Cholangiocarcinoma lymph node metastasis image detection method and system based on frequency perception and semantic constraint fusion

Abstract

The application discloses a kind of based on frequency perception and semantic constraint fusion's bile duct cancer lymph node metastasis image detection method and system.The present application is directed to the size of suspicious lymph node in the CT / MRI image of bile duct cancer lymph node metastasis is small, weak boundary, low contrast and easily interfered by complex anatomical structure, YOLO-World network is improved, frequency perception gradual refinement module is added in image branch, weak boundary and subtle texture features are enhanced by space convolution branch and frequency domain residual branch joint;Semantics constraint context attention module is added in image text fusion stage, multi-scale visual features and text features are aggregated by direction constraint and semantic correlation constraint, reduce non-target anatomical structure false response.The present application can output suspicious lymph node candidate frame, confidence and text alignment score, improve the detection ability and positioning accuracy of suspicious lymph node candidate region in the image of bile duct cancer lymph node metastasis.

Description

Technical Field

[0001] This invention belongs to the fields of medical image processing, target detection, and visual-language fusion in computer vision, specifically relating to an image detection method and system for cholangiocarcinoma lymph node metastasis based on frequency perception and semantic constraint fusion. Background Technology

[0002] Cholangiocarcinoma is characterized by its high malignancy and tendency to metastasize to lymph nodes. The status of lymph node metastasis is of great significance for preoperative imaging assessment, surgical planning, and subsequent treatment planning. CT and MRI are commonly used imaging methods in the preoperative imaging assessment of cholangiocarcinoma, providing anatomical information on the tumor area, blood vessels, bile ducts, soft tissues, and surrounding lymph nodes.

[0003] However, suspicious lymph nodes in images of cholangiocarcinoma lymph node metastasis are usually small, have indistinct borders, and low contrast. They are often located within complex anatomical backgrounds such as blood vessels, bile ducts, intestines, and soft tissue boundaries, making them easily confused with surrounding structures and leading to missed and false detections during the detection process. Therefore, accurately locating candidate regions of suspicious lymph nodes related to cholangiocarcinoma lymph node metastasis in CT / MRI images is an important issue in intelligent medical image analysis.

[0004] In recent years, with the widespread application of deep learning in computer vision and the rapid development of GPU computing power, object detection algorithms have been widely used in various image analysis tasks. The YOLO series of detection models have advantages such as high speed, compact structure, and convenient deployment. The YOLO-World model further incorporates image features and textual prompts, achieving open-vocabulary object detection and providing a new approach to object localization in medical imaging.

[0005] However, when YOLO-World is directly used for CT / MRI image detection of lymph node metastases in cholangiocarcinoma, the following problems still exist: On the one hand, suspicious metastatic lymph nodes are often small in size and have weak boundaries, and their texture and contour information are easily weakened during multiple convolutions and downsampling processes; on the other hand, the structures such as blood vessels, bile ducts, intestines and soft tissue boundaries in CT / MRI images are similar to lymph nodes in terms of local visual features, and ordinary image-text fusion methods are prone to responding to non-target anatomical structures, thus causing false detections. Summary of the Invention

[0006] In CT / MRI images of lymph node metastases in cholangiocarcinoma, suspicious lymph nodes often present challenges due to their small size, weak boundary features, low contrast, and susceptibility to confusion with non-target anatomical structures such as blood vessels, bile ducts, intestines, and soft tissue boundaries. Deep learning methods face significantly increased difficulty in identifying and detecting these lymph nodes in this scenario. Furthermore, the morphology, size, and grayscale representation of lymph nodes vary among different patients, modalities, and scanning sequences, further complicating the detection task.

[0007] Therefore, to address the above problems, this invention provides a method and system for detecting lymph node metastasis in cholangiocarcinoma based on frequency awareness and semantic constraint fusion. This method optimizes the YOLO-World network structure, introduces a frequency awareness progressive refinement module in the image branch, and introduces a semantic constraint contextual attention module in the image-text fusion stage. This enhances the model's sensitivity to small-sized, low-contrast suspicious lymph node regions in CT / MRI images, effectively improving the accuracy and recall of candidate regions for detecting lymph node metastasis in cholangiocarcinoma images.

[0008] In a first aspect, the present invention provides an imaging method for detecting lymph node metastasis in cholangiocarcinoma based on frequency-aware and semantic constraint fusion, comprising the following steps:

[0009] CT / MRI images of patients with cholangiocarcinoma were acquired and annotated to obtain a dataset of CT / MRI images of lymph node metastases in cholangiocarcinoma.

[0010] The YOLO-World network is trained using a public dataset to obtain a pre-trained YOLO-World network model.

[0011] The YOLO-World pre-trained network model structure was modified as follows: a frequency-aware progressive refinement module was added to the image branch, and a semantic constraint context attention module was added to the image-text fusion stage;

[0012] The CT / MRI image dataset of cholangiocarcinoma lymph node metastasis was input into the modified YOLO-World pre-trained network model and retrained to obtain the target network model.

[0013] The target network model is loaded into the cholangiