Image Processing Method and Apparatus, Medical Imaging Device, and Storage Medium

Abstract

An image processing method, apparatus, medical imaging device, and storage medium belong to the field of medical diagnosis technologies. The method includes: obtaining to-be-corrected projection data collected by a medical imaging device (101); performing reconstruction processing on the to-be-corrected projection data to obtain a to-be-corrected image, and inputting the to-be-corrected image into a first correction model to obtain a first corrected image output by the first correction model (102); inputting the to-be-corrected projection data into a second correction model to obtain corrected projection data output by the second correction model, and performing reconstruction processing on the corrected projection data to obtain a second corrected image (103); generating a target corrected image corresponding to the to-be-corrected projection data based on the first corrected image and the second corrected image (104). The probability of artifacts in the target corrected image is relatively low, thereby effectively improving the quality of images obtained by a medical imaging device for clinical treatment.

Description

Technical Field

[0001] This application relates to the field of medical diagnostic technologies, and particularly to an image processing method and apparatus, a medical imaging device, and a storage medium. Background Art

[0002] Currently, medical imaging devices using technologies such as Computed Tomography (CT), spectral CT, Cone Beam CT (CBCT), and Magnetic Resonance Imaging (MRI) have been widely used in clinical medical imaging diagnosis.

[0003] During the imaging process of a medical imaging device, the medical imaging device needs to emit X-rays towards the target area of a patient, capture the projection data generated after the X-rays pass through the patient's target area, and then perform reconstruction processing on the projection data to obtain an image for clinical treatment.

[0004] However, the projection data collected by current medical imaging devices has a large error, resulting in artifacts (such as beam hardening artifacts, scatter artifacts, noise artifacts, ring artifacts, and metal artifacts, etc.) in the subsequently reconstructed image, and thus the quality of the reconstructed image is poor. Summary of the Invention

[0005] Embodiments of this application provide an image processing method and apparatus, a medical imaging device, and a storage medium. It can solve the problem that the quality of the image reconstructed by the medical imaging device in the prior art is poor. The technical solutions are as follows:

[0006] On the one hand, an image processing method is provided. The method includes:

[0007] Obtain the projection data to be corrected collected by a medical imaging device;

[0008] Perform reconstruction processing on the projection data to be corrected to obtain an image to be corrected, and input the image to be corrected into a first correction model to obtain a first corrected image output by the first correction model. The first correction model is used to eliminate the artifacts in the input image to be corrected;

[0009] Input the projection data to be corrected into a second correction model to obtain the corrected projection data output by the second correction model, and perform reconstruction processing on the corrected projection data to obtain a second corrected image. The second correction model is used to eliminate the data corresponding to the artifacts in the input projection data to be corrected;

[0010] Generate a target corrected image corresponding to the projection data to be corrected based on the first corrected image and the second corrected image.

[0011] Optionally, generating a target corrected image corresponding to the projection data to be corrected based on the first corrected image and the second corrected image includes:

[0012] Perform image block extraction processing on the first corrected image to obtain a plurality of first sub-images;

[0013] Perform image block extraction processing on the second corrected image to obtain a plurality of second sub-images corresponding one-to-one to the plurality of first sub-images, and the position of each second sub-image in the second corrected image is the same as the position of the corresponding first sub-image in the first corrected image;

[0014] In each of the first sub-images and the corresponding second sub-images, determine the image with a higher degree of artifact elimination;

[0015] Generate a target corrected image corresponding to the projection data to be corrected based on the plurality of images with a higher degree of artifact elimination.

[0016] Optionally, determining the image with a higher degree of artifact elimination in each of the first sub-images and the corresponding second sub-images includes:

[0017] Input the first sub-image and the second sub-image into the selection model at the same time, and determine the image output by the selection model as the image with a higher degree of artifact elimination. The selection model is used to select the image with a higher degree of artifact elimination from the received plurality of sub-images.

[0018] Optionally, before determining the image with a higher degree of artifact elimination in each of the first sub-images and the corresponding second sub-images, the method further includes:

[0019] Perform a plurality of first training processes on the binary classification model to obtain the selection model;

[0020] Wherein, each first training process includes:

[0021] Input two images into the binary classification model at the same time, and the two images include an image containing artifacts and an image not containing artifacts;

[0022] Judge whether the image output by the binary classification model is the image not containing artifacts;

[0023] If the image output by the binary classification model is not the image not containing artifacts, adjust the model parameters of the binary classification model.

[0024] Optionally, in each of the first sub-images and the corresponding second sub-images, determining the image with a higher degree of artifact elimination includes:

[0025] Determining a first artifact image corresponding to the first sub-image and a second artifact image corresponding to the second sub-image;

[0026] Determining a structural similarity index between the first artifact image and a third sub-image, and a structural similarity index between the second artifact image and the third sub-image, where the third sub-image is located within the image to be corrected, and the position of the third sub-image in the image to be corrected is the same as the position of the first sub-image in the first corrected image;

[0027] In each of the first sub-images and the corresponding second sub-images, determining the sub-image corresponding to the artifact image with a larger structural similarity index to the third sub-image as the image with a higher degree of artifact elimination.

[0028] Optionally, determining a first artifact image corresponding to the first sub-image and a second artifact image corresponding to the second sub-image includes:

[0029] Determining the first artifact image based on the first sub-image and the third sub-image;

[0030] Determining the second artifact image based on the second sub-image and the third sub-image.

[0031] Optionally, generating a target corrected image corresponding to the projection data to be corrected based on the first corrected image and the second corrected image includes:

[0032] In the first corrected image and the second corrected image, determining the corrected image with a higher degree of artifact elimination as the target corrected image corresponding to the projection data to be corrected.

[0033] Optionally, before inputting the image to be corrected into the first correction model to obtain the first corrected image output by the first correction model, the method further includes:

[0034] Performing a plurality of second training processes on the first artifact elimination model to obtain the first correction model;

[0035] Wherein, each second training process includes:

[0036] Inputting an image containing artifacts into the first artifact elimination model to obtain the image output by the first artifact elimination model;

[0037] Adjust the model parameters of the first artifact elimination model based on the image output by the first artifact elimination model and the image without artifacts.

