as photon counting CT have superior imaging capabilities and can obtain ultra-high resolution, high-quality images, their current application costs are high, making it difficult to quickly meet market demand. This is especially true for the urgent need for ultra-high resolution, high-quality images in the diagnosis of certain special symptoms, such as coronary artery CT angiography. The inventors of this application have creatively solved this technical problem by using a second device with superior imaging capabilities, such as the aforementioned second device, as a training data source. This allows low-resolution medical image data provided by a common first device, such as the aforementioned first device, to be processed into high-resolution, high-quality images. These high-resolution, high-quality images are essentially generated by the second device, providing patients with more accurate diagnoses at a low cost. Attached Figure Description
[0036] Figure 1 A schematic flowchart of the medical image processing method provided in the embodiments of this application;
[0037] Figure 2 A flowchart illustrating a method for preprocessing an image using a DICOM preprocessing module, as provided in an embodiment of this application.
[0038] Figure 3 A schematic diagram illustrating the training process of the diffusion model provided in this application embodiment;
[0039] Figure 4 A schematic diagram of the diffusion and reverse diffusion process of the diffusion model provided in this application embodiment;
[0040] Figure 5 This is a schematic diagram of the structure of a medical image processing system provided in an embodiment of this application;
[0041] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation
[0042] The following specific examples illustrate the implementation of this application. Those skilled in the art can readily understand other advantages and effects of this application from the content disclosed in this specification. This application can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be noted that, unless otherwise specified, the following embodiments and features described therein can be combined with each other.
[0043] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of this application. Therefore, the drawings only show the components related to this application and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.
[0044] In the following description, numerous details are explored to provide a more thorough explanation of embodiments of the present application. However, it will be apparent to those skilled in the art that embodiments of the present application may be practiced without these specific details. In other embodiments, well-known structures and devices are shown in block diagram form rather than in detail to avoid obscuring embodiments of the present application.
[0045] With the development of artificial intelligence technology, especially deep learning, optimization methods based on this technology are increasingly being applied to the field of medical image processing, such as image denoising and improving spatial resolution. Deep learning models can automatically learn features from large amounts of data and use them to improve image quality, for example, by learning the mapping relationship between low-resolution and high-resolution images to improve spatial resolution, and by training models to identify and remove noise in images. However, numerous studies have shown that these models have relatively limited ability to improve spatial resolution and cannot improve image resolution according to expected values. During extensive experimentation, the inventors of this application creatively discovered that the current limitations of improving spatial resolution through model training are mainly concentrated in two points. First, training data limitations: if the model is trained using conventional CT images with normal resolution, the model lacks sufficient information to learn how to effectively improve the spatial resolution of images. Second, model design biases: due to the limitations in improving image resolution, many model architectures or training strategies tend to optimize denoising tasks rather than resolution improvement. Clearly, while deep learning-based methods have shown great potential in medical image processing, they still face challenges in effectively and significantly improving spatial resolution in specific applications.
[0046] To address the aforementioned technical problems in existing technologies, this application creatively uses high-resolution, high-quality images as a training benchmark, effectively improving the optimization level of image resolution. Furthermore, this application introduces a diffusion model as the image quality improvement network and uses a mean-preserving degradation operator to optimize and reconstruct low-resolution images. The specific solutions of this application will be further described in detail below.
[0047] like Figure 1 As shown, this application provides a medical image processing method, including the following steps:
[0048] S10, Obtain raw data;
[0049] S20. Perform image reconstruction on the raw data to generate an image to be processed;
[0050] S30. Preprocess the image to be processed;
[0051] S40. Input the preprocessed image to be processed into the trained image optimization model to obtain a target image, such that the resolution of the target image is higher than the resolution of the image to be processed.
[0052] The image optimization model is obtained by training a diffusion model. The training process of the diffusion model starts with high-dose, high-resolution image data and ends with target high-resolution image data. The generated data comes from a first device, and the high-dose, high-resolution image data used to train the diffusion model comes from a second device. The second device has better image processing capabilities than the first device, and under the same external conditions, the image resolution generated by the second device is higher than that generated by the first device.
