Artifact correction method, computer device and readable storage medium

Abstract

The application relates to an artifact correction method, a computer device and a computer readable storage medium. The method comprises the following steps: acquiring preoperative image data; performing artifact identification on the preoperative image data by using a pre-trained artifact identification model to obtain a target artifact region; acquiring attribute information of the target artifact region; when it is determined according to the attribute information that the target artifact region needs to be subjected to artifact correction, performing artifact correction on the target artifact region needing to be subjected to artifact correction by using a pre-trained artifact correction model. The method can improve processing efficiency.

Description

Technical Field

[0001] This application relates to the field of medical image processing technology, and in particular to an artifact correction method, computer equipment, and readable storage medium. Background Technology

[0002] Computed tomography (CT) is an advanced medical imaging technique that uses X-ray beams to scan specific areas of the body and reconstruct lesions, providing crucial information for diagnosis. During preoperative surgical planning, metal artifacts can occur due to the presence of metals in the patient's body. For example, because these metals are high-density materials, their presence causes significant attenuation of X-rays during scanning, appearing as bright or dark stripes or radial areas in CT images. These metal artifacts can negatively impact the effectiveness of preoperative surgical planning.

[0003] Traditional techniques eliminate the effects of metal artifacts by correcting them. However, current direct correction methods result in a large amount of data processing and reduced efficiency. Summary of the Invention

[0004] Therefore, it is necessary to provide an artifact correction method, device, and storage medium that can improve processing efficiency to address the aforementioned technical problems.

[0005] In a first aspect, this application provides an artifact correction method, the method comprising:

[0006] Obtain preoperative imaging data;

[0007] The pre-trained artifact recognition model is used to identify artifacts in the preoperative image data to obtain the target artifact region.

[0008] Obtain the attribute information of the target artifact region;

[0009] When it is determined from the attribute information that the target artifact region needs artifact correction, the artifact correction model obtained in the pre-trained model is used to perform artifact correction on the target artifact region that needs artifact correction.

[0010] In one embodiment, the feature is that, after the artifact correction model obtained through pre-training performs artifact correction on the target artifact region requiring artifact correction, the process includes:

[0011] Obtain the attribute information of the corrected artifact region, take the corrected artifact region as the target artifact region, and continue to execute the step of determining whether the target artifact region needs to be corrected based on the attribute information, until it is determined based on the attribute information that the target artifact region does not need to be corrected.

[0012] In one embodiment, obtaining the attribute information of the target artifact region includes:

[0013] At least one of the following attributes is obtained as attribute information: the number of target artifact regions, the region location, the region volume ratio, and the region gray value intensity distribution;

[0014] When it is determined from the attribute information that the target artifact region needs artifact correction, before performing artifact correction on the target artifact region that needs artifact correction using a pre-trained artifact correction model, the method further includes:

[0015] Determine the weight corresponding to each attribute information and the reference index value corresponding to each attribute information;

[0016] Calculate the correction index value of the target artifact region based on the weight and the reference index value;

[0017] Determine whether the target artifact region needs artifact correction based on the correction index value.

[0018] In one embodiment, the preoperative image data is used to identify artifacts using a pre-trained artifact recognition model to obtain the target artifact region, including:

[0019] The feature extraction module of the pre-trained artifact recognition model extracts features from the preoperative image data to obtain the features to be processed.

[0020] The features to be processed are respectively input into the center point prediction module, center point offset prediction module and target object length, width and height prediction module of the artifact recognition model for processing to obtain the corresponding artifact region information.

[0021] The target artifact region is calculated based on the information of each artifact region.

[0022] In one embodiment, the preoperative imaging data is a CT image sequence; the artifact recognition model is a three-dimensional convolutional neural network model; and / or the artifact correction model is a three-dimensional convolutional neural network model.

[0023] In one embodiment, the training method for the artifact recognition model or artifact correction model includes:

[0024] Acquire sample medical image data, wherein the sample medical image data carries tags;

[0025] The initial model is used to process the sample medical image data to obtain the model processing result;

[0026] The model loss is calculated based on the labels and the model processing results.

[0027] When the model loss does not meet the requirements, the network parameters of the initial model are optimized, and the sample medical image data are processed by the optimized model to obtain the model processing result. The step of calculating the model loss based on the label and the model processing result continues until the model loss meets the requirements, thus obtaining the artifact recognition model or artifact correction model.

[0028] In one embodiment, the label is a pre-labeled artifact region; the step of processing the sample medical image data using an initial model to obtain the model processing result includes:

[0029] The initial model's feature extraction module extracts features from the sample medical image data to obtain sample features; the sample features are then input into the initial model's center point prediction module, center point offset prediction module, and target object length, width, and height prediction module for processing to obtain the corresponding sample artifact region information;

[0030] The calculation of the model loss based on the label and the model processing result includes:

[0031] The position of the first center point and the first length, width and height information are obtained based on the pre-marked artifact region;

[0032] The first loss value is calculated by comparing the artifact region information obtained by the center point prediction module with the first center point position of the pre-labeled artifact region.

[0033] The second loss value is obtained by calculating the position of the first center point of the artifact region pre-labeled by the artifact region information obtained from the center point offset prediction module.

[0034] The third loss value is calculated by combining the artifact region information obtained by the length, width and height prediction module of the target object with the first length, width and height information of the pre-labeled artifact region.

[0035] The model loss is calculated based on the first loss value, the second loss value, and the third loss value.

[0036] In one embodiment, the step of calculating a first loss value by comparing the artifact region information obtained by the center point prediction module with the first center point position of the pre-labeled artifact region includes:

[0037] Based on the pre-labeled artifact regions, positive and negative voxels of the samples are determined. The artifact region information is processed by the center point prediction module to obtain the predicted positive voxels, predicted negative voxels, and the second center point position. The first loss of the positive voxels and the predicted positive voxels is calculated, the second loss of the negative voxels and the predicted negative voxels is calculated, and the third loss of the first center point position and the second center point position is calculated. Based on the first loss, the second loss, and the third loss, the first loss value corresponding to the center point prediction module is calculated.

