A medical imaging method and system

Abstract

Embodiments of the present specification provide a medical imaging method and system, the method comprising: acquiring imaging data; acquiring a first model and a second model, wherein the image processing style of the first model is different from that of the second model; determining a model style difference based on the first model and the second model; determining a first output image based on the first model and the imaging data; determining a second input image based on the first output image and the model style difference; and determining a target image based on the second model and the second input image.

Description

Technical Field

[0001] This manual relates to the field of medical imaging, and in particular to a medical imaging method and system. Background Technology

[0002] Existing medical imaging reconstruction techniques mostly employ filtered backprojection for image reconstruction. During the filtering process, the choice of the filter kernel function determines the style of the generated tomographic image. Taking CT equipment as an example, after generating CT tomographic images, AI algorithms are needed for further processing. For instance, one AI algorithm can remove stripe artifacts from the CT tomographic image, while another can remove metallic artifacts. However, each AI algorithm may only support a specific style of CT tomographic image; therefore, existing AI algorithms are typically used in isolation.

[0003] Therefore, there is a need to provide a medical imaging method and system for connecting at least two AI algorithms to process the same imaging data multiple times. Summary of the Invention

[0004] One embodiment of this specification provides a medical imaging method, the method comprising: acquiring imaging data; acquiring a first model and a second model, wherein the first model and the second model have different image processing styles; determining model style differences based on the first model and the second model; determining a first output image based on the first model and the imaging data; determining a second input image based on the first output image and the model style differences; and determining a target image based on the second model and the second input image.

[0005] In some embodiments, the model style difference includes a sharpness difference, which is manifested in the fact that the first filter function corresponding to the first model is different from the second filter function corresponding to the second model.

[0006] In some embodiments, acquiring imaging data includes: determining the imaging data based on data acquired by the detector of a medical device using at least one filtering kernel function.

[0007] In some embodiments, the model style difference includes the filter kernel suppression ratios of the first filter function and the second filter function.

[0008] In some embodiments, determining the second input image based on the first output image and the model style difference includes: obtaining a first output noise image and a first output denoised image based on the first output image using a denoising model; converting the first output noise image to the frequency domain using a fast Fourier transform to obtain noise-generated data; convolving the noise-generated data with the filter kernel suppression rate to obtain smoothed noise-generated data; converting the smoothed noise-generated data back to the image domain using an inverse fast Fourier transform to obtain a smoothed noise image; and adding the smoothed noise image to the first output denoised image as the second input image.

[0009] In some embodiments, the model style difference is obtained by the following steps: processing the imaging data using a first filtering function to obtain a first style image; processing the imaging data using a second filtering function to obtain a second style image; and subtracting the first style image from the second style image to obtain the model style difference.

[0010] In some embodiments, the method further includes: determining a first output image based on the first model and the first style image; determining a second input image based on the first output image and the style difference of the model; and determining a target image based on the second model and the second input image.

[0011] One embodiment of this specification provides a medical imaging system, the system comprising: an image acquisition module for acquiring imaging data; a model acquisition module for acquiring a first model and a second model, wherein the first model and the second model have different image processing styles; a difference determination module for determining model style differences based on the first model and the second model; a first processing module for determining a first output image based on the first model and the imaging data; a second processing module for determining a second input image based on the first output image and the model style differences; and a third processing module for determining a target image based on the second model and the second input image.

[0012] One embodiment of this specification provides a medical imaging device, the device comprising: at least one storage medium storing computer instructions; and at least one processor executing the computer instructions to implement the above-described medical imaging method.

[0013] One embodiment of this specification provides a computer-readable storage medium that stores computer instructions, which, when read by a computer, enable the computer to execute the medical imaging method described above.

[0014] The beneficial effects of the embodiments of this specification include at least the following: (1) The medical imaging method determines the image differences corresponding to two styles, superimposes the image differences onto the first output image of the current model (e.g., the first model), obtains the target image, and thus obtains the input image of the next model (e.g., the second model), so that the same CT data can be continuously processed by at least two AI algorithms. (2) By using machine learning algorithms to train and generate a difference extraction model, the relationship between various dimensions of data (e.g., the filtering kernel function corresponding to the two styles, X-ray energy spectrum information, CT system hardware information, CT system software information, and at least one of the reconstruction methods and image differences) can be explored, thereby improving the accuracy of determining image differences. (3) By converting the first output noise image to the frequency domain through fast Fourier transform, noise-generated data is obtained. The noise-generated data is convolved with the filtering kernel suppression rate to obtain smoothed noise-generated data. The noise difference between the target image and the first output noise image is determined. The result of adding the smoothed noise image to the first output denoised image can obtain a more accurate target image. (4) By using the filtering coefficients of the first filtering kernel function and the filtering coefficients of the second filtering kernel function, the image differences corresponding to the first style and the second style can be determined more accurately and quickly. Attached Figure Description

[0015] This specification will be further described by way of exemplary embodiments, which will be described in detail with reference to the accompanying drawings. These embodiments are not limiting; in these embodiments, the same reference numerals denote the same structures, wherein:

[0016] Figure 1 These are application scenario diagrams of exemplary medical imaging systems shown in some embodiments of this specification;

[0017] Figure 2 This is a block diagram of an exemplary medical imaging apparatus according to some embodiments of this specification;

[0018] Figure 3 This is a flowchart illustrating an exemplary medical imaging method according to some embodiments of this specification;

[0019] Figure 4 This is a flowchart illustrating the determination of a second input image according to some embodiments of this specification;

[0020] Figure 5 This is a schematic diagram illustrating the continuous processing of the same CT data using multiple models according to some embodiments of this specification. Detailed Implementation

[0021] To more clearly illustrate the technical solutions of the embodiments in this specification, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are merely some examples or embodiments of this specification. For those skilled in the art, these drawings can be applied to other similar scenarios without creative effort. Unless obvious from the context or otherwise specified, the same reference numerals in the drawings represent the same structures or operations.