[0053] Specifically, in step S10, raw image data is acquired from a medical imaging device (e.g., a conventional CT scanner). More specifically, the raw data originates from a first device, which is a non-photon-counting CT scanner or another medical imaging device with lower image processing capabilities than the second device. Typically, the first device has lower cost and technical complexity than the second device, and its market application began earlier. For example, the first device is a conventional X-ray CT scanner, which uses an X-ray source and detector array to generate images by measuring the intensity differences of X-rays passing through different parts of the human body, and this technology relies on the acquisition and conversion of analog signals. The second device is a photon-counting CT scanner, which uses more advanced detector technology, enabling it to directly count individual X-ray photons and form images based on the photon energy information, thus providing higher contrast resolution without sacrificing spatial resolution. As those skilled in the art will understand, the first device and the second device described in this application are not limited to non-photon counting CT and photon counting CT as described above. With the continuous development of technology, other medical imaging devices with higher image processing capabilities than photon counting CT may also emerge. In this case, the first device may also be a medical imaging device including photon counting CT, and the second device may be other more advanced medical imaging devices.
[0054] In step S20, image reconstruction is performed on the raw data to generate an image to be processed. Specifically, since the raw data is data acquired by an imaging device and has not undergone any substantial post-processing, the image reconstruction process is used to transform the raw data into a two-dimensional or three-dimensional visualization image, i.e., to generate the image to be processed. As those skilled in the art will understand, the image reconstruction process includes filtered back-projection (FBP), iterative reconstruction (IR), etc.
[0055] In step S30, the image to be processed is preprocessed to be provided to the image optimization model. Specifically, as follows: Figure 2 As shown, the preprocessing of the image to be processed using the DICOM preprocessing module includes:
[0056] S301. Extract pixel data from the image to be processed, for example, extract pixel information from a DICOM file of normal resolution image data (such as a 512*512 matrix);
[0057] S302. Using an interpolation method, such as bicubic interpolation, the pixel data is processed into virtual high-resolution image data (such as a 1024*1024 matrix).
[0058] S303. Normalize the processed data so that all pixel values fall within a specific range (e.g., between 0 and 1 or -1 and 1) to prepare for further data processing, making subsequent processing steps more efficient and accurate. Examples of normalization methods include linear normalization and Z-score standardization.
[0059] In step S40, the preprocessed image to be processed is input into the trained image optimization model to obtain a target image, such that the resolution of the target image is higher than the resolution of the image to be processed. Specifically, the image optimization model is obtained by training a diffusion model. The training process of the diffusion model starts with high-dose high-resolution image data (such as a 1024*1024 matrix) and ends with the target high-resolution image data. Please refer to [reference needed]. Figure 3 The diagram shown illustrates the training process of the diffusion model provided in this embodiment of the application. The process includes:
[0060] S401. Construct a multi-dose training dataset, which includes multiple paired high-dose high-resolution data and low-dose normal-resolution data, both of which originate from the second device.
[0061] Specifically, step S401 includes:
[0062] S4010. Collect paired high-dose high-resolution data and low-dose normal-resolution data using a multi-dose raw data collection module;
[0063] S4011. Extract pixel information from the DICOM file of the low-dose normal resolution image data (such as a 512*512 matrix) using the DICOM preprocessing module;
[0064] S4012. The pixel data is processed using an interpolation method, such as bicubic interpolation, to obtain low-dose simulated high-resolution image data (e.g., a 1024*1024 matrix).
[0065] S4013. Normalize the low-dose simulated high-resolution image data and the high-dose high-resolution data, and then pair the normalized data to form a training dataset.
[0066] S402. Input the training dataset into the diffusion model.
[0067] S403. Train the diffusion model to obtain the image optimization model. For details, please refer to... Figure 4 As shown, the diffusion model gradually adds noise to the high-dose, high-resolution image data through a diffusion process, causing it to degenerate into corresponding low-dose, low-resolution image data. Then, it gradually eliminates the noise in the low-dose, low-resolution image data through a reverse diffusion process to generate high-resolution image data. Through continuous training, the high-resolution image data gradually approaches the high-dose, high-resolution image data and serves as the target high-resolution image data. This enables the reverse diffusion process of the diffusion model to obtain a high-dose, high-resolution image similar to that obtained from the second device when receiving low-dose image data.