[0038] The step of calculating the second loss value by taking the first center point position of the artifact region pre-annotated by the artifact region information obtained from the center point offset prediction module includes:

[0039] The predicted offset value is determined by the center point offset prediction module, the true offset value is determined by the pre-labeled artifact region, and the second loss value corresponding to the center point offset prediction module is calculated based on the predicted offset value, the true offset value and the number of center points.

[0040] The step of calculating the third loss value by combining the artifact region information obtained from the length, width, and height prediction module of the target object with the first length, width, and height information of the pre-labeled artifact region includes:

[0041] The second length, width, and height information is determined by the length, width, and height prediction module of the target object. Based on the first length, width, and height information, the second length, width, and height information, and the number of center points, the third loss value corresponding to the length, width, and height prediction module of the target object is calculated.

[0042] In one embodiment, the label is a matched artifact-free region; the calculation of the model loss based on the label and the model processing result includes:

[0043] Determine the corresponding pixel pairs of the model processing results and the matched artifact-free regions;

[0044] The model loss is calculated based on the pixel values ​​of the pixel pairs.

[0045] Secondly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method in any of the above embodiments.

[0046] Thirdly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the methods in any of the above embodiments.

[0047] The aforementioned artifact correction method, computer equipment, and computer-readable storage medium, after acquiring preoperative image data, first use a pre-trained artifact recognition model to identify artifacts in the preoperative image data to determine the target artifact region. Then, based on the attribute information of the target artifact region, it determines whether artifact correction is needed. Only the target artifact region that needs artifact correction is corrected. This way, not all artifacts need to be corrected, reducing the amount of data processing and improving processing efficiency. Attached Figure Description

[0048] Figure 1 This is a diagram illustrating the application environment of an artifact correction method in one embodiment.

[0049] Figure 2 This is a flowchart illustrating an artifact correction method in one embodiment;

[0050] Figure 3 A flowchart of an artifact correction method in another embodiment;

[0051] Figure 4 This is a schematic diagram showing the cross-sectional, coronal, and sagittal plane positions in one embodiment;

[0052] Figure 5 This is a schematic diagram of an artifact level of H in one embodiment;

[0053] Figure 6 This is a schematic diagram of an artifact level of M in one embodiment;

[0054] Figure 7 This is a schematic diagram of an artifact level of L in one embodiment;

[0055] Figure 8 This is a schematic diagram of a CT image sequence in one embodiment;

[0056] Figure 9 This is a comparative illustration of three-dimensional convolution and two-dimensional convolution in one embodiment;

[0057] Figure 10 This is a flowchart illustrating the target model training method in one embodiment;

[0058] Figure 11 This is a schematic diagram of the structure of the target model in one embodiment;

[0059] Figure 12 This is a schematic diagram of the encoder structure in one embodiment;

[0060] Figure 13 This is a schematic diagram of the decoder structure in one embodiment;

[0061] Figure 14This is a flowchart illustrating the training process of an artifact recognition model in one embodiment.

[0062] Figure 15 This is a structural entity diagram of an artifact recognition model in one embodiment;

[0063] Figure 16 This is a flowchart illustrating the training process of an artifact correction model in one embodiment.

[0064] Figure 17 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0065] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0066] The artifact correction method provided in this application can be applied to, for example... Figure 1 In the application environment shown, terminal 102 communicates with medical imaging device 104 via a network. A data storage system can store the data that terminal 102 needs to process. The data storage system can be integrated onto terminal 102 or placed in the cloud or on another network server.

[0067] Terminal 102 acquires preoperative image data from medical imaging device 104, and then uses a pre-trained artifact recognition model to identify artifacts in the preoperative image data to obtain target artifact regions. It also extracts attribute information of the target artifact regions. When the attribute information determines that a target artifact region requires artifact correction, the pre-trained artifact correction model is used to correct the artifact in that region. Therefore, artifact correction is only performed on target artifact regions that require correction, eliminating the need to correct all artifacts, reducing data processing volume, and improving processing efficiency.

[0068] The terminal 102 may be, but is not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices. The medical imaging device 104 includes, but is not limited to, various imaging devices, such as CT imaging devices (CT: Computed Tomography, which uses a precisely collimated X-ray beam and a highly sensitive detector to perform a series of cross-sectional scans around a part of the human body, and can reconstruct precise three-dimensional images of tumors, etc. through CT scans), magnetic resonance imaging devices (which are a type of tomographic imaging that uses the magnetic resonance phenomenon to obtain electromagnetic signals from the human body and reconstruct human body information images), positron emission tomography (PET) devices, positron emission tomography (PET / MR) systems, etc.

[0069] In one embodiment, such as Figure 2 As shown, an artifact correction method is provided, which is applied to... Figure 1 Taking the terminal in the example, the explanation includes the following steps:

[0070] S202: Obtain preoperative imaging data.

[0071] Specifically, preoperative imaging data refers to images of the target area acquired before surgery, such as the surgical area, or preoperative medical imaging data acquired using medical imaging equipment. Specifically, this preoperative imaging data can be joint imaging data, dental imaging data, etc., without further specific limitations. Preoperative imaging data can be image sequences, such as CT image sequences, which allow for the extraction of depth information.

[0072] The preoperative imaging data is primarily used for generating surgical plans based on this data before surgery. However, to mitigate the impact of artifact regions, these regions are first identified and corrected to prevent inaccurate surgical planning caused by artifacts.

[0073] S204: Use a pre-trained artifact recognition model to identify artifacts in preoperative image data to obtain the target artifact region.