[0038] Optionally, before inputting the projection data to be corrected into the second correction model to obtain the corrected projection data output by the second correction model, the method further includes:

[0039] Performing a plurality of third training processes on the second artifact elimination model to obtain the second correction model;

[0040] Wherein, each of the third training processes includes:

[0041] Input a projection data containing artifact data into the second artifact elimination model to obtain the projection data output by the second artifact elimination model;

[0042] Adjust the model parameters of the second artifact elimination model based on the projection data output by the second artifact elimination model and the projection data without artifact data.

[0043] On the other hand, an image processing apparatus is provided, the apparatus includes:

[0044] An acquisition module, configured to acquire projection data to be corrected collected by a medical imaging device;

[0045] A first reconstruction module, configured to perform reconstruction processing on the projection data to be corrected to obtain an image to be corrected;

[0046] A first correction module, configured to input the image to be corrected into a first correction model to obtain a first corrected image output by the first correction model, the first correction model being used to eliminate artifacts in the input image to be corrected;

[0047] A second correction module, configured to input the projection data to be corrected into a second correction model to obtain the corrected projection data output by the second correction model;

[0048] A second reconstruction module, configured to perform reconstruction processing on the corrected projection data to obtain a second corrected image, the second correction model being used to eliminate data corresponding to artifacts in the input projection data to be corrected;

[0049] A generation module, configured to generate a target corrected image corresponding to the projection data to be corrected based on the first corrected image and the second corrected image.

[0050] In yet another aspect, a medical imaging device is provided, including: a processor, and a memory for storing executable instructions of the processor; wherein, the processor is configured to execute the image processing method described in any one of the above.

[0051] In another aspect, a computer-readable storage medium is provided, in which instructions are stored. When the readable storage medium runs on a processing component, the processing component is caused to execute any one of the above-mentioned image processing methods.

[0052] The beneficial effects brought by the technical solutions provided in the embodiments of the present application at least include:

[0053] After the medical imaging device acquires the projection data to be corrected, the projection data to be corrected can be reconstructed to obtain a to-be-corrected graph, and a first correction model is used to eliminate the artifacts in the to-be-corrected image to obtain a first corrected image; at the same time, a second correction model can be used to eliminate the data corresponding to the artifacts in the to-be-corrected projection data, and the projection data after the artifact data is eliminated is reconstructed to obtain a second corrected image. Subsequently, after generating a target corrected image corresponding to the projection data to be corrected acquired by the medical imaging device based on the first corrected image and the second corrected image, the images in each region of the target corrected image are all: among the images in the corresponding region of the first corrected image and the images in the corresponding region of the second corrected image, the images with a higher degree of artifact elimination, so that the degree of artifact elimination in each region of the target corrected image is relatively high, and the probability of the target corrected image containing artifacts is relatively low, thereby effectively improving the quality of the images obtained by the medical imaging device for clinical treatment. Description of the Drawings

[0054] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0055] Figure 1 is a flowchart of an image processing method provided by an embodiment of the present application;

[0056] Figure 2 is a flowchart of another image processing method provided by an embodiment of the present application;

[0057] Figure 3 is a flowchart of a method for generating a target corrected image corresponding to the projection data to be corrected provided by an embodiment of the present application;

[0058] Figure 4 is an effect diagram of obtaining a plurality of first sub-images after performing image block extraction processing on a first corrected image provided by an embodiment of the present application;

[0059] Figure 5It is an effect diagram of obtaining multiple second sub-images after performing image block extraction processing on a second corrected image provided by an embodiment of the present application;

[0060] Figure 6 It is an effect diagram of generating a target corrected image corresponding to the to-be-corrected projection data provided by an embodiment of the present application;

[0061] Figure 7 It is a structural block diagram of an image processing device provided by an embodiment of the present application;

[0062] Figure 8 It is a structural block diagram of a generation module provided by an embodiment of the present application. Detailed implementation manners

[0063] To make the objectives, technical solutions, and advantages of the present application clearer, the following will further describe the embodiments of the present application in detail with reference to the accompanying drawings.

[0064] Please refer to Figure 1 , Figure 1 It is a flowchart of an image processing method provided by an embodiment of the present application. This image processing method is applied to a medical imaging device, and it can process the images obtained by the medical imaging device. This image processing method may include:

[0065] Step 101, obtain the to-be-corrected projection data collected by the medical imaging device.

[0066] Step 102, perform reconstruction processing on the to-be-corrected projection data to obtain a to-be-corrected image, and input the to-be-corrected image into a first correction model to obtain a first corrected image output by the first correction model.

[0067] This first correction model is used to eliminate the artifacts in the input to-be-corrected image.

[0068] Step 103, input the to-be-corrected projection data into a second correction model to obtain the corrected projection data output by the second correction model, and perform reconstruction processing on the corrected projection data to obtain a second corrected image.

[0069] This second correction model is used to eliminate the data corresponding to the artifacts in the input to-be-corrected projection data.

[0070] Step 104, generate a target corrected image corresponding to the to-be-corrected projection data based on the first corrected image and the second corrected image.

[0071] In summary, for the image processing method provided in the embodiments of the present application, after the medical imaging device acquires the projection data to be corrected, the projection data to be corrected can be reconstructed to obtain an image to be corrected, and the first correction model is used to eliminate the artifacts in the image to be corrected to obtain a first corrected image; at the same time, the second correction model can be used to eliminate the data corresponding to the artifacts in the projection data to be corrected, and the projection data after the artifact data is eliminated is reconstructed to obtain a second corrected image. Subsequently, after generating a target corrected image corresponding to the projection data to be corrected acquired by the medical imaging device based on the first corrected image and the second corrected image, the probability of the target corrected image containing artifacts is relatively low, thereby effectively improving the quality of the image obtained by the medical imaging device for clinical treatment.