[0022] It should be understood that the terms “system,” “device,” “unit,” and / or “module” used herein are one way to distinguish different components, elements, parts, sections, or assemblies at different levels. However, if other terms can achieve the same purpose, they may be replaced by other expressions.

[0023] As indicated in this specification and claims, unless the context clearly indicates otherwise, the words "a," "an," "an," and / or "the" do not specifically refer to the singular and may also include the plural. Generally speaking, the terms "comprising" and "including" only indicate the inclusion of expressly identified steps and elements, which do not constitute an exclusive list, and the method or apparatus may also include other steps or elements.

[0024] Flowcharts are used in this specification to illustrate the operations performed by the system according to embodiments of this specification. It should be understood that the preceding or following operations are not necessarily performed in exact order. Instead, the steps can be processed in reverse order or simultaneously. Furthermore, other operations can be added to these processes, or one or more steps can be removed from them.

[0025] Existing medical imaging reconstruction techniques mostly employ filtered backprojection for image reconstruction. During the filtering process, the choice of the filter kernel function determines the style of the generated tomographic image. Taking CT equipment as an example, after generating CT tomographic images, AI algorithms are needed for further processing. For instance, a first AI algorithm can remove stripe artifacts from the CT tomographic image, while a second AI algorithm can remove metallic artifacts. However, each AI algorithm may only support a specific style of CT tomographic image, therefore existing AI algorithms are used individually. This manual, using CT equipment as an example, introduces a medical imaging method that enables continuous processing of the same imaging data by at least two AI algorithms, quickly obtaining a relatively accurate target image. This manual uses CT equipment as an example to illustrate the medical imaging method and system. It should be noted that the method provided in this manual can be applied to, but is not limited to, computed tomography (CT) scanners, magnetic resonance imaging (MRI) scanners, and positron emission tomography (PET) scanners.

[0026] Figure 1This is a schematic diagram illustrating an application scenario of an exemplary medical imaging system according to some embodiments of this specification. In some embodiments, such as Figure 1 As shown, the medical imaging system 100 may include a processing device 110, a network 120, a user terminal 130, a storage device 140, and a CT device 150.

[0027] The processing device 110 can be used to process data from at least one component of the medical imaging system 100 or an external data source (e.g., a cloud data center). For example, the processing device 110 can acquire imaging data; acquire a first model and a second model, wherein the first model and the second model have different image processing styles; determine model style differences based on the first model and the second model; determine a first output image based on the first model and the imaging data; determine a second input image based on the first output image and the model style differences; and determine a target image based on the second model and the second input image.

[0028] In some embodiments, the processing device 110 may include a central processing unit (CPU), an application-specific integrated circuit (ASIC), an application-specific instruction set processor (ASIP), a graphics processing unit (GPU), a physical processing unit (PPU), a digital signal processor (DSP), a field-programmable gate array (FPGA), a programmable logic device (PLD), a controller, a microcontroller unit, a reduced instruction set computer (RISC), a microprocessor, or any combination thereof. In some embodiments, the processing device 110 may be a single server or a group of servers. In some embodiments, the processing device 110 may be local or remote. In some embodiments, the processing device 110 may be implemented on a cloud platform. By way of example only, a cloud platform may include a private cloud, a public cloud, a hybrid cloud, a community cloud, a distributed cloud, an internal cloud, a multi-tiered cloud, or any combination thereof.

[0029] Network 120 may include any suitable network capable of facilitating information and / or data exchange between the medical imaging system 100. In some embodiments, components of the medical imaging system 100 (e.g., processing device 110, user terminal 130, storage device 140, and / or CT device 150) may exchange information and / or data via network 120. For example, processing device 110 may establish a connection with user terminal 130 and / or CT device 150 via network 120. In some embodiments, network 120 may be any one or more of a wired network or a wireless network. In some embodiments, network 120 may include one or more network access points. For example, network 120 may include wired or wireless network access points, such as base stations and / or network switching points, through which one or more components of the medical imaging system 100 may connect to network 120 to exchange data and / or information.

[0030] User terminal 130 may be a terminal device used by a user (e.g., a doctor, nurse, etc.). In some embodiments, user terminal 130 may include a mobile phone, tablet, computer, etc. In some embodiments, user terminal 130 may include display components (e.g., a display screen), interaction components (e.g., a mouse, keyboard, etc.). In some embodiments, user terminal 130 may interact with at least one component of medical imaging system 100 or an external data source (e.g., a cloud data center). For example, user terminal 130 may receive a second output image from processing device 110.

[0031] Storage device 140 can be used to store data, instructions, and / or any other information. In some embodiments, storage device 140 can store data and / or information acquired from at least one component of medical imaging system 100 or an external data source. For example, storage device 140 can store a first output image, a second input image, and a target image. In some embodiments, storage device 140 can store data and / or instructions used by processing device 110 to perform or use in order to complete the exemplary methods described herein. For example, storage device 140 can store medical imaging instructions for execution by processing device 110.