[0074] Specifically, the pre-trained artifact recognition model is used to identify artifacts in preoperative image data to determine the target artifact region. This artifact recognition model can be a 3D convolutional neural network model, so the identified target artifact region is a 3D region; for example, the identified target artifact region is surrounded by a 3D bounding box. Specific limitations of this artifact recognition model are detailed below and will not be repeated here.

[0075] The target artifact region refers to the area in the preoperative imaging data where metallic artifacts are present.

[0076] Optionally, the terminal first preprocesses the preoperative image data and then uses an artifact recognition model for identification. Preprocessing includes, but is not limited to, adjusting window width and window level to enable the artifact model to process the preoperative image data in a targeted manner.

[0077] S206: Obtain the attribute information of the target artifact region.

[0078] Specifically, the attribute information includes, but is not limited to, the number of target artifact regions, their location, their volume percentage, and their grayscale intensity distribution. The location of a region can be represented by the position of its center point and its length, width, and height. The volume percentage of the artifact region is calculated as the intersection volume of the artifact region and the surgical selection region divided by the volume of the surgical selection region. The grayscale intensity distribution is the grayscale intensity distribution of the artifact region.

[0079] S208: When it is determined from the attribute information that the target artifact region needs to be corrected, the artifact correction model obtained in the pre-trained model is used to correct the artifacts in the target artifact region that needs to be corrected.

[0080] Specifically, the terminal determines whether the artifact region needs correction based on the attribute information of the artifact region. If correction is required, the artifact correction model is used for correction; otherwise, the artifact region is not processed.

[0081] The determination of whether a region needs correction based on its attribute information can be achieved by using a weighted average of multiple attributes. In other embodiments, calculations can also be performed based on a single attribute; this is not specifically limited here. Optionally, the type of attribute information can be preset by the user or selected based on the system's computational load; this is not specifically limited here either.

[0082] The pre-trained artifact correction model is used to correct artifacts in artifact regions. This model can be a 3D convolutional neural network, where the input target artifact region to be corrected is a 3D region; for example, if the input is a CT image sequence, the output is the corrected image sequence, also 3D data. Specific limitations of this artifact correction model are detailed below and will not be repeated here.

[0083] The aforementioned artifact correction method, after acquiring preoperative image data, first uses a pre-trained artifact recognition model to identify artifacts in the preoperative image data to determine the target artifact region. Then, based on the attribute information of the target artifact region, it determines whether artifact correction is needed. Only the target artifact region that needs artifact correction is corrected. This way, not all artifacts need to be corrected, reducing the amount of data processing and improving processing efficiency.

[0084] In one embodiment, after the target artifact region that needs artifact correction is corrected by the pre-trained artifact correction model, the process includes: obtaining the attribute information of the corrected artifact region, taking the corrected artifact region as the target artifact region, and continuing to execute the step of determining whether the target artifact region needs artifact correction based on the attribute information, until it is determined based on the attribute information that the target artifact region does not need artifact correction.

[0085] Specifically, in combination Figure 3 As shown, Figure 3 The flowchart below shows an artifact correction method in another embodiment, in which the aforementioned processing procedure is the same as... Figure 2 The embodiments shown are basically the same. After artifact correction, the attribute information of the corrected artifact region is also obtained, and the attribute information of the corrected artifact region is used to determine whether the corrected artifact region still needs artifact correction. If so, artifact correction continues until the attribute information determines that the target artifact region does not need artifact correction. In this way, the accuracy of the final artifact correction is ensured by multiple corrections.

[0086] In the above embodiments, multiple corrections ensure the accuracy of the final artifact correction, thus improving the existing preoperative planning process and expanding the applicability of image-guided surgical planning. Artifact recognition and scoring enhance the intelligence and automation of preoperative planning risk analysis. Correcting artifacts with low-risk scores improves image quality and enhances the reliability and accuracy of preoperative planning.

[0087] In one embodiment, obtaining attribute information of the target artifact region includes: obtaining at least one of the following as attribute information: the number of target artifact regions, region location, region volume ratio, and region grayscale intensity distribution; determining whether the target artifact region needs artifact correction based on the attribute information includes: determining the weight corresponding to each attribute information and the reference index value corresponding to each attribute information; calculating the correction index value of the target artifact region based on the weight and the reference index value; and determining whether the target artifact region needs artifact correction based on the correction index value.

[0088] Specifically, after identifying the target artifact region from the preoperative image data, the attribute information of the target artifact region is obtained. This attribute information includes, but is not limited to, the number of target artifact regions, the region location, the region volume ratio, and the region grayscale intensity distribution. The number of artifact regions is the number of artifacts contained in the preoperative image data. The region location can refer to the approximate location of the region, such as on the left leg, right leg, or whether it conflicts with the surgical area. The volume ratio of the artifact region is equal to the intersection volume of the artifact region and the surgical selection region / the volume of the surgical selection region. The region grayscale intensity distribution is the grayscale intensity distribution of the artifact region. This embodiment does not limit the type of the above attribute information. Those skilled in the art can obtain any type of attribute information of the artifact region as needed.

[0089] When determining whether a target artifact region needs artifact correction, a weighting method can be used to assign weights to each attribute information to obtain a level. The level, i.e., the correction index value, is then used to determine whether the target artifact region needs artifact correction.

[0090] To facilitate understanding, this embodiment uses two factors—"the location of the artifact" and "the volume percentage of the region where the artifact is located"—as examples to quantify and unify the scoring criteria.

[0091] Specifically, in combination Figure 4 As shown, the patient's CT image is displayed in three sections: coronal, sagittal, and transverse. This includes the "target artifact region" identified by the artifact recognition model and the patient's "surgical area" selected by the doctor.