[0072] Please refer to Figure 2 , Figure 2 which is a flowchart of another image processing method provided in the embodiments of the present application. This image processing method is applied to a medical imaging device and can process the images obtained by the medical imaging device. This image processing method may include:

[0073] Step 201, acquire the projection data to be corrected collected by the medical imaging device.

[0074] In the embodiments of the present application, the medical imaging device can acquire the projection data to be corrected collected by it. By way of example, the medical imaging device may be a CT imaging device, an energy spectrum CT imaging device, a CBCT imaging device, an MRI imaging device, etc. During the imaging process of the medical imaging device, the projection data can be collected first. However, since the error of the projection data collected by the medical imaging device is relatively large, if the corresponding image is directly reconstructed subsequently, the image will contain artifacts. Therefore, in order to reduce the probability of the reconstructed image containing artifacts, the embodiments of the present application need to process the reconstructed image to eliminate the artifacts in the reconstructed image, that is, sequentially execute the following steps 202 and 203; at the same time, it is also necessary to process the projection data collected by the medical imaging device to eliminate the artifacts in the reconstructed image subsequently, that is, sequentially execute the following steps 204 and 205.

[0075] It should be noted that in the present application, steps 204 and 205 can be sequentially executed while steps 202 and 203 are sequentially executed; or steps 204 and 205 can be sequentially executed after steps 202 and 203 are sequentially executed first. The embodiments of the present application do not make any limitations in this regard.

[0076] Step 202, perform a reconstruction process on the projection data to be corrected to obtain an image to be corrected.

[0077] In the embodiments of the present application, a medical imaging device may perform reconstruction processing on the to-be-corrected projection data collected thereby to obtain a to-be-corrected image. It should be noted that the to-be-corrected image contains artifacts, and the artifacts in the corrected image are relatively clear.

[0078] Step 203: Input the to-be-corrected image into the first correction model to obtain a first corrected image output by the first correction model.

[0079] In the embodiments of the present application, a medical imaging device may input the to-be-corrected image into the first correction model to obtain a first corrected image output by the first correction model.

[0080] In the present application, the first correction model is used to eliminate the artifacts in the input to-be-corrected image. Thus, after the medical imaging device inputs the to-be-corrected image into the first correction model, the clarity of the artifacts in the first corrected image output by the first correction model is lower than that of the artifacts in the to-be-corrected image. Optionally, the first correction model may be an artificial intelligence (abbreviation: AI) model obtained by learning and training using a deep neural network or a convolutional neural network, etc.

[0081] It should be noted that before inputting the to-be-corrected image into the first correction model to obtain a first corrected image output by the first correction model, the medical imaging device also needs to obtain the first correction model. In the present application, the medical imaging device needs to perform a plurality of second training processes on the first artifact elimination model to obtain the first correction model.

[0082] However, since a large amount of data needs to be processed in the process of training the first artifact elimination model to obtain the first correction model, in order to improve the training efficiency, in the present application, a training server with stronger computing power may be used to train the first artifact elimination model to obtain the first correction model. After the training is completed, the first correction model may be sent to the medical imaging device so that the medical imaging device can obtain the first correction model. In this case, the training server may perform a plurality of second training processes on the first artifact elimination model to obtain the first correction model.

[0083] Exemplarily, each second training process may include:

[0084] Step A1: Input an image containing artifacts into the first artifact elimination model to obtain an image output by the first artifact elimination model.

[0085] Step B1: Adjust the model parameters of the first artifact elimination model based on the image output by the first artifact elimination model and the image not containing artifacts.

[0086] In this application, the similarity between the image output by the first artifact removal model and the image without artifacts can be compared, and the model parameters of the first artifact removal model can be adjusted according to the comparison result. It should be noted that as the number of training times increases, the similarity between the image output by the first artifact removal model and the image without artifacts becomes higher and higher.

[0087] It should be noted that during multiple executions of the second training process on the first artifact removal model, if the cut-off condition is reached, the first artifact removal model after the last execution of the second training process can be determined as the first calibration model.

[0088] Optionally, the cut-off condition may include: performing a specified number of second training processes; or, when continuously performing n second training processes, the similarity between the image output by the first artifact removal model and the image without artifacts is greater than the similarity threshold each time. Where n is an integer greater than 1.

[0089] After the first artifact removal model is trained to obtain the first calibration model, at least part of the artifacts in the image containing artifacts can be removed through the first calibration model. Thus, if an image containing artifacts is input to the first calibration model, the clarity of the artifacts in the image output by the first calibration model is much lower than the clarity of the artifacts in the image containing artifacts.

[0090] In the embodiments of this application, in order to ensure that the first calibration model trained subsequently can accurately remove the artifacts in images with different display contents, during the training process of the first artifact removal model, images with different display contents containing artifacts can be input into the first artifact removal model, and then the similarity between the image output by the first artifact removal model and the image without artifacts but with the same display content as this image can be compared to adjust the model parameters of the first artifact removal model.

[0091] Therefore, before training the first artifact removal model, multiple images without artifacts and a large number of images containing artifacts can be obtained in advance. Where each image without artifacts corresponds to multiple images containing artifacts, and the display content of each image without artifacts is the same as the display content of the corresponding multiple images containing artifacts. Exemplarily, the multiple images without artifacts may include: head images without artifacts, chest images without artifacts, leg images without artifacts, etc. Correspondingly, the large number of images containing artifacts includes: multiple head images containing artifacts corresponding to the head images without artifacts, multiple chest images containing artifacts corresponding to the chest images without artifacts, and multiple leg images containing artifacts corresponding to the leg images without artifacts, etc.

[0092] Step 204: Input the projection data to be corrected into the second correction model to obtain the corrected projection data output by the second correction model.

[0093] In the embodiments of the present application, the medical imaging device may input the projection data to be corrected into the second correction model to obtain the corrected projection data output by the second correction model.

[0094] In the present application, the second correction model is used to eliminate the data corresponding to the artifacts in the input projection data to be corrected. Thus, after the medical imaging device inputs the projection data to be corrected into the second correction model, the proportion of the artifact data contained in the corrected projection data output by the second correction model is lower than the proportion of the artifact data contained in the projection data to be corrected. Subsequently, when the corrected projection data is processed to obtain the second corrected image, the clarity of the artifacts in the second corrected image is lower than the clarity of the artifacts in the image to be corrected. Optionally, the second correction model may be an AI model obtained by learning and training using a deep neural network or a convolutional neural network, etc.

[0095] It should be noted that before inputting the projection data to be corrected into the second correction model to obtain the corrected projection data output by the second correction model, the medical imaging device also needs to obtain the second correction model. In the present application, the medical imaging device needs to perform multiple third training processes on the second artifact elimination model to obtain the second correction model.

[0096] However, since a large amount of data needs to be processed in the process of training the second artifact elimination model to obtain the second correction model, in order to improve the training efficiency, in the present application, a training server with stronger computing power may be used to train the second artifact elimination model to obtain the second correction model. After the training is completed, the second correction model may be sent to the medical imaging device so that the medical imaging device can obtain the second correction model. In this case, the training server may perform multiple third training processes on the second artifact elimination model to obtain the second correction model.

[0097] Exemplarily, each third training process may include:

[0098] Step A2: Input a projection data containing artifact data into the second artifact elimination model to obtain the projection data output by the second artifact elimination model.

[0099] Step B2: Based on the projection data output by the second artifact elimination model and the projection data not containing artifact data, adjust the model parameters of the second artifact elimination model.

[0100] In the embodiments of the present application, the similarity between the projection data output by the second artifact elimination model and the projection data without artifact data can be compared, and the model parameters of the second artifact elimination model can be adjusted according to the comparison result. It should be noted that as the number of training times increases, the similarity between the projection data output by the second artifact elimination model and the projection data without artifact data becomes higher and higher.

[0101] It should be noted that during multiple executions of the third training process on the second artifact elimination model, if the cut-off condition is reached, the second artifact elimination model after the last execution of the third training process can be determined as the second calibration model.

[0102] Optionally, the cut-off condition may include: executing the third training process a specified number of times; or, when the third training process is continuously executed m times, the similarity between the projection data output by the second artifact elimination model and the projection data without artifact data is greater than the similarity threshold each time. Where m is an integer greater than 1.

[0103] After the second artifact elimination model is trained to obtain the second calibration model, at least part of the artifact data in the projection data containing artifact data can be eliminated through the second calibration model. Thus, if the projection data containing artifact data is input to the second calibration model, the proportion of artifact data in the projection data output by the second calibration model is much lower than the proportion of artifact data in the projection data with artifact data.

[0104] In the embodiments of the present application, in order to ensure that the subsequent trained second calibration model can accurately eliminate the artifact data in the projection data of different projection contents, during the training process of the second artifact elimination model, the projection data of different projection contents containing artifact data can be input into the second artifact elimination model, and then the similarity between the projection data output by the second artifact elimination model and the projection data without artifact data but with the same projection content as the projection data is compared to adjust the model parameters of the second artifact elimination model.

[0105] To this end, before training the second artifact elimination model, multiple projection data without artifact data and a large amount of projection data with artifact data can be obtained in advance. Among them, each projection data without artifact data corresponds to multiple projection data with artifact data, and the projection content of each projection data without artifact data is consistent with the projection content of the corresponding multiple projection data with artifact data. By way of example, the multiple projection data without artifact data may include: head projection data without artifact data, chest projection data without artifact data, leg projection data without artifact data, and the like. Correspondingly, the large amount of projection data with artifact data includes: multiple head projection data with artifact data corresponding to the head projection data without artifact data, multiple chest projection data with artifact data corresponding to the chest projection data without artifact data, and multiple leg projection data with artifact data corresponding to the leg projection data without artifact data, and the like.

[0106] Step 205: Perform a reconstruction process on the corrected projection data to obtain a second corrected image.

[0107] In an embodiment of the present application, the medical imaging device may perform a reconstruction process on the corrected projection data to obtain a second corrected image. Since at least some of the projection data in the projection data to be corrected are eliminated by the second correction model in step 204, the clarity of the artifacts in the second corrected image obtained after the reconstruction process on the corrected projection data is lower than the clarity of the artifacts in the image to be corrected.

[0108] Step 206: Generate a target corrected image corresponding to the projection data to be corrected based on the first corrected image and the second corrected image.

[0109] In an embodiment of the present application, the medical imaging device may generate a target corrected image corresponding to the projection data to be corrected collected by the medical imaging device based on the first corrected image and the second corrected image.

[0110] It should be noted that there are multiple possible implementation manners for the medical imaging device to generate a target corrected image corresponding to the projection data to be corrected based on the first corrected image and the second corrected image. In the embodiments of the present application, the following two possible implementation manners are taken as examples for illustrative description:

[0111] In the first possible implementation manner, the medical imaging device generating a target corrected image corresponding to the projection data to be corrected based on the first corrected image and the second corrected image may include:

[0112] In the medical imaging device, among the first corrected image and the second corrected image, the corrected image with a higher degree of artifact elimination is determined as the target corrected image corresponding to the projection data to be corrected.

[0113] In the embodiments of the present application, since the errors of the projection data collected by the medical imaging device during each imaging process are different, the first corrected image is an image obtained by first reconstructing the projection data to be corrected and then eliminating artifacts using the first correction model, and the second corrected image is an image obtained by first eliminating the artifact data in the projection data to be corrected using the second correction model and then reconstructing the projection data to be corrected with the artifact data eliminated. Therefore, the degree of artifact elimination of the first corrected image may be different from that of the second corrected image. The medical imaging device can determine the corrected image with a higher degree of artifact elimination as the target corrected image corresponding to the projection data to be corrected, thereby improving the quality of the image obtained by the medical imaging device for clinical treatment.

[0114] It should be noted that in the first corrected image and the second corrected image, the method for determining the corrected image with a higher degree of artifact elimination can refer to the corresponding content in the following embodiments, and the embodiments of the present application will not elaborate here.