[0032] In some embodiments, storage device 140 may include mass storage, removable storage, or any combination thereof. In some embodiments, storage device 140 may be implemented on a cloud platform. By way of example only, cloud platforms may include private cloud, public cloud, hybrid cloud, community cloud, distributed cloud, internal cloud, multi-tiered cloud, or any combination thereof. In some embodiments, storage device 140 may be integrated into processing device 110 and / or user terminal 130.

[0033] CT (Computed Tomography) equipment 150 can be used to scan a target object to obtain scan data of that target object. The target object can be biological or non-biological. For example, the target object can be a patient, a man-made object, etc. The target object can include specific parts, organs, tissues, and / or body parts of a patient. By way of example only, the scanned object can include the head, brain, neck, body, shoulder, arm, chest, heart, stomach, blood vessels, soft tissue, knee, foot, etc., or combinations thereof. CT equipment 150 can include a gantry 151, a detector 152, a radiation source 153, and a scanning bed 154. The detector 152 and the radiation source 153 can be mounted relative to each other on the gantry 151. The target object can be placed on the scanning bed 154 and moved into the detector channel of the CT scanner 110. For ease of illustration, a reference coordinate system is introduced, which can include an X-axis, a Y-axis, and a Z-axis. The Z-axis refers to the direction in which the target object is moved into and / or out of the detector channel of the CT equipment 150. The X-axis and Y-axis can form a plane perpendicular to the Z-axis. X-ray source 153 emits X-rays to scan a target object located on scanning bed 154. The target object can be a living organism (such as a patient or animal) or a non-living organism (such as a human model or a water film). Detector 152 detects the radiation (such as X-rays) emitted by X-ray source 153. In some embodiments, detector 152 may include multiple detector units. Detector units may include scintillation detectors (such as cesium iodide detectors) or gas detectors. Detector units may be arranged in a single row or multiple rows. It should be noted that, in suitable scenarios, CT equipment 150 can be replaced by other medical devices. For example, including but not limited to computed tomography (CT) equipment, magnetic resonance imaging (MRI) equipment, positron emission tomography (PET) equipment, etc.

[0034] It should be noted that the above description of the medical imaging system 100 is for illustrative purposes only and is not intended to limit the scope of this specification. Those skilled in the art can make various modifications or variations based on the description herein. For example, the medical imaging system 100 may also include one or more other components, or one or more of the components described above may be omitted. However, these changes and modifications will not depart from the scope of this specification.

[0035] Figure 2 This is a block diagram of an exemplary medical imaging system according to some embodiments of this specification. In some embodiments, the medical imaging system 200 may be implemented using a processing device 110. In some embodiments, such as Figure 2 As shown, the medical imaging system 200 may include an image acquisition module 210, a model acquisition module 220, a difference determination module 230, a first processing module 240, a second processing module 250, and a third processing module 260.

[0036] The image acquisition module 210 can be used to acquire imaging data.

[0037] The model acquisition module 220 can be used to acquire a first model and a second model, wherein the first model and the second model have different image processing styles.

[0038] The difference determination module 230 can be used to determine the model style differences based on the first model and the second model.

[0039] The first processing module 240 can determine the first output image based on the first model and the imaging data.

[0040] The second processing module 250 can be used to determine the second input image based on the first output image and the model style differences.

[0041] The third processing module 260 can be used to determine the target image based on the second model and the second input image.

[0042] It should be noted that the above description of the medical imaging system 200 and its modules is for convenience only and should not be construed as limiting this specification to the scope of the illustrated embodiments. It is understood that those skilled in the art, after understanding the principles of the system, may arbitrarily combine the various modules or construct subsystems connected to other modules without departing from these principles. In some embodiments, Figure 2 The image acquisition module 210, model acquisition module 220, difference determination module 230, first processing module 240, second processing module 250, and third processing module 260 disclosed herein can be different modules within a single system, or a single module can implement the functions of two or more of the aforementioned modules. For example, the modules can share a single storage module, or each module can have its own separate storage module. Such variations are all within the scope of protection of this specification.

[0043] Figure 3 This is a flowchart illustrating an exemplary medical imaging method according to some embodiments of this specification. In some embodiments, process 300 may be executed by medical imaging system 100 (e.g., processing device 110) or medical imaging system 200. For example, process 300 may be stored in storage device 140 in the form of a program or instructions, and process 300 may be implemented when processing device 110 or medical imaging system 200 executes the instructions. The operational schematic diagrams of process 300 presented below are illustrative. In some embodiments, the process may be accomplished using one or more additional operations not described and / or one or more operations not discussed. Additionally, Figure 3 The order of operations shown in the diagram and described below in process 300 is not restrictive. Figure 3As shown, process 300 may include the following steps.

[0044] Step 310: Acquire imaging data. In some embodiments, step 310 may be performed by the image acquisition module 210.

[0045] Imaging data can be reconstructed images, such as CT scans or MRI images.

[0046] In some embodiments, the image acquisition module 210 can reconstruct the scan data acquired by the CT device using a reconstruction algorithm corresponding to the first style to obtain an image of the first style.

[0047] In some embodiments, the image acquisition module 210 can determine imaging data based on data acquired by the detector of the medical device using at least one filtering kernel function. For example, the image acquisition module 210 can reconstruct the data acquired by the detector of the medical device using a filtering kernel function corresponding to a first style, generating a first-style image as imaging data.

[0048] Step 320: Obtain the first model and the second model. In some embodiments, step 320 may be performed by the model acquisition module 220.