[0092] Specifically, in combination Figures 5 to 7 As shown, the reference index value analysis for "the location of artifacts" is as follows: Figure 5 or Figure 6 As shown, if the artifact region is within the surgical selection area or the two intersect, it means that the artifact region will directly affect the subsequent planning of the surgical area. The reference index value is set to "1". Figure 7 As shown, if the identified artifact area is not within the surgical selection area, it means that the artifact area will directly affect the subsequent planning of the surgical area, and the reference index value is set to "0". Other values ​​can also be selected for the reference index value, which are not limited here.

[0093] Analysis of reference index value for "volume percentage of the region where artifacts are located": Volume percentage of the region where artifacts are located = Intersection volume of the artifact region and the surgical selection region / Volume of the surgical selection region.

[0094] like Figure 5 The volume percentage of the artifact region shown is 0.6, as indicated. Figure 6The volume percentage of the artifact region shown is 0.2, as indicated. Figure 7 The volume percentage of the artifact region shown is 0.

[0095] The weighting factors represent the degree of influence of each attribute information on the correction index value. In this example, the two attribute information each account for 50%. Figure 5 The final correction index value is calculated as follows: 1 (the location of the artifact affects the result) * 0.5 (the weight of this factor) + 0.6 (the volume percentage of the region where the artifact is located) * 0.5 (the weight of this factor) = 0.8. In other embodiments, the weights can be set in other ways, which are not specifically limited here.

[0096] The final correction index values ​​are set into three levels: H (representing high risk, set to this level if the result of step 4 is between 0.7 and 1); M (representing medium risk, set to this level if the result of step 4 is between 0.5 and 0.7); and L (representing low risk, set to this level if the result of step 4 is between 0 and 0.5). H indicates that the existing artifacts will have a significant impact on the subsequent surgical planning, and it is not recommended for the surgeon to continue with subsequent surgical planning. M indicates that the existing artifacts will have a moderate impact on the subsequent surgical planning, and it is recommended that the surgeon perform artifact correction before continuing with subsequent surgical planning. L indicates that the existing artifacts will have a minimal impact on the subsequent surgical planning, and no artifact correction is required before continuing with subsequent surgical planning. In this embodiment, this level setting is only for distance; in other embodiments, other values ​​can be used, and no specific limitation is made here. Figure 5 The artifact level shown is H (high risk). For example... Figure 6 The artifact level shown is M (medium risk). For example... Figure 7 The artifact level shown is L (low risk). The terminal will automatically calculate the level based on the three-dimensional coordinate relationship between the "artifact area" and the "surgical area" according to the set scoring criteria.

[0097] In the above embodiments, the correction index value is determined by the attribute information of the target artifact region, and the need for correction is determined based on the correction index value. The judgment is made by using quantitative standards, which is more accurate.

[0098] In one embodiment, the pre-trained artifact recognition model is used to identify artifacts in preoperative image data to obtain the target artifact region. This includes: extracting features from the preoperative image data using the feature extraction module of the pre-trained artifact recognition model to obtain features to be processed; inputting the features to be processed into the center point prediction module, center point offset prediction module, and length, width, and height prediction module of the artifact recognition model for processing to obtain corresponding artifact region information; and calculating the target artifact region based on the artifact region information.

[0099] Specifically, in this embodiment, the artifact recognition model includes a feature extraction module, a center point prediction module, a center point offset prediction module, and a target object length, width, and height prediction module. First, preoperative image data is input into the feature extraction module to obtain the features to be processed. Then, the features to be processed are input into the center point prediction module, the center point offset prediction module, and the target object length, width, and height prediction module to obtain the artifact region information calculated by each module. Finally, the target artifact region is calculated based on the artifact region information.

[0100] The artifact region information includes, but is not limited to, the position of the center point and the length, width and height of the artifact region. The target artifact region is calculated by mathematical statistical values ​​of the artifact region information obtained by the center point prediction module, the center point offset prediction module and the length, width and height prediction module of the target object, making the target artifact region more accurate.

[0101] In the above embodiments, features are extracted by constructing a three-dimensional fully convolutional backbone network. The center point and dimensions of the three-dimensional box of the region where the artifact is located are obtained by constructing a center point prediction (HeatMap) branch, a center point offset prediction branch (Offset) and a target object length, width and depth prediction branch (Height, Width and Depth). The position of the target in the original image is finally obtained by combining the prediction results of the three branches.

[0102] In one embodiment, the preoperative imaging data is a CT image sequence; the artifact recognition model is a three-dimensional convolutional neural network model, and / or the artifact correction model is a three-dimensional convolutional neural network model.

[0103] Specifically, in combination Figure 8 As shown, in Figure 8 Artifacts in CT images manifest as artifacts appearing across multiple layers within a single CT sequence. Furthermore, the morphology of bones and joints varies significantly across different cross-sectional planes in joint CT sequences. Figure 8 The image shows the morphological information of bones and joints at different levels, as well as information in the artifact regions. Information from a single CT scan is too limited; training the artifact recognition network and artifact correction model on a single scan is detrimental to model convergence and affects the overall accuracy. To fully consider the spatial information of artifact regions and bone and joint morphology, this embodiment constructs the joint CT sequence as follows: Figure 8 The three-dimensional volume shown is then subjected to subsequent network training operations such as three-dimensional convolution to obtain the final network model.

[0104] Specifically, in combination Figure 9In this embodiment, three-dimensional convolution is used for feature extraction. A CT image sequence consists of a series of two-dimensional tomographic images; from an overall perspective, it is three-dimensional data with spatial information. The knee joint CT image sequence adds another spatial dimension. Traditional two-dimensional convolution can only extract planar features from a single slice, resulting in the loss of spatial information in the image. Unlike two-dimensional convolution, three-dimensional convolution has a depth dimension in its input, which manifests as multiple continuous slices in CT images. Therefore, its convolution kernel also adds one dimension. The differences between two-dimensional and three-dimensional convolution are as follows: Figure 9 As shown.

[0105] In the above embodiments, in order to make full use of the three-dimensional spatial information of CT image data, both the artifact recognition model and the artifact correction model adopt a three-dimensional convolutional neural network model. The three-dimensional volume constructed from the joint CT sequence is subjected to subsequent three-dimensional convolution and other network training operations to obtain the final network model, which fully considers the spatial information of the artifact region and the shape of the target object.

[0106] In one embodiment, such as Figure 10 As shown, a target model training method is provided, which is then applied to... Figure 1 Taking the terminal in the example, the explanation includes the following steps:

[0107] S1002: Obtain sample medical image data, which carries labels.

[0108] Specifically, the sample medical image data refers to the collected sample data. When training the artifact recognition network for the artifact recognition model, it is necessary to first acquire the medical image dataset and complete data annotation, such as using software like 3D slicer to annotate the 3D target boxes where artifacts are located. Then, the dataset is divided into training, validation, and test sets, and data augmentation operations are performed on the data in the training set samples. When training the artifact correction network for the artifact correction model, it is necessary to first acquire the original medical image dataset and the corresponding artifact CT images. The artifact CT images can be obtained by simulating artifact generation. Then, the dataset is divided into training, validation, and test sets, and data augmentation operations are performed on the data in the training set samples.

[0109] S1004: The model processing results are obtained by processing the sample medical image data through the initial model.

[0110] The backbone network of the artifact recognition or artifact correction model can include an encoder structure, a decoder structure, and a transition section. For details, please refer to [link to relevant documentation]. Figure 11As shown, the initial model also includes the three parts mentioned above. The encoder structure is used to extract the low-level features of the task and perform feature dimensionality reduction. The number of encoders can be determined according to different projects. In this embodiment, the number is 4, but in other embodiments, it can be other numbers. The specific structure of the encoder can be found in [reference needed]. Figure 12 The encoder structure includes a 3D pooling layer, a 3D convolutional layer, and a residual module. The 3D pooling layer is used for feature dimensionality reduction; in this embodiment, a 3D convolution with a kernel size of 3x3x3 and a stride of 2 is used for dimensionality reduction. The 3D convolutional layer is used to extract low-level features. It consists of a 3D convolution operation with a kernel size of 3x3x3 and a stride of 1, a normalization operation, a ReLU activation operation, and a random deactivation operation. The number of 3D convolutional layers can be determined according to different projects; in this embodiment, there are 3 layers.

[0111] The residual module is used to add the encoder's input and output to alleviate the problem of vanishing gradients in the network.

[0112] The decoder structure is used to extract high-level features and recover feature scale. The number of decoders can vary depending on the project; in this embodiment, there are four decoders, but other embodiments may use different numbers. For the specific structure of the decoders, please refer to [link to relevant documentation]. Figure 13 The decoder structure includes a 3D deconvolution layer, a 3D convolution layer, and a residual module. The 3D deconvolution layer is used for feature scale recovery; in this embodiment, a 3x3x3 kernel with a stride of 2 is used for scale recovery. The 3D convolution layer is used to extract high-level features. It consists of a 3x3x3 kernel with a stride of 1, a normalization operation, a ReLU activation operation, and a random deactivation operation. The number of 3D convolution layers can vary depending on the project; in this embodiment, there are 3. The residual module adds the decoder's input and output to alleviate the vanishing gradient problem.

[0113] The artifact recognition module can calculate the model processing results in the following way: perform the calculation of multi-artifact recognition network training, and send the calculated feature maps to the category center point prediction (HeatMap) branch, the center point offset prediction branch (Offset) and the target object length, height and width prediction branches (Height, Width and depth) respectively. By combining the prediction results of the three branches, the position of the target in the original image is finally obtained.

[0114] S1006: The model loss is calculated based on the labels and the model processing results.

[0115] S1008: When the model loss does not meet the requirements, the network parameters of the initial model are optimized, and the sample medical image data are processed by the optimized model to obtain the model processing result. The step of calculating the model loss based on the label and the model processing result is continued until the model loss meets the requirements, and the target model is obtained. The target model is the artifact recognition model or artifact correction model in any of the above embodiments.

[0116] The definition of loss can be found below, and will not be specifically limited here.

[0117] Specifically, in this embodiment, the model is trained by adjusting the loss function until the model loss meets the requirements, thus obtaining the target model.

[0118] In the above embodiments, an artifact recognition model or an artifact correction model is obtained through model training to facilitate subsequent artifact correction.

[0119] In one embodiment, the target model is an artifact recognition model, and the labels are pre-labeled artifact regions. The initial model processes the sample medical image data to obtain model processing results, including: extracting sample features from the sample medical image data using the feature extraction module of the initial model; inputting the sample features into the center point prediction module, center point offset prediction module, and target object length, width, and height prediction module of the initial model for processing to obtain corresponding artifact region information; calculating the model loss based on the labels and model processing results, including: obtaining the first center point position and first length, width, and height information based on the pre-labeled artifact regions; calculating a first loss value by comparing the artifact region information obtained from the center point prediction module with the first center point position of the pre-labeled artifact regions; calculating a second loss value by comparing the artifact region information obtained from the center point offset prediction module with the first center point position of the pre-labeled artifact regions; calculating a third loss value by comparing the artifact region information obtained from the target object length, width, and height prediction module with the first length, width, and height information of the pre-labeled artifact regions; and calculating the model loss based on the first loss value, second loss value, and third loss value.

[0120] In one embodiment, calculating a first loss value by comparing the artifact region information obtained by the center point prediction module with the first center point position of the pre-labeled artifact region includes: determining sample positive voxels and sample negative voxels based on the pre-labeled artifact region; processing the artifact region information through the center point prediction module to obtain predicted positive voxels, predicted negative voxels, and a second center point position; calculating a first loss for the sample positive voxels and predicted positive voxels; calculating a second loss for the sample negative voxels and predicted negative voxels; calculating a third loss for the first center point position and the second center point position; and calculating the first loss value corresponding to the center point prediction module based on the first loss, second loss, and third loss; and shifting the artifact region obtained by the center point offset prediction module... The second loss value is calculated by using the position of the first center point of the artifact region pre-annotated by the domain information. This includes: determining the predicted offset value through the center point offset prediction module, determining the true offset value based on the pre-annotated artifact region, and calculating the second loss value corresponding to the center point offset prediction module based on the predicted offset value, the true offset value, and the number of center points. The third loss value is calculated by using the artifact region information obtained by the target object's length, width, and height prediction module and the first length, width, and height information of the pre-annotated artifact region. This includes: determining the second length, width, and height information through the target object's length, width, and height prediction module, and calculating the third loss value corresponding to the target object's length, width, and height prediction module based on the first length, width, and height information, the second length, width, and height information, and the number of center points.

[0121] Specifically, see Figure 14 As shown, Figure 14 This is a flowchart illustrating the training process of an artifact recognition model in one embodiment. In this embodiment, the first step is dataset collection, specifically including acquiring a medical image dataset and completing data annotation. For example, using software such as 3dslicer, the 3D bounding boxes containing artifacts are annotated, along with the center coordinates, length, width, and height of the bounding boxes. The dataset is then divided into training, validation, and test sets, and data augmentation is performed on the training set samples. Next, preprocessing is performed, such as adjusting window width and window level. Third, the artifact recognition model network parameters are calculated forward. Specifically, this involves calculating the training parameters for the multi-artifact recognition network, and the calculated feature maps are fed to the category center point prediction (HeatMap) branch, the center point offset prediction branch (Offset), and the target object's length, width, and depth prediction branches (Height, Width, and Depth). By combining the prediction results from these three branches, the final position of the target in the original image is obtained. Fourth, calculate the combined loss, for example, the loss function used is the combined loss function of the three task branches. Fifth, when the network test results of the validation set and the test set meet the expected conditions, the network training can be terminated, the final artifact recognition is saved and the final artifact recognition model is output.

[0122] Specifically, in combination Figure 15 As shown, Figure 15 This is a structural entity diagram of an artifact recognition model in one embodiment. In order to accurately identify the 3D bounding box contained in the artifact region, this embodiment constructs a 3D fully convolutional backbone network, i.e., a feature extraction module to extract features. By constructing a center point prediction (HeatMap) branch, a center point offset prediction branch (Offset) and a target object length, width and depth prediction branch (Height, Width and Depth), the center point and length, width and depth of the 3D box in the artifact region are obtained. By combining the prediction results of the three branches, the position of the target in the original image is finally obtained.

[0123] For ease of understanding, the overall loss function of the artifact recognition model in this embodiment includes three parts: center point loss (HeatMap) k Center point offset loss (Offset) offset And the target object's length, width, and depth loss (Height, Width, and Depth) size .

[0124] The loss function for the artifact recognition model is:

[0125] l det =l k +w s1 ×l size +w s2 ×l offset

[0126] Where w si The weights for each element are dynamically adjusted during the training process.

[0127] Specifically, in combination Figure 14 As shown, the center point loss l k The loss function is:

[0128] l k =w l1 ×l pos +w l2 ×l neg +w l3 ×l dis

[0129] Among them, w li The weights for each element are dynamically adjusted during the training process. First, for the real heatmap generated from keypoints, voxels with a threshold greater than 0 are extracted as positive voxels, and voxels with a threshold less than or equal to 0 are extracted as negative voxels. Then, using the black and white voxel range of the real heatmap, predicted positive and negative voxels are extracted from the predicted heatmap. The loss between real positive voxels and predicted positive voxels is calculated as l. posThe loss between the true negative voxels and the predicted negative voxels is calculated as l. neg , l dis This is the average Euclidean distance between the predicted anatomical landmark locations and the actual anatomical landmark locations.

[0130] The center point offset function of the artifact recognition model is l offset for:

[0131]

[0132] in, Here, p represents the predicted offset value output by the network, where p represents the center position of the current bounding box, and R represents the downsampling factor. This represents the offset value from the true value, and N represents the number of center points.

[0133] The length, width, and height loss function of the artifact recognition model size for:

[0134]

[0135] in, S represents the predicted length, width, and height values ​​output by the network. P The actual length, width, and height values ​​are given, and N represents the number of center points.

[0136] The above embodiments provide a deep learning-based metal artifact recognition model. The model's input is the original CT image sequence, and its output is the three-dimensional target region where artifacts exist. A risk level score for the artifacts is determined using established scoring principles. The scoring principles are based on the attribute information of the target artifact region within the identified three-dimensional target region.

[0137] In one embodiment, the target model is an artifact correction model, and the label is the matched artifact-free region; the model loss is calculated based on the label and the model processing result, including: determining the corresponding pixel pairs of the model processing result and the matched artifact-free region; and calculating the model loss based on the pixel value of the pixel pair.

[0138] Among them, combined Figure 16 As shown, Figure 16This is a flowchart illustrating the training process of an artifact correction model in one embodiment. In this embodiment, the first step is dataset collection, specifically acquiring the original medical image dataset and corresponding artifact CT images. These artifact CT images can be obtained by simulating artifact generation. The dataset is then divided into training, validation, and test sets, and data augmentation is performed on the training set samples. Next, preprocessing is performed, such as adjusting window width and window level. Third, the artifact correction model network parameters are calculated forward, and the correction result is output. Fourth, the loss is calculated, for example using the mean squared error loss function. Specifically, this embodiment extracts features by constructing a three-dimensional fully convolutional backbone network and outputs the correction result through the artifact-corrected image output module.

[0139] The artifact correction model defines the mean squared error loss function as follows:

[0140]

[0141] Where w×h×d represents the total number of pixels, I (i，j，k) Let K be the pixel value of the labeled image at pixel (i,j,k). (i，j，k) This is the pixel value of the image output by the network at pixel (i,j,k).

[0142] The above embodiments provide a deep learning-based metal artifact correction model. The input of the model is the CT image sequence to be corrected, and the output of the model is the CT image sequence after artifact correction. Artifact correction is performed on low-risk artifacts, improving the quality of the original image and facilitating efficient preoperative planning. It can be applied to shutdown replacement surgery robots to improve automation and intelligence.

[0143] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0144] Based on the same inventive concept, this application also provides an artifact correction apparatus for implementing the artifact correction method described above. The solution provided by this apparatus is similar to the implementation described in the above method; therefore, the specific limitations in one or more artifact correction apparatus embodiments provided below can be found in the limitations of the artifact correction method described above, and will not be repeated here.

[0145] In one embodiment, an artifact correction apparatus is provided, comprising:

[0146] The preoperative imaging data acquisition module is used to acquire preoperative imaging data.

[0147] The artifact recognition module is used to perform artifact recognition on the preoperative image data using a pre-trained artifact recognition model to obtain the target artifact region.

[0148] The attribute information acquisition module is used to acquire the attribute information of the target artifact region.

[0149] The artifact correction module is used to perform artifact correction on the target artifact region that needs artifact correction by using a pre-trained artifact correction model when it is determined from the attribute information that the target artifact region needs artifact correction.

[0150] In one embodiment, the artifact correction module is further configured to obtain attribute information of the corrected artifact region, take the corrected artifact region as the target artifact region, and continue to execute the step of determining whether the target artifact region needs to be corrected based on the attribute information, until it is determined based on the attribute information that the target artifact region does not need to be corrected.

[0151] In one embodiment, the attribute information acquisition module is further configured to acquire at least one of the following as attribute information: the number of target artifact regions, the region location, the region volume ratio, and the region grayscale intensity distribution.

[0152] The device further includes:

[0153] The judgment module is used to determine the weights corresponding to each attribute information and the reference index values ​​corresponding to each attribute information; calculate the correction index value of the target artifact region based on the weights and the reference index values; and determine whether the target artifact region needs artifact correction based on the correction index values.

[0154] In one embodiment, the artifact recognition module is further configured to extract features from the preoperative image data using a feature extraction module of a pre-trained artifact recognition model to obtain features to be processed; input the features to be processed into the center point prediction module, center point offset prediction module, and length, width, and height prediction module of the artifact recognition model for processing to obtain corresponding artifact region information; and calculate the target artifact region based on the artifact region information.

[0155] In one embodiment, the preoperative imaging data is a CT image sequence; the artifact recognition model is a three-dimensional convolutional neural network model; and / or the artifact correction model is a three-dimensional convolutional neural network model.

[0156] In one embodiment, a target model training apparatus is provided, comprising:

[0157] The sample acquisition module is used to acquire sample medical image data, which carries tags.

[0158] The model processing module is used to process the sample medical image data using an initial model to obtain the model processing result.

[0159] The loss calculation module is used to calculate the model loss based on the label and the model processing result.

[0160] The training module is used to optimize the network parameters of the initial model when the model loss does not meet the requirements, and to process the sample medical image data through the optimized model to obtain the model processing result. The module continues to execute the step of calculating the model loss based on the label and the model processing result until the model loss meets the requirements, and then obtains the target model. The target model is the artifact recognition model or artifact correction model described in any of the above embodiments.

[0161] In one embodiment, the above-mentioned model processing module is further configured to extract features from the sample medical image data through the feature extraction module of the initial model to obtain sample features; and input the sample features into the center point prediction module, center point offset prediction module and target object length, width and height prediction module of the initial model for processing to obtain the corresponding artifact region information of the sample.

[0162] The aforementioned loss calculation module is further configured to obtain the first center point position and first length, width, and height information based on the pre-labeled artifact region; calculate a first loss value by combining the artifact region information obtained by the center point prediction module with the first center point position of the pre-labeled artifact region; calculate a second loss value by combining the artifact region information obtained by the center point offset prediction module with the first center point position of the pre-labeled artifact region; calculate a third loss value by combining the artifact region information obtained by the target object length, width, and height prediction module with the first length, width, and height information of the pre-labeled artifact region; and calculate the model loss based on the first loss value, the second loss value, and the third loss value.

[0163] In one embodiment, the loss calculation module is further configured to determine positive and negative voxels of the sample based on pre-labeled artifact regions, process the artifact region information through the center point prediction module to obtain predicted positive voxels, predicted negative voxels, and a second center point position, calculate a first loss for the positive voxels and the predicted positive voxels, calculate a second loss for the negative voxels and the predicted negative voxels, calculate a third loss for the first center point position and the second center point position, and calculate a first loss value corresponding to the center point prediction module based on the first loss, the second loss, and the third loss; determine a predicted offset value through the center point offset prediction module, determine a true offset value based on the pre-labeled artifact regions, and calculate a second loss value corresponding to the center point offset prediction module based on the predicted offset value, the true offset value, and the number of center points; determine second length, width, and height information through the target object length, width, and height prediction module, and calculate a third loss value corresponding to the target object length, width, and height prediction module based on the first length, width, and height information, the second length, width, and height information, and the number of center points.

[0164] In one embodiment, the loss calculation module is further configured to determine the corresponding pixel pairs of the model processing result and the matched artifact-free regions; and calculate the model loss based on the pixel values ​​of the pixel pairs.

[0165] The various modules in the aforementioned artifact correction device and target model training device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0166] In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 17As shown, the computer device includes a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When executed by the processor, the computer program implements an artifact correction method and a target model training method. The display screen can be an LCD screen or an e-ink display screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0167] Those skilled in the art will understand that Figure 17 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0168] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0169] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above method embodiments.

[0170] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0171] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0172] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0173] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. An artifact correction method, characterized in that, The method includes: Obtain preoperative imaging data; The pre-trained artifact recognition model is used to identify artifacts in the preoperative image data to obtain the target artifact region, which refers to the region in the preoperative image data where there are metallic artifacts. Obtain the attribute information of the target artifact region; When it is determined that the target artifact region needs to be corrected based on the attribute information, the target artifact region that needs to be corrected is corrected by a pre-trained artifact correction model, including: inputting the target artifact region to be corrected into the pre-trained artifact correction model to obtain the corrected image sequence. The step of obtaining the attribute information of the target artifact region includes: The number of target artifact regions, region location, region volume ratio, and region gray value intensity distribution are obtained as attribute information, wherein the region volume ratio is the ratio of the intersection volume of the artifact region and the surgical selection region to the volume of the surgical selection region. When it is determined from the attribute information that the target artifact region needs artifact correction, before performing artifact correction on the target artifact region that needs artifact correction using a pre-trained artifact correction model, the method further includes: Determine the weight corresponding to each attribute information and the reference index value corresponding to each attribute information, wherein the reference index value corresponding to the region location is determined based on whether it conflicts with the surgical region; Calculate the correction index value of the target artifact region based on the weight and the reference index value; Determining whether the target artifact region needs artifact correction based on the correction index value includes: determining whether the target artifact region has an impact on subsequent surgical planning based on the correction index value; if the target artifact region has little impact on the surgical planning described later, then artifact correction is not required.

2. The method according to claim 1, characterized in that, After the artifact correction model obtained through pre-training performs artifact correction on the target artifact region that needs artifact correction, the process includes: Obtain the attribute information of the corrected artifact region, take the corrected artifact region as the target artifact region, and continue to execute the step of determining whether the target artifact region needs to be corrected based on the attribute information, until it is determined based on the attribute information that the target artifact region does not need to be corrected.

3. The method according to claim 1, characterized in that, The preoperative image data is used to identify artifacts using a pre-trained artifact recognition model to obtain the target artifact region, including: The feature extraction module of the pre-trained artifact recognition model extracts features from the preoperative image data to obtain the features to be processed. The features to be processed are respectively input into the center point prediction module, center point offset prediction module and target object length, width and height prediction module of the artifact recognition model for processing to obtain the corresponding artifact region information. The target artifact region is calculated based on the information of each artifact region.

4. The method according to claim 1, characterized in that, The preoperative imaging data is a CT image sequence; the artifact recognition model is a three-dimensional convolutional neural network model, and / or the artifact correction model is a three-dimensional convolutional neural network model.

5. The method according to claim 1, characterized in that, The training methods for the artifact recognition model or artifact correction model include: Acquire sample medical image data, wherein the sample medical image data carries tags; The initial model is used to process the sample medical image data to obtain the model processing result; The model loss is calculated based on the labels and the model processing results. When the model loss does not meet the requirements, the network parameters of the initial model are optimized, and the sample medical image data are processed by the optimized model to obtain the model processing result. The step of calculating the model loss based on the label and the model processing result continues until the model loss meets the requirements, thus obtaining the artifact recognition model or artifact correction model.

6. The method according to claim 5, characterized in that, The labels are pre-labeled artifact regions; the process of processing the sample medical image data using an initial model to obtain the model processing result includes: The initial model's feature extraction module extracts features from the sample medical image data to obtain sample features; the sample features are then input into the initial model's center point prediction module, center point offset prediction module, and target object length, width, and height prediction module for processing to obtain the corresponding sample artifact region information; The calculation of the model loss based on the label and the model processing result includes: The position of the first center point and the first length, width and height information are obtained based on the pre-marked artifact region; The first loss value is calculated by comparing the artifact region information obtained by the center point prediction module with the first center point position of the pre-labeled artifact region. The second loss value is obtained by calculating the position of the first center point of the artifact region pre-labeled by the artifact region information obtained from the center point offset prediction module. The third loss value is calculated by combining the artifact region information obtained by the length, width and height prediction module of the target object with the first length, width and height information of the pre-labeled artifact region. The model loss is calculated based on the first loss value, the second loss value, and the third loss value.

7. The method according to claim 6, characterized in that, The step of calculating the first loss value by comparing the artifact region information obtained by the center point prediction module with the first center point position of the pre-labeled artifact region includes: Based on the pre-labeled artifact regions, positive and negative voxels of the samples are determined. The artifact region information is processed by the center point prediction module to obtain the predicted positive voxels, predicted negative voxels, and the second center point position. The first loss of the positive voxels and the predicted positive voxels is calculated, the second loss of the negative voxels and the predicted negative voxels is calculated, and the third loss of the first center point position and the second center point position is calculated. Based on the first loss, the second loss, and the third loss, the first loss value corresponding to the center point prediction module is calculated. The step of calculating the second loss value by taking the first center point position of the artifact region pre-annotated by the artifact region information obtained from the center point offset prediction module includes: The predicted offset value is determined by the center point offset prediction module, the true offset value is determined by the pre-labeled artifact region, and the second loss value corresponding to the center point offset prediction module is calculated based on the predicted offset value, the true offset value and the number of center points. The step of calculating the third loss value by combining the artifact region information obtained from the length, width, and height prediction module of the target object with the first length, width, and height information of the pre-labeled artifact region includes: The second length, width, and height information is determined by the length, width, and height prediction module of the target object. Based on the first length, width, and height information, the second length, width, and height information, and the number of center points, the third loss value corresponding to the length, width, and height prediction module of the target object is calculated.

8. The method according to claim 5, characterized in that, The label is the matched artifact-free region; the calculation of the model loss based on the label and the model processing result includes: Determine the corresponding pixel pairs of the model processing results and the matched artifact-free regions; The model loss is calculated based on the pixel values ​​of the pixel pairs.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 8.

Images