[0115] In a second possible implementation, please refer to Figure 3 , Figure 3 which is a flowchart of a method for generating a target corrected image corresponding to the projection data to be corrected provided by the embodiments of the present application. Based on the first corrected image and the second corrected image, the medical imaging device generates a target corrected image corresponding to the projection data to be corrected, which may include:

[0116] Step 2061: Perform image block extraction processing on the first corrected image to obtain a plurality of first sub-images.

[0117] In the embodiments of the present application, the medical imaging device may perform image block extraction processing on the first corrected image to obtain a plurality of first sub-images.

[0118] Exemplarily, as Figure 4 shown, Figure 4 is an effect diagram of obtaining a plurality of first sub-images after performing image block extraction processing on the first corrected image provided by the embodiments of the present application. After performing image block extraction processing on the first corrected image A, a plurality of first sub-images can be obtained. For example, the plurality of first sub-images may be: the first sub-image a1, the first sub-image a2, the first sub-image a3, and the first sub-image a4.

[0119] Step 2062: Perform image block extraction processing on the second corrected image to obtain a plurality of second sub-images.

[0120] In the embodiment of the present application, the medical imaging device may perform image block extraction processing on the second corrected image to obtain a plurality of second sub-images. Among them, the plurality of second sub-images correspond one-to-one to the plurality of first sub-images, and the position of each second sub-image in the second corrected image is the same as the position of the corresponding first sub-image in the first corrected image. In the present application, the area of the region where each second sub-image is located may be the same as the area of the region where the corresponding first sub-image is located.

[0121] Exemplarily, as Figure 5 shown, Figure 5 is an effect diagram of obtaining a plurality of second sub-images after performing image block extraction processing on the second corrected image provided by the embodiment of the present application. After performing image block extraction processing on the second corrected image B, a plurality of second sub-images can be obtained. For example, the plurality of second sub-images may be respectively: second sub-image b1, second sub-image b2, second sub-image b3, and second sub-image b4. Among them, the second sub-image b1 may correspond to the first sub-image a1, the second sub-image b2 may correspond to the first sub-image a2, the second sub-image b3 may correspond to the first sub-image a3, and the second sub-image b4 may correspond to the first sub-image a4.

[0122] Step 2063: In each first sub-image and the corresponding second sub-image, determine the image with a higher degree of artifact elimination.

[0123] In the embodiment of the present application, the medical imaging device determines the image with a higher degree of artifact elimination in each first sub-image and the corresponding second sub-image, and there are multiple exemplary implementation manners. The embodiment of the present application takes the following two exemplary implementation manners as examples for description:

[0124] In the first exemplary implementation manner, the medical imaging device determining the image with a higher degree of artifact elimination in each first sub-image and the corresponding second sub-image may include:

[0125] The medical imaging device inputs the first sub-image and the second sub-image into the selection model at the same time, and determines the image output by the selection model as the image with a higher degree of artifact elimination.

[0126] Among them, the selection model is used to select the image with a higher degree of artifact elimination from the received multiple sub-images. In this way, after the medical imaging device inputs the first sub-image and the second sub-image into the selection model at the same time, the selection model can select the image with a higher degree of artifact elimination from the first sub-image and the second sub-image, and output the selected sub-image. Optionally, the selection model may be an AI model obtained by learning and training using a deep neural network or a convolutional neural network, etc.

[0127] It should be noted that the first sub-image and the second sub-image input to the selection model simultaneously are two corresponding sub-images. For example, as Figure 4 and Figure 5 shown, the first sub-image a1 and the second sub-image b1 can be input to the selection model simultaneously. Suppose the sub-image output by the selection model is the first sub-image a1, then the first sub-image a1 can be determined as the image with a higher degree of artifact elimination.

[0128] It should also be noted that before inputting the first sub-image and the second sub-image to the selection model simultaneously and determining the image output by the selection model as the image with a higher degree of artifact elimination, the medical imaging device also needs to obtain the selection model. In this application, the medical imaging device needs to perform multiple first training processes on the binary classification model to obtain the selection model.

[0129] However, since a large amount of data needs to be processed in the process of training the binary classification model to obtain the selection model, in order to improve the training efficiency, in this application, a training server with stronger computing power can be used to train the binary classification model to obtain the selection model. After the training is completed, the selection model can be sent to the medical imaging device so that the medical imaging device can obtain the selection model. In this case, the training server can perform multiple first training processes on the binary classification model to obtain the selection model.

[0130] Exemplarily, each first training process may include:

[0131] Step A3: Input two images to the binary classification model simultaneously.

[0132] In this application, the two images may include an image with artifacts and an image without artifacts.

[0133] Step B3: Determine whether the image output by the binary classification model is an image without artifacts.

[0134] In this application, after inputting an image with artifacts and an image without artifacts to the binary classification model simultaneously, it can be determined whether the image output by the binary classification model is the image without artifacts. It should be noted that the image contents of the image with artifacts and the image without artifacts input to the binary classification model simultaneously may be the same or different. The embodiments of this application do not make any limitations in this regard.

[0135] Exemplarily, if the image output by the binary classification model is the image without artifacts, repeat Step A3; if the image output by the binary classification model is not the image without artifacts, perform the following Step C3.

[0136] Step C3: If the image output by the binary classification model is not an image without artifacts, adjust the model parameters of the binary classification model.

[0137] It should be noted that during multiple executions of the first training process on the binary classification model, if the cut-off condition is reached, the binary classification model after the last execution of the first training process can be determined as the selected model.

[0138] Optionally, the cut-off condition may include: executing a specified number of third training processes; or, when continuously executing the first training process k times, the image output by the binary classification model each time is the image without artifacts. Here, k is an integer greater than 1.

[0139] After training the binary classification model to obtain the selected model, the selected model can be used to select the sub-image with a higher degree of artifact elimination from the first sub-image and its corresponding second sub-image.

[0140] In the second exemplary implementation, for a medical imaging device to determine the image with a higher degree of artifact elimination in each first sub-image and its corresponding second sub-image, it may include:

[0141] Step A4: Determine the first artifact image corresponding to the first sub-image and the second artifact image corresponding to the second sub-image.

[0142] In the embodiments of the present application, the medical imaging device can determine the first artifact image corresponding to the first sub-image based on the first sub-image and the third sub-image, and determine the second artifact image corresponding to the second sub-image based on the second sub-image and the third sub-image.

[0143] Among them, the third sub-image is located within the image to be corrected, and the position of the third sub-image in the image to be corrected is the same as the position of the first sub-image in the first corrected image.

[0144] Exemplarily, since the first sub-image is an image after artifact removal and the third sub-image is an image before artifact removal, therefore, by subtracting the pixel values of each pixel point in the first sub-image from the pixel values of each pixel point in the third sub-image, the first artifact image corresponding to the first sub-image can be obtained. Similarly, since the second sub-image is an image after artifact removal, therefore, by subtracting the pixel values of each pixel point in the second sub-image from the pixel values of each pixel point in the third sub-image, the second artifact image corresponding to the second sub-image can be obtained.

[0145] Step B4: Determine the structural similarity index between the first artifact image and the third sub-image, and the structural similarity index between the second artifact image and the third sub-image.

[0146] In the embodiments of the present application, the medical imaging device can respectively determine the structural similarity index between the first artifact image and the third sub-image, and the structural similarity index between the second artifact image and the third sub-image.

[0147] It should be noted that the value range of the structural similarity index of two images is usually [0, 1]. The larger the value, the more similar the two images are.

[0148] Suppose the two images are respectively: the image before artifact elimination (for example, it can be the third sub-image) and the artifact image corresponding to the image after artifact elimination (for example, it can be the first sub-image or the second sub-image). Then, the larger the structural similarity index of the two images, the more similar the two images are, and the higher the degree of artifact elimination in the image after artifact elimination.

[0149] Step C4: In each first sub-image and the corresponding second sub-image, determine the sub-image corresponding to the artifact image with a larger structural similarity index to the third sub-image as the image with a higher degree of artifact elimination.

[0150] In the embodiments of the present application, the medical imaging device can, in each first sub-image and the corresponding second sub-image, determine the sub-image corresponding to the artifact image with a larger structural similarity index to the third sub-image as the image with a higher degree of artifact elimination.

[0151] For example, as Figure 4 and Figure 5 shown, suppose the structural similarity index between the first artifact image corresponding to the first sub-image a2 and the third sub-image is less than the structural similarity index between the first artifact image corresponding to the second sub-image b2 and the third sub-image. Then, the second sub-image b2 can be determined as the image with a higher degree of artifact elimination.

[0152] Step 2064: Generate a target correction image corresponding to the projection data to be corrected based on multiple images with a higher degree of artifact elimination.

[0153] In the embodiments of the present application, through the above step 2063, after the medical imaging device determines the images with a higher degree of artifact elimination in each first sub-image and the corresponding second sub-image, the medical imaging device can generate a target correction image corresponding to the projection data to be corrected based on multiple images with a higher degree of artifact elimination.

[0154] For example, as Figure 4 and Figure 5As shown, assume that in the first sub-image a1 and the second sub-image b1, the first sub-image a1 is determined as the image with a higher degree of artifact elimination; in the first sub-image a2 and the second sub-image b2, the second sub-image b2 is determined as the image with a higher degree of artifact elimination; in the first sub-image a3 and the second sub-image b3, the second sub-image b3 is determined as the image with a higher degree of artifact elimination; in the first sub-image a4 and the second sub-image b4, the first sub-image a4 is determined as the image with a higher degree of artifact elimination. Then, please refer to Figure 6 , Figure 6 is an effect diagram of a target correction image corresponding to the to-be-corrected projection data provided by an embodiment of the present application. Based on the first sub-image a1, the second sub-image b2, the second sub-image b3, and the first sub-image a4, a target correction image C corresponding to the to-be-corrected projection data can be generated. The target correction image C is composed of the first sub-image a1, the second sub-image b2, the second sub-image b3, and the first sub-image a4.

[0155] In this case, in the process of generating a target correction image corresponding to the to-be-corrected projection data based on the first correction image and the second correction image, the artifact elimination degree in each region of the first correction image and the artifact elimination image in each region of the second correction image are comprehensively considered. The image in each region of the target correction image is: the image with a higher degree of artifact elimination among the image in the corresponding region of the first correction image and the image in the corresponding region of the second correction image, so that the artifact elimination degree in each region of the target correction image is relatively high, further improving the quality of the image obtained by the medical imaging device for clinical treatment.

[0156] It should be noted that the order of the steps of the image processing method provided by the embodiments of the present application can be appropriately adjusted, and the steps can also be increased or decreased accordingly according to the situation. Any method that can be easily thought of by those skilled in the art within the technical scope disclosed in the present application should be covered within the protection scope of the present application, so it will not be elaborated here.

[0157] In summary, for the image processing method provided in the embodiments of the present application, after the medical imaging device acquires the projection data to be corrected, the projection data to be corrected can be reconstructed to obtain an image to be corrected, and the first correction model is used to eliminate the artifacts in the image to be corrected to obtain a first corrected image. At the same time, the second correction model can be used to eliminate the data corresponding to the artifacts in the projection data to be corrected, and the projection data after the elimination of the artifact data is reconstructed to obtain a second corrected image. Subsequently, after generating a target corrected image corresponding to the projection data to be corrected acquired by the medical imaging device based on the first corrected image and the second corrected image, the images in each region of the target corrected image are all: the images in the corresponding regions of the first corrected image and the second corrected image, with a higher degree of artifact elimination, such that the degree of artifact elimination in each region of the target corrected image is relatively high, and the probability of the target corrected image containing artifacts is relatively low, thereby effectively improving the quality of the images obtained by the medical imaging device for clinical treatment.

[0158] The embodiments of the present application further provide an image processing apparatus, which can be integrated into a medical imaging device. As Figure 7 shown, Figure 7 FIG. is a structural block diagram of an image processing apparatus provided by an embodiment of the present application. The image processing apparatus 300 may include:

[0159] An acquisition module 301, configured to acquire projection data to be corrected collected by a medical imaging device.

[0160] A first reconstruction module 302, configured to reconstruct the projection data to be corrected to obtain an image to be corrected.

[0161] A first correction module 303, configured to input the image to be corrected into a first correction model to obtain a first corrected image output by the first correction model, where the first correction model is used to eliminate the artifacts in the input image to be corrected.

[0162] A second correction module 304, configured to input the projection data to be corrected into a second correction model to obtain the corrected projection data output by the second correction model.

[0163] A second reconstruction module 305, configured to reconstruct the corrected projection data to obtain a second corrected image, where the second correction model is used to eliminate the data corresponding to the artifacts in the input projection data to be corrected.

[0164] A generation module 306, configured to generate a target corrected image corresponding to the projection data to be corrected based on the first corrected image and the second corrected image.

[0165] Optionally, as Figure 8 shown, Figure 8It is a structural block diagram of a generation module provided by an embodiment of the present application. The generation module 306 may include:

[0166] A first image block extraction unit 3061, configured to perform image block extraction processing on the first corrected image to obtain a plurality of first sub-images.

[0167] A second image block extraction unit 3062, configured to perform image block extraction processing on the second corrected image to obtain a plurality of second sub-images that correspond one-to-one to the plurality of first sub-images. The position of each second sub-image in the second corrected image is the same as the position of the corresponding first sub-image in the first corrected image.

[0168] A determination unit 3063, configured to determine an image with a higher degree of artifact elimination in each first sub-image and the corresponding second sub-image.

[0169] A generation unit 3064, configured to generate a target corrected image corresponding to the projection data to be corrected based on the plurality of images with a higher degree of artifact elimination.

[0170] Optionally, the determination unit 3063 is configured to: input the first sub-image and the second sub-image into the selection model simultaneously, and determine the image output by the selection model as the image with a higher degree of artifact elimination. The selection model is used to select an image with a higher degree of artifact elimination from the received plurality of sub-images.

[0171] Optionally, the image processing device 300 may further include: a first training module, configured to: perform a plurality of first training processes on the binary classification model to obtain the selection model.

[0172] Wherein, each first training process includes: inputting two images into the binary classification model simultaneously, the two images including an image with artifacts and an image without artifacts; determining whether the image output by the binary classification model is an image without artifacts; if the image output by the binary classification model is not an image without artifacts, adjusting the model parameters of the binary classification model.

[0173] Optionally, the determination unit 3063 is configured to: determine a first artifact image corresponding to the first sub-image and a second artifact image corresponding to the second sub-image; determine the structural similarity index between the first artifact image and a third sub-image, and the structural similarity index between the second artifact image and the third sub-image. The third sub-image is located in the image to be corrected, and the position of the third sub-image in the image to be corrected is the same as the position of the first sub-image in the first corrected image; in each first sub-image and the corresponding second sub-image, determine the sub-image corresponding to the artifact image with a larger structural similarity index to the third sub-image as the image with a higher degree of artifact elimination.

[0174] Optionally, the determination unit 3063 is configured to: determine a first artifact image based on the first sub-image and the third sub-image; determine a second artifact image based on the second sub-image and the third sub-image.

[0175] Optionally, the generation module 306 is configured to: determine, from the first corrected image and the second corrected image, the corrected image with a higher degree of artifact elimination as the target corrected image corresponding to the projection data to be corrected.

[0176] Optionally, the image processing apparatus 300 may further include: a second training module configured to: perform a plurality of second training processes on the first artifact elimination model to obtain a first correction model.

[0177] Wherein, each second training process includes: inputting an image containing artifacts into the first artifact elimination model to obtain an image output by the first artifact elimination model; adjusting the model parameters of the first artifact elimination model based on the image output by the first artifact elimination model and the image without artifacts.

[0178] Optionally, the image processing apparatus 300 may further include: a third training module configured to: perform a plurality of third training processes on the second artifact elimination model to obtain a second correction model.

[0179] Wherein, each third training process includes: inputting a projection data containing artifact data into the second artifact elimination model to obtain a projection data output by the second artifact elimination model; adjusting the model parameters of the second artifact elimination model based on the projection data output by the second artifact elimination model and the projection data without artifact data.

[0180] In summary, after the image processing apparatus provided in the embodiments of the present application acquires the projection data to be corrected by a medical imaging device, it can perform reconstruction processing on the projection data to be corrected to obtain a to-be-corrected graph, and use the first correction model to eliminate artifacts in the to-be-corrected image to obtain a first corrected image; at the same time, it can use the second correction model to eliminate the data corresponding to the artifacts in the projection data to be corrected, and perform reconstruction processing on the projection data after the artifact data is eliminated to obtain a second corrected image. Subsequently, after generating a target corrected image corresponding to the projection data to be corrected acquired by the medical imaging device based on the first corrected image and the second corrected image, the images in each region of the target corrected image are all: the images in the corresponding regions of the first corrected image and the second corrected image, the images with a higher degree of artifact elimination, so that the degree of artifact elimination in each region of the target corrected image is higher, and the probability of the target corrected image containing artifacts is lower, thereby effectively improving the quality of the image obtained by the medical imaging device for clinical treatment.

[0181] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the devices, modules, and units described above can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0182] In practical applications, each module and each unit under the module in this embodiment can be implemented by devices such as a central processing unit (CPU), a microprocessor (MPU), a digital signal processor (DSP), or a field programmable gate array (FPGA) and a modem on an image processing device.

[0183] The embodiment of the present application also provides a medical imaging device. The medical imaging device may include: a processor, and a memory for storing executable instructions of the processor. Wherein, the processor is configured to execute Figure 1 or Figure 2 the image processing method shown.

[0184] The embodiment of the present application also provides a computer-readable storage medium. Instructions are stored in the computer-readable storage medium, and when the computer-readable storage medium runs on a processing component, the processing component is caused to execute Figure 1 or Figure 2 the image processing method shown.

[0185] In the present application, the terms "first" and "second" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance. The term "plurality" refers to two or more, unless otherwise clearly defined.

[0186] Those of ordinary skill in the art can understand that all or part of the steps of implementing the above embodiments can be completed by hardware, or can be completed by a program instructing relevant hardware. The program can be stored in a computer-readable storage medium. The storage medium mentioned above can be a read-only memory, a disk, an optical disc, etc.

[0187] The above are only optional embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. An image processing method, characterized in that, The method includes: Obtaining projection data to be corrected collected by a medical imaging device; Performing reconstruction processing on the projection data to be corrected to obtain an image to be corrected, and inputting the image to be corrected into a first correction model to obtain a first corrected image output by the first correction model, where the first correction model is used to eliminate artifacts in the input image to be corrected; Inputting the projection data to be corrected into a second correction model to obtain corrected projection data output by the second correction model, and performing reconstruction processing on the corrected projection data to obtain a second corrected image, where the second correction model is used to eliminate data corresponding to artifacts in the input projection data to be corrected; Generating a target corrected image corresponding to the projection data to be corrected based on the first corrected image and the second corrected image.

2. The method according to claim 1, characterized in that, Generating a target corrected image corresponding to the projection data to be corrected based on the first corrected image and the second corrected image, including: Performing image block extraction processing on the first corrected image to obtain a plurality of first sub-images; Performing image block extraction processing on the second corrected image to obtain a plurality of second sub-images corresponding one-to-one to the plurality of first sub-images, where the position of each second sub-image in the second corrected image is the same as the position of the corresponding first sub-image in the first corrected image; Determining, in each of the first sub-images and the corresponding second sub-image, an image with a higher degree of artifact elimination; Generating a target corrected image corresponding to the projection data to be corrected based on the plurality of images with a higher degree of artifact elimination.

3. The method according to claim 2, characterized in that, Determining, in each of the first sub-images and the corresponding second sub-image, an image with a higher degree of artifact elimination, including: Inputting the first sub-image and the second sub-image into a selection model simultaneously, and determining the image output by the selection model as the image with a higher degree of artifact elimination, where the selection model is used to select an image with a higher degree of artifact elimination from the received plurality of sub-images.

4. The method according to claim 3, characterized in that, Before determining, in each of the first sub-images and the corresponding second sub-image, an image with a higher degree of artifact elimination, the method further includes: Performing a plurality of first training processes on a binary classification model to obtain the selection model; Wherein, each first training process includes: Inputting two images into the binary classification model simultaneously, where the two images include an image with artifacts and an image without artifacts; Determining whether the image output by the binary classification model is the image without artifacts; If the image output by the binary classification model is not the image without artifacts, adjusting the model parameters of the binary classification model.

5. The method according to claim 2, characterized in that, Determining, in each of the first sub-images and the corresponding second sub-image, an image with a higher degree of artifact elimination, including: Determining a first artifact image corresponding to the first sub-image and a second artifact image corresponding to the second sub-image; Determine the structural similarity index between the first artifact image and the third sub-image, and the structural similarity index between the second artifact image and the third sub-image. The third sub-image is located within the image to be corrected, and the position of the third sub-image in the image to be corrected is the same as the position of the first sub-image in the first corrected image; In each of the first sub-images and the corresponding second sub-images, determine the sub-image corresponding to the artifact image with a larger structural similarity index to the third sub-image as the image with a higher degree of artifact elimination.

6. The method according to claim 5, characterized in that, Determine the first artifact image corresponding to the first sub-image and the second artifact image corresponding to the second sub-image, including: Based on the first sub-image and the third sub-image, determine the first artifact image; Based on the second sub-image and the third sub-image, determine the second artifact image.

7. The method according to claim 1, characterized in that, Generate a target corrected image corresponding to the projection data to be corrected based on the first corrected image and the second corrected image, including: In the first corrected image and the second corrected image, determine the corrected image with a higher degree of artifact elimination as the target corrected image corresponding to the projection data to be corrected.

8. The method according to any one of claims 1 to 7, characterized in that, Before inputting the image to be corrected into the first correction model to obtain the first corrected image output by the first correction model, the method further includes: Perform multiple second training processes on the first artifact elimination model to obtain the first correction model; Wherein, each second training process includes: Input an image containing artifacts into the first artifact elimination model to obtain the image output by the first artifact elimination model; Based on the image output by the first artifact elimination model and the image without artifacts, adjust the model parameters of the first artifact elimination model.

9. The method according to any one of claims 1 to 7, characterized in that Before inputting the projection data to be corrected into the second correction model to obtain the corrected projection data output by the second correction model, the method further includes: Perform multiple third training processes on the second artifact elimination model to obtain the second correction model; Wherein, each third training process includes: Input a projection data containing artifact data into the second artifact elimination model to obtain the projection data output by the second artifact elimination model; Based on the projection data output by the second artifact elimination model and the projection data without artifact data, adjust the model parameters of the second artifact elimination model.

10. An image processing apparatus, characterized in that The apparatus includes: An acquisition module, configured to acquire projection data to be corrected collected by a medical imaging device; A first reconstruction module, configured to perform reconstruction processing on the projection data to be corrected to obtain an image to be corrected; A first correction module, configured to input the image to be corrected into a first correction model to obtain a first corrected image output by the first correction model, where the first correction model is used to eliminate artifacts in the input image to be corrected; A second correction module, configured to input the projection data to be corrected into a second correction model to obtain the corrected projection data output by the second correction model; A second reconstruction module, configured to perform a reconstruction process on the corrected projection data to obtain a second corrected image, where the second correction model is used to eliminate data corresponding to artifacts in the input projection data to be corrected; A generation module, configured to generate a target corrected image corresponding to the projection data to be corrected based on the first corrected image and the second corrected image.

11. A medical imaging device, characterized in that Comprising: A processor, and a memory for storing executable instructions of the processor; wherein, the processor is configured to execute the image processing method according to any one of claims 1 to 9.

12. A computer-readable storage medium, characterized in that Instructions are stored in the computer-readable storage medium, and when the readable storage medium runs on the processing component, the processing component is caused to execute the image processing method according to any one of claims 1 to 9.

Images