[0068] More in detail, such as Figure 4 As shown, the diffusion process, also known as the forward process or degradation process, is based on the high-dose, high-resolution image data ( Figure 4 The leftmost image (Original represents the original image) serves as the starting point for diffusion, with the low-dose, low-resolution image data ( Figure 4 The two middle images (Degraded images represent degraded images) serve as points in the diffusion process. A mean-preserving degradation operator is used to progressively add noise with a zero mean to the high-dose, high-resolution image data, continuously degrading it into low-dose, low-resolution image data. This effectively reduces the number of noise addition steps and accelerates the model's inference speed. Taking CT images as an example, the inference time for a single CT image can be reduced by 0.05s-0.1s, for example, 0.05s, 0.075s, and 0.1s. The mean-preserving degradation operator is defined as follows:
[0069] x t =α t x0+(1-α t )x T
[0070] In the formula, x0 represents high-dose, high-resolution CT image data; x T This is low-dose simulation of high-resolution CT image data; T is the total number of diffusion steps, T is greater than or equal to 5 and less than or equal to 20, for example 5, 10, 15, 20; t takes one of 1, 2, ..., T, α t Let α be the noise factor corresponding to step t. t <α t-1 α t Based on this, experiments have shown that the noise added using the mean-preserving degradation operator of this application more closely matches the true noise distribution of CT images in the original scan data, which is helpful for noise simulation and learning.
[0071] The reverse diffusion process, also known as the reverse process or denoising process, trains the model using a large number of paired graph data pairs from the multi-dose training dataset, enabling the diffusion model to learn from low-dose, low-resolution image data. Figure 4 From the two middle images to high-dose, high-resolution data ( Figure 4 The rightmost image (representing the generated image) shows the reverse diffusion process, which completes the denoising training of the diffusion model. This allows the trained diffusion model to utilize the reverse diffusion process to progressively optimize and generate high-dose, high-resolution target images from normal-resolution images.
[0072] S404. Train the diffusion model to obtain the image optimization model.
[0073] After obtaining the trained denoising diffusion model, an image optimization model is obtained. The image to be processed can then be input into the image optimization model. The reverse diffusion process of the diffusion model is used to gradually denoise and optimize the image to obtain a high-dose, high-resolution image. Then, post-processing of the optimized image is performed according to specific needs, such as modifying the header information of DICOM.
[0074] Based on similar concepts described above, such as Figure 5 As shown, this application also provides a medical image processing system 11, which includes a medical image acquisition device 111 and a processor 112.
[0075] The medical image acquisition device 111 is used to acquire biometric data, and is also defined as a first device. This medical image acquisition device may be, for example, an X-ray imaging device, a computed tomography (CT) device, or a magnetic resonance imaging (MRI) device.
[0076] The processor 112 is configured to:
[0077] The raw data is used to reconstruct the image to generate the image to be processed;
[0078] The image to be processed is preprocessed;
[0079] The preprocessed image to be processed is input into a trained image optimization model to obtain a target image, such that the resolution of the target image is higher than the resolution of the image to be processed. The image optimization model is obtained by training a diffusion model. The training process of the diffusion model starts with high-dose high-resolution image data and ends with target high-resolution image data. The generated data comes from a first device, and the high-dose high-resolution image data used to train the diffusion model comes from a second device. The second device has better image processing capabilities than the first device. Under the same external conditions, the resolution of the image generated by the second device is higher than that of the image generated by the first device.
[0080] As those skilled in the art will understand, the processor 112 can be integrated into the first device 111 or set independently of the first device 111. Furthermore, the functions of the processor 112 can be correspondingly configured as functional modules. One or more of these functional modules can be selectively integrated into the processor 112, as long as the purpose of this application is achieved. This application does not impose any particular limitation in this regard. In addition, it should be noted that the medical image processing system 11 provided in the above embodiments and the medical image processing method provided in the above embodiments belong to the same concept. Therefore, the functions of the processor 112 have been included in the specific implementation of the medical image processing method and have been described in detail, and will not be repeated here.
[0081] like Figure 6 As shown, this application also provides an electronic device for implementing the above-described medical image processing method.
[0082] The electronic device 1 includes a memory 12, a processor 13, and a bus, and also includes a computer program, such as a medical image processing program, stored in the memory 12 and capable of running on the processor 13.
[0083] The memory 12 includes at least one type of readable storage medium, including flash memory, portable hard drive, multimedia card, card-type memory (e.g., SD or DX memory), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 12 can be an internal storage unit of the electronic device 1, such as a portable hard drive. In other embodiments, the memory 12 can be an external storage device of the electronic device 1, such as a plug-in portable hard drive, smart media card (SMC), secure digital card (SD), flash card, etc., equipped on the electronic device 1. Furthermore, the memory 12 can include both internal and external storage units of the electronic device 1. Moreover, the memory can also be an interface capable of interfacing with remote or virtual storage devices such as cloud storage to transmit information stored in these devices to the electronic device. The memory 12 can be used not only to store application software and various types of data installed on the electronic device 1, such as code for medical image processing, but also to temporarily store data that has been output or will be output.
[0084] In some embodiments, the processor 13 may be composed of integrated circuits, such as a single packaged integrated circuit or multiple integrated circuits with the same or different functions, including combinations of one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and various control chips. The processor 13 is the control unit of the electronic device 1, connecting various components of the electronic device 1 via various interfaces and lines. It executes programs or modules (such as computed tomography image processing programs) stored in the memory 12, and calls data stored in the memory 12 to perform various functions and process data of the electronic device 1.
[0085] The processor 13 executes the operating system of the electronic device 1 and various installed applications. The processor 13 executes the applications to implement the steps in the above-described medical image processing method.
[0086] For example, the computer program can be divided into one or more modules, which are stored in the memory 12 and executed by the processor 13 to complete this application. The one or more modules can be a series of computer program instruction segments capable of performing specific functions, describing the execution process of the computer program in the electronic device 1. For example, the computer program can be divided into functional modules respectively used to execute each step of the medical image processing method of this application. As those skilled in the art will understand, this application does not particularly limit the specific division. Furthermore, as those skilled in the art will understand, the processor 13 of the electronic device 1 can be the same as or different from the processor 112 of the medical image processing system, as long as the inventive purpose of this application can be achieved; this application does not particularly limit this.
[0087] Furthermore, one embodiment of this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a computer's processor, causes the computer to perform the medical image processing method as described above.
[0088] In summary, this application discloses a medical image processing method, system, device, and medium. The process involves acquiring raw data; reconstructing the raw data to generate an image to be processed; preprocessing the image to be processed; and inputting the preprocessed image to be processed into a trained image optimization model to obtain a target image, such that the resolution of the target image is higher than the resolution of the image to be processed. The image optimization model is obtained by training a diffusion model, with the training process starting with high-dose, high-resolution image data and ending with the target high-resolution image data. Specifically, the low-dose raw data originates from a first device, while the high-dose, high-resolution image data used to train the diffusion model originates from a second device. The second device has superior image processing capabilities compared to the first device, and under the same external conditions, the image resolution generated by the second device is higher than that generated by the first device. Based on the inventors' inventive and groundbreaking thinking, this application innovatively changes the traditional path for improving medical image resolution. First, the focus of traditional training models is shifted from continuous denoising to adjusting the source of input data. All training data is generated by a second device with higher image processing capabilities, such as the newly emerging photon-counting CT scanner. This ensures the quality of high-resolution DICOM images and the entire training dataset. Second, this application introduces a diffusion model as a deblurring and denoising network, which can effectively distinguish noise from meaningful structural information in the image. This allows for the removal of noise while preserving important edges and details as much as possible, better reducing or eliminating artifacts caused by the imaging process, such as blurring due to partial volume effects, thereby improving the overall image clarity and readability. Furthermore, although newly emerging imaging devices like photon-counting CT scanners have superior imaging capabilities and can obtain ultra-high resolution, high-quality images… While images are currently widely used, their high application cost makes it difficult to quickly meet market demand, especially given the urgent need for ultra-high resolution, high-quality images in the diagnosis of certain special symptoms, such as coronary artery CT angiography. The inventors of this application have creatively solved this technical problem by using a second device with superior imaging capabilities, such as the aforementioned second device, as a training data source. This allows processed images with standard resolution provided by a common first device to be transformed into high-resolution, high-quality images. These high-resolution, high-quality images are essentially generated by the second device, enabling more accurate diagnosis for patients at a lower cost. Specifically, the model is trained using data generated by the relatively expensive second device, and then applied to the low-cost first device, significantly improving the image quality of the low-cost device.
[0089] The above embodiments are merely illustrative of the principles and effects of this application and are not intended to limit this application. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of this application. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in this application should still be covered by the claims of this application.
Claims
1. A medical image processing method, characterized in that, include: Obtain raw data; The raw data is used to reconstruct the image to generate the image to be processed; The image to be processed is preprocessed; The preprocessed image to be processed is input into the trained image optimization model to obtain a target image, such that the resolution of the target image is higher than the resolution of the image to be processed. The image optimization model is obtained by training a diffusion model. The training process of the diffusion model starts with high-dose high-resolution image data and ends with target high-resolution image data. The raw data originates from the first device, and the high-dose, high-resolution image data used to train the diffusion model originates from the second device. The second device has better image processing capabilities than the first device, and under the same external conditions, the image resolution generated by the second device is higher than that generated by the first device.
2. The medical image processing method according to claim 1, characterized in that, The second device is a photon counting CT, and / or the first device is a non-photon counting CT.
3. The medical image processing method according to claim 1, characterized in that, The equivalent external conditions include using the same radiation dose during the imaging process.
4. The medical image processing method according to claim 1, characterized in that, Preprocessing the image to be processed includes: Extract pixel data from the image to be processed; The pixel data is processed into virtual high-resolution image data using interpolation. The virtual high-resolution image data is then normalized.
5. The medical image processing method according to claim 1, characterized in that, The image optimization model is obtained by training a diffusion model, and the training of the diffusion model includes: A multi-dose training dataset is constructed, which includes multiple paired high-dose high-resolution data and low-dose normal-resolution data, both of which are derived from the second device.
6. The medical image processing method according to claim 5, characterized in that, The low-dose normal resolution data is processed into low-dose simulated high-resolution image data through interpolation, and then paired with the high-dose high-resolution data as input data and target images for training.
7. The medical image processing method according to claim 5, characterized in that, The training process of the diffusion model includes: Noise is gradually added to the high-dose, high-resolution image data through a diffusion process using a mean-preserving degradation operator, causing it to degrade into the corresponding low-dose, low-resolution image data. Then, the noise in the low-dose, low-resolution image data is gradually eliminated through a reverse diffusion process to generate high-resolution image data. Through continuous training, the high-resolution image data gradually approaches the high-dose, high-resolution image data and is used as the target high-resolution image data.
8. A medical image processing system, characterized in that, include: A medical image acquisition device for acquiring biodata, the medical image acquisition device being defined as a first device; Processor, the processor being configured to: The raw data is used to reconstruct the image to generate the image to be processed; The image to be processed is preprocessed; The preprocessed image to be processed is input into the trained image optimization model to obtain a target image, such that the resolution of the target image is higher than the resolution of the image to be processed. The image optimization model is obtained by training a diffusion model. The training process of the diffusion model starts with high-dose high-resolution image data and ends with target high-resolution image data. The generated data comes from a first device, and the high-dose high-resolution image data used to train the diffusion model comes from a second device. The second device has better image processing capabilities than the first device. Under the same external conditions, the image resolution generated by the second device is higher than that generated by the first device.
9. An electronic device, characterized in that: The electronic device includes: One or more processors; A storage device for storing one or more programs, which, when executed by the one or more processors, cause the electronic device to perform the method as described in any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed by the computer's processor, causes the computer to perform the method according to any one of claims 1 to 7.