[0049] The first and second models can be used to process imaging data / images. The first and second models can be machine learning models, such as convolutional neural network models. In some embodiments, the first and second models can be preset image processing models. The first and second models can be obtained through a network. In some embodiments, the first and second models have different image processing styles. For example, the first model can correspond to a first style, and the second model can correspond to a second style. In some embodiments, the style can include sharpness. Sharpness is related to high-frequency components in the CT image. A sharp image reduces blur by increasing high-frequency components, enhancing image edges while increasing image noise. A smooth image filters out high-frequency components, thereby reducing image noise and making the image somewhat blurry. Style can represent the sharpness or blurriness of an image. In some embodiments, a style can correspond to at least one filtering function. For example, the first style can correspond to a high-pass filtering function, and the second style can correspond to a low-pass filtering function. In some embodiments, the first style image is applied to the first model, and the first model processes the first style image based on a first AI algorithm. Different models can perform different processing on CT tomographic images of corresponding styles based on different AI algorithms. For example, the first model can remove stripe artifacts from a first-style image based on a first AI algorithm, and the second model can remove metal artifacts from a second-style image based on a second AI algorithm.

[0050] Step 330: Based on the first model and the second model, determine the model style differences. In some embodiments, step 330 may be performed by the difference determination module 230.

[0051] Model style differences can characterize the differences in how well a model processes imaging data. For example, model style differences can specifically manifest as the differences between CT tomographic images corresponding to a first style and those corresponding to a second style. Specifically, model style differences can include the differences in noise between the CT tomographic images corresponding to the first and second styles.

[0052] In some embodiments, model style differences may include differences in sharpness, which are manifested in the fact that the first filter function corresponding to the first model is different from the second filter function corresponding to the second model. For example, the first style may correspond to a high-pass filter function as the first filter function, and the second style may correspond to a low-pass filter function as the second filter function, with the first filter function and the second filter function being different.

[0053] In some embodiments, model style differences may include the suppression ratios of the kernels of a first filtering function and a second filtering function. The suppression ratio characterizes the style differences between models and can be calculated using existing techniques based on the filtering coefficients of the first and second filtering functions. For example, the suppression ratio can be the ratio between two functions. The suppression ratio is generally expressed as a value between 0 and 1; a value greater than 0 and less than 1 indicates a difference in image style at that frequency. A larger value indicates a greater difference between the functions.

[0054] In some embodiments, the difference determination module 230 may determine the filter kernel suppression rate based on the filter coefficients of the first filter kernel function and the filter coefficients of the second filter kernel function.

[0055] For example, the difference determination module 230 can determine the filter kernel suppression rate based on the following formula: uppression Rate=SOFT Filter / SHARP Filter;

[0056] Wherein, suppression rate is the suppression rate of the filter kernel, SHARP Filter is the filter coefficient of the first filter kernel function, and SOFT Filter is the filter coefficient of the second filter kernel function.

[0057] In some embodiments, the image differences between the first style and the second style can be determined relatively accurately and quickly using the filtering coefficients of the first filtering kernel function and the filtering coefficients of the second filtering kernel function.

[0058] In some embodiments, the sharpness of the first style image can be greater than the sharpness of the second style image corresponding to the second style. High-frequency information in a style image with lower sharpness has already been filtered out, making it impossible to further process it to obtain a style image with relatively higher sharpness. Therefore, the sharpness of the first style image needs to be greater than the sharpness of the second style image corresponding to the second style.

[0059] In some embodiments, the difference determination module 230 can determine the image difference as the model style difference based on the first style and the second style corresponding to the first style image in any way.

[0060] For example, the difference determination module 230 can determine the image difference based on the difference between the second style and the first style. As an example only, the difference determination module 230 can subtract the image domains of the sample first-style image corresponding to the first style and the sample second-style image corresponding to the second style to obtain the image difference as the model style difference. As another example, the difference determination module 230 can subtract the noise of the sample first-style image corresponding to the first style and the noise of the sample second-style image corresponding to the second style to obtain the image difference as the model style difference.

[0061] In some embodiments, image differences can be determined relatively quickly based on the difference between the second style and the first style. For example, the difference determination module 230 can process the imaging data using a first filtering function to obtain a first-style image; process the imaging data using a second filtering function to obtain a second-style image; and calculate the difference between the first-style image and the second-style image to obtain the model style difference.

[0062] For example, the difference determination module 230 can obtain a first output noise image and a first output denoised image based on the first output image using a denoising model; transform the first output noise image to the frequency domain using a Fast Fourier Transform to obtain noise-generated data; convolve the noise-generated data with the filter kernel suppression rate to obtain smoothed noise-generated data; transform the smoothed noise-generated data back to the image domain using an Inverse Fast Fourier Transform to obtain a smoothed noise image; and add the smoothed noise image to the first output denoised image as the second input image. For more details on determining the second input image, please refer to [link to relevant documentation]. Figure 4 The details and related descriptions will not be repeated here.

[0063] For example, the difference determination module 230 can determine image differences using a difference extraction model based on at least one of the following: filtering kernel functions corresponding to two styles, X-ray energy spectrum information, CT system hardware information, CT system software information, and reconstruction methods. As an example only, the difference determination module 230 can determine image differences using a difference extraction model based on at least one of the following: filtering kernel functions, X-ray energy spectrum information, CT system hardware information, software information, and reconstruction methods corresponding to the first style, and at least one of the following: filtering kernel functions, X-ray energy spectrum information, CT system hardware information, software information, and reconstruction methods corresponding to the second style.

[0064] X-ray energy spectrum information may include at least the X-ray wavelength and / or frequency.

[0065] CT system hardware information can be information related to the hardware of a CT device (e.g., CT device 150). For example, CT system hardware information may include information related to the X-ray generator, filters, collimators, detectors, and / or analog-to-digital converters of the CT device. By way of example only, CT system hardware information may include at least one or more of the following: number of detector units per row, number of effective channels per layer of detectors, target material, tube voltage, tube current, exposure time, effective tube heat capacity, and effective high-voltage generator power. As another example, CT system hardware information may also include detector type information, such as photon-counting detector (PCD), energy integration detector (EID), etc. The aforementioned detector type information may further include the scanning mode of each detector, such as macro mode, high-resolution mode (HR), and ultra-high-resolution mode (UHR) of a photon-counting detector. As yet another example, CT system hardware information may also include scanning method information, such as axial, helical, and topological scanning methods.

[0066] CT system software information can include at least information such as filters used in the process of converting imaging data into the image domain. For example, CT system software information may include filter kernels in backprojection reconstruction algorithms or bundle filters in loop removal algorithms.

[0067] Reconstruction methods may include at least back-projection reconstruction (FBP), model-based iterative reconstruction (MBIR), deep learning-based reconstruction algorithms, and reconstruction algorithms that combine deep learning and iterative methods.

[0068] In some embodiments, the difference determination module 230 may obtain at least one of the following: the filtering kernel function, X-ray energy spectrum information, CT system hardware information, software information and reconstruction method corresponding to the first style, and the filtering kernel function, X-ray energy spectrum information, CT system hardware information, software information and reconstruction method corresponding to the second style, through the processing device 110, user terminal 130, storage device 140, CT device 150 and / or external data source.

[0069] In some embodiments, the difference determination module 230 can determine the difference in filtering kernel functions based on the first filtering kernel function corresponding to the first style and the second filtering kernel function corresponding to the second style; it can determine the difference in X-ray energy spectrum based on the X-ray energy spectrum information corresponding to the first style and the second style; it can determine the difference in CT system hardware based on the CT system hardware information corresponding to the first style and the second style; it can determine the difference in CT system software based on the software information corresponding to the first style and the second style; and it can determine the difference in reconstruction methods based on the reconstruction method corresponding to the first style and the second style.

[0070] In some embodiments, step 330 can be implemented using a difference extraction model. The difference extraction model can be a machine learning model that determines image differences based on at least one of filtering kernel functions, X-ray energy spectrum information, CT system hardware information, CT system software information, and reconstruction methods. The input to the difference extraction model can include at least one of two styles based on filtering kernel functions, X-ray energy spectrum information, CT system hardware information, software information, and reconstruction methods. For example, the difference extraction model can include at least one of filtering kernel function differences, X-ray energy spectrum differences, CT system hardware differences, CT system software differences, and / or reconstruction method differences.

[0071] In some embodiments, the difference determination module 230 can train an initial difference extraction model using at least two first training samples to generate a trained difference extraction model. The first training samples may include at least one of the following: differences in filter kernel functions, X-ray energy spectrum, CT system hardware, CT system software, and / or reconstruction methods between a first sample style and a second sample style. The labels of the first training samples may include image differences between the first sample style and the second sample style. The labels of the first training samples can be determined based on the differences between the images corresponding to the first sample style and the images corresponding to the second sample style. The difference determination module 230 can train the initial difference extraction model using at least two first training samples and their corresponding labels until the initial difference extraction model meets preset conditions, thus obtaining a trained difference extraction model. In some embodiments, the difference determination module 230 can iteratively update the parameters of the initial difference extraction model using at least two first training samples to ensure that the initial difference extraction model meets preset conditions. These preset conditions may include loss function convergence, loss function value being less than a preset value, or the number of iterations being greater than a preset number.

[0072] In some embodiments, the difference extraction model may include, but is not limited to, neural networks (NN), convolutional neural networks (CNN), deep neural networks (DNN), recurrent neural networks (RNN), or any combination thereof. For example, the difference extraction model may be a model formed by combining convolutional neural networks and deep neural networks.

[0073] In some embodiments, after training is completed, the difference determination module 230 can input at least one of the following: the filtering kernel function, X-ray energy spectrum information, CT system hardware information, software information and reconstruction method corresponding to the first style, and at least one of the filtering kernel function, X-ray energy spectrum information, CT system hardware information, software information and reconstruction method corresponding to the second style, into the difference extraction model. The difference extraction model outputs the image differences corresponding to the first style and the second style.

[0074] In some embodiments, after training is completed, the difference determination module 230 determines the image differences using the trained difference extraction model, then superimposes the image differences onto the first output image to obtain the superimposed image. The superimposed image is then used as input to the second model to obtain the second output image. An effectiveness evaluation model can be used to assess whether the second output image meets the image requirements and determine whether the difference extraction model needs further adjustment. The effectiveness evaluation model can be a machine learning model used to evaluate whether the second output image meets the image requirements. The input to the effectiveness evaluation model can include the second output image output by the second model, and the output of the effectiveness evaluation model can include an evaluation result determining whether the second output image meets the image requirements.

[0075] In some embodiments, image requirements may include at least one of noise requirements, color requirements, brightness requirements, texture requirements, etc.

[0076] In some embodiments, when the effect evaluation model evaluates the second output image as not meeting the image requirements, the difference determination module 230 can retrain the difference extraction model.

[0077] In some embodiments, a difference extraction model can be trained using machine learning algorithms to explore the relationship between data from various dimensions (e.g., the filter kernel functions corresponding to two styles, X-ray energy spectrum information, CT system hardware information, CT system software information, and at least one of the reconstruction methods and image differences), thereby improving the accuracy of determining image differences.

[0078] Step 340: Based on the first model and the imaging data, determine the first output image. In some embodiments, step 330 may be performed by the first processing module 240.

[0079] The first output image may be a first style image. In some embodiments, the first processing module 240 may input the first style image into a first model, and the first model may perform image processing on the first style image based on a first AI algorithm to generate the first output image. For example, the first processing module 240 may input the first style image into the first model, and the first model may perform stripe artifact removal processing on the first style image based on the first AI algorithm to obtain a stripe artifact-removed first style image.

[0080] Step 350: Determine the second input image based on the first output image and the model style differences. In some embodiments, step 350 may be performed by the second processing module 250.

[0081] The second input image can be the input image of the second model. For example, the second input image can be the first output image itself, or an image after superimposing image differences / model style differences onto the first output image.

[0082] Step 360: Determine the target image based on the second model and the second input image. In some embodiments, step 360 may be performed by a third processing module 260.

[0083] In some embodiments, the third processing module 260 can superimpose image differences / model style differences onto the first output image in any manner to obtain a target image. The target image can be the final processed image. For example, the target image can be the obtained noise-free image.

[0084] For example, the third processing module 260 can superimpose image differences onto the first output image using an image generation model to obtain a target image. The image generation model can be a machine learning model that generates the target image based on image differences and the output image. The input to the image generation model can include image differences and the output image, and the output of the image generation model can include the target image. In some embodiments, the image generation model can include a generative adversarial network (GAN).

[0085] In some embodiments, the third processing module 260 can pre-construct an initial image generation model, which includes a generator and a discriminator. The initial image generation model is then trained using at least two second training samples. These second training samples may include sample image differences, a first sample output image, and a sample target image. The generator of the initial image generation model receives the sample image differences and the first sample output image as input, and its output is a virtual target image. The discriminator receives the sample target image and the virtual target image output by the generator as input. The discriminator compares the sample target image and the virtual target image to determine the probability that the virtual target image was generated by the generator. Based on the discriminator's judgment, a backpropagation algorithm is used to feed back the results to the generator, guiding it to generate more realistic virtual target images. Simultaneously, the discriminator improves its own discrimination ability. Iterative training is performed using a loss function, with the two working against each other, until the virtual target image generated by the generator becomes indistinguishable from the sample target image by the discriminator. This reaches a Nash equilibrium state, or the number of iterations reaches a threshold, completing the training of the initial image generation model and obtaining the image generation model.

[0086] In some embodiments, the third processing module 260 may input the second input image into the second model, and the second model may perform image processing on the second input image based on the second AI algorithm to generate a second output image as the target image. For example, the third processing module 260 may input the second input image into the second model, and the second model may perform metal artifact removal processing on the second input image based on the second AI algorithm to obtain a second input image after metal artifact removal.

[0087] Figure 5 This is a schematic diagram illustrating the continuous processing of the same CT data using multiple models according to some embodiments of this specification, such as... Figure 5As shown, in some embodiments, process 300 can be applied to scenarios where multiple models (e.g., a first model, a second model, a third model, ..., the Nth model) continuously process the same CT data. After the second model outputs a second output image, the processing device 110 or the medical imaging system 200 can determine the image differences based on the second and third styles, superimpose the image differences onto the second output image to obtain a second target image, and use the second target image as input to the third model to obtain a third output image. The third model processes the target image based on a third AI algorithm. This process is repeated until the image differences are determined based on the (N-1)th and Nth styles, and the image differences are superimposed onto the (N-1)th output image to obtain the (N-1)th target image. The (N-1)th target image is then used as input to the Nth model to obtain the Nth output image. The Nth model processes the target image based on the Nth AI algorithm. The sharpness of the i-th style is less than that of the (i-1)th style, and i is a natural number greater than 1 and less than or equal to N.

[0088] In some embodiments, the processing device 110 or the medical imaging system 200 may also directly determine the image difference based on the j-th style and the m-th style, superimpose the image difference onto the j-th output image of the j-th model to obtain the j-th target image, and use the j-th target image as the input of the m-th model to obtain the m-th output image. The sharpness of the m-th style is less than that of the j-th style.

[0089] In some embodiments, the medical imaging method determines the image differences corresponding to two styles, superimposes the image differences onto the first output image of the current model (e.g., the first model), obtains the target image, and thus obtains the input image of the next model (e.g., the second model), enabling the same CT data to be processed continuously by at least two AI algorithms.

[0090] It should be noted that the above description of process 300 is for illustrative purposes only and does not limit the scope of this specification. Those skilled in the art can make various modifications and changes to process 300 under the guidance of this specification. However, these modifications and changes remain within the scope of this specification.

[0091] Figure 4This is a flowchart illustrating the determination of a second input image according to some embodiments of this specification. In some embodiments, process 400 may be executed by medical imaging system 100 (e.g., processing device 110) or medical imaging system 200. For example, process 400 may be stored in storage device 140 in the form of a program or instructions, and process 400 may be implemented when processing device 110 or medical imaging system 200 executes the instructions. The operational schematic diagram of process 400 presented below is illustrative. In some embodiments, the process may be accomplished using one or more additional operations not described and / or one or more operations not discussed. Additionally, Figure 4 The order of operations shown in the diagram and described below in process 400 is not restrictive. Figure 4 As shown, process 400 may include the following steps.

[0092] Step 410: Based on the first output image, obtain the first output noisy image and the first output denoised image using the denoising model.

[0093] The first output noisy image can be an image composed of noise in the first output image. The first output denoised image can be the image after removing noise from the first output image.

[0094] The denoising model can be a machine learning model that processes the output image to obtain an output noisy image and an output denoised image. The input to the denoising model can include the output image, and the output can include the output noisy image and the output denoised image corresponding to the output image. The difference determination module 230 can train the initial denoising model using at least two third training samples to generate a trained denoising model. The third training samples can include the sample output image, and the labels of the third training samples can include the sample output noisy image and the sample output denoised image corresponding to the sample output image. In some embodiments, the structure and training of the denoising model are similar to those of the difference extraction model. For a description of the structure and training of the denoising model, please refer to the description of the structure and training of the difference extraction model; it will not be repeated here.

[0095] Step 420: The first output noisy image is converted to the frequency domain by fast Fourier transform to obtain the noise generation data.

[0096] Step 430: Convolve the noise-generated data with the filtering kernel suppression rate to obtain smoothed noise-generated data.

[0097] Step 440: The smoothed noise data is transformed back into the image domain through inverse fast Fourier transform to obtain a smoothed noise image.

[0098] Step 450: The result of adding the smoothed noise image to the first output denoised image is used as the target image. A more accurate target image can be obtained through the above process.

[0099] It should be noted that the above description of process 400 is for illustrative purposes only and does not limit the scope of this specification. Those skilled in the art can make various modifications and changes to process 400 under the guidance of this specification. However, these modifications and changes remain within the scope of this specification.

[0100] The beneficial effects of the embodiments of this specification include at least the following: (1) The medical imaging method determines the image differences corresponding to two styles, superimposes the image differences onto the first output image of the current model (e.g., the first model), obtains the target image, and thus obtains the input image of the next model (e.g., the second model), so that the same CT data can be continuously processed by at least two AI algorithms. (2) By using machine learning algorithms to train and generate a difference extraction model, the relationship between various dimensions of data (e.g., the filtering kernel function corresponding to the two styles, X-ray energy spectrum information, CT system hardware information, CT system software information, and at least one of the reconstruction methods and image differences) can be explored, thereby improving the accuracy of determining image differences. (3) By converting the first output noise image to the frequency domain through fast Fourier transform, noise-generated data is obtained. The noise-generated data is convolved with the filtering kernel suppression rate to obtain smoothed noise-generated data. The noise difference between the target image and the first output noise image is determined. The result of adding the smoothed noise image to the first output denoised image can obtain a more accurate target image. (4) By using the filtering coefficients of the first filtering kernel function and the filtering coefficients of the second filtering kernel function, the image differences corresponding to the first style and the second style can be determined more accurately and quickly.

[0101] The solutions described in this specification and embodiments, if involving the processing of personal information, will be processed only under the premise of having a legal basis (such as obtaining the consent of the personal information subject, or being necessary for the performance of a contract), and will only be processed within the scope stipulated or agreed upon. A user's refusal to process personal information beyond what is necessary for basic functions will not affect the user's use of basic functions.

[0102] The basic concepts have been described above. Obviously, for those skilled in the art, the detailed disclosure above is merely illustrative and does not constitute a limitation of this specification. Although not explicitly stated herein, those skilled in the art may make various modifications, improvements, and corrections to this specification. Such modifications, improvements, and corrections are suggested in this specification and therefore remain within the spirit and scope of the exemplary embodiments described herein.

[0103] Furthermore, this specification uses specific terms to describe embodiments thereof. For example, "an embodiment," "one embodiment," and / or "some embodiments" refer to a particular feature, structure, or characteristic associated with at least one embodiment of this specification. Therefore, it should be emphasized and noted that references to "an embodiment," "one embodiment," or "an alternative embodiment" in different locations throughout this specification do not necessarily refer to the same embodiment. Moreover, certain features, structures, or characteristics in one or more embodiments of this specification can be appropriately combined.

[0104] Furthermore, unless expressly stated in the claims, the order of processing elements and sequences, the use of numbers and letters, or other names described in this specification are not intended to limit the order of the processes and methods described herein. Although various examples have been discussed in the foregoing disclosure of some embodiments of the invention that are currently considered useful, it should be understood that such details are for illustrative purposes only, and the appended claims are not limited to the disclosed embodiments; rather, the claims are intended to cover all modifications and equivalent combinations that conform to the spirit and scope of the embodiments described herein. For example, while the system components described above can be implemented using hardware devices, they can also be implemented solely using software solutions, such as installing the described system on existing servers or mobile devices.

[0105] Similarly, it should be noted that, in order to simplify the description disclosed herein and thus aid in the understanding of one or more embodiments of the invention, the foregoing description of embodiments in this specification may sometimes combine multiple features into a single embodiment, drawing, or description thereof. However, this method of disclosure does not imply that the subject matter of this specification requires more features than those mentioned in the claims. In fact, the embodiments contain fewer features than all the features of a single embodiment disclosed above.

[0106] In some embodiments, numbers describing the quantity of components and attributes are used. It should be understood that such numbers used in the description of embodiments are modified in some examples with the terms "approximately," "approximately," or "generally." Unless otherwise stated, "approximately," "approximately," or "generally" indicates that the numbers are allowed to vary by ±20%. Accordingly, in some embodiments, the numerical parameters used in the specification and claims are approximate values, which may be changed depending on the characteristics required by individual embodiments. In some embodiments, numerical parameters should take into account specified significant digits and employ a general method of digit reservation. Although the numerical ranges and parameters used to confirm their breadth of range in some embodiments of this specification are approximate values, in specific embodiments, such values ​​are set as precisely as feasible.

[0107] For each patent, patent application, patent application publication, and other material, such as articles, books, specifications, publications, and documents, referenced in this specification, the entire contents of which are incorporated herein by reference. This excludes historical application documents that are inconsistent with or conflict with the content of this specification, as well as documents that limit the broadest scope of the claims in this specification (currently or subsequently appended to this specification). It should be noted that in the event of any inconsistency or conflict between the descriptions, definitions, and / or terminology used in the supplementary materials to this specification and the content of this specification, the descriptions, definitions, and / or terminology used in this specification shall prevail.

[0108] Finally, it should be understood that the embodiments described in this specification are merely illustrative of the principles of the embodiments described herein. Other variations may also fall within the scope of this specification. Therefore, alternative configurations of the embodiments described herein are intended to be illustrative rather than limiting, and should be considered consistent with the teachings of this specification. Accordingly, the embodiments described herein are not limited to those explicitly introduced and described herein.

Claims

1. A medical imaging method, characterized in that, The method includes: Acquire imaging data; Obtain a first model and a second model, wherein the first model and the second model have different image processing styles; Based on the first model and the second model, the model style difference is determined; the model style difference characterizes the difference in the effect of the first model and the second model on image data processing; Based on the first model and the imaging data, a first output image is determined; The second input image is determined based on the first output image and the model style differences; Based on the second model and the second input image, the target image is determined; The first filter function corresponding to the first model is different from the second filter function corresponding to the second model; Determining the second input image based on the first output image and the model style difference specifically includes: obtaining the noise of the first output image based on the first output image, and determining the second input image based on the first output image, the noise of the first output image, and the model style difference; The model style difference includes the suppression ratio of the filter kernels of the first filter function and the second filter function; the suppression ratio of the filter kernel is determined based on the filter coefficients of the first filter kernel function and the second filter kernel function. Obtaining noise from the first output image based on the first output image, and determining the second input image based on the first output image, the noise from the first output image, and the model style difference, including: Based on the first output image, a first output noisy image and a first output denoised image are obtained using a denoising model. The first output noise image is converted to the frequency domain using a fast Fourier transform to obtain noise generation data; The noise-generated data is convolved with the filter kernel suppression rate to obtain smoothed noise-generated data; The smoothed noise data is converted back to the image domain using an inverse fast Fourier transform to obtain a smoothed noise image; The result of adding the smoothed noise image to the first output denoised image is used as the second input image.

2. The method according to claim 1, characterized in that, The differences in model style include differences in sharpness.

3. The method according to claim 2, characterized in that, The acquisition of imaging data includes: The imaging data is determined by using data acquired from the detector of the medical device and passing it through at least one filtering kernel function.

4. The method according to claim 2, characterized in that, The model style differences are obtained through the following steps: The imaging data is processed using a first filtering function to obtain a first-style image; The imaging data is processed using a second filtering function to obtain a second-style image; The difference between the first style image and the second style image is obtained to obtain the style difference of the model.

5. The method as described in claim 4, characterized in that, The method further includes: Based on the first model and the first style image, the first output image is determined; The second input image is determined based on the first output image and the model style differences; The target image is determined based on the second model and the second input image.

6. A medical imaging system, characterized in that, The system includes: The image acquisition module is used to acquire imaging data; The model acquisition module is used to acquire a first model and a second model, wherein the first model and the second model have different image processing styles; The difference determination module is used to determine the model style difference based on the first model and the second model; the model style difference characterizes the difference in the effect of the first model and the second model on image data processing; The first processing module determines a first output image based on the first model and the imaging data; The second processing module is used to determine a second input image based on the first output image and the model style difference. Specifically, determining the second input image based on the first output image and the model style difference includes: obtaining noise from the first output image; determining the second input image based on the first output image, the noise from the first output image, and the model style difference; the model style difference includes the suppression ratio of the filter kernels of the first and second filter functions; the suppression ratio of the filter kernels is determined based on the filter coefficients of the first and second filter kernel functions; obtaining noise from the first output image; determining the second input image based on the first output image, the noise from the first output image, and the model style difference includes: obtaining a first output noise image and a first output denoised image based on the first output image using a denoising model; converting the first output noise image to the frequency domain using a fast Fourier transform to obtain noise-generated data; convolving the noise-generated data with the suppression ratio of the filter kernel to obtain smoothed noise-generated data; converting the smoothed noise-generated data back to the image domain using an inverse fast Fourier transform to obtain a smoothed noise image; and using the result of adding the smoothed noise image and the first output denoised image as the second input image. The third processing module is used to determine the target image based on the second model and the second input image; The first filtering function corresponding to the first model is different from the second filtering function corresponding to the second model.

7. A medical imaging device, characterized in that, The device includes: At least one storage medium that stores computer instructions; At least one processor executes the computer instructions to implement the method of any one of claims 1 to 5.

8. A computer-readable storage medium storing computer instructions that, when read by a computer, execute the method as described in any one of claims 1 to 5.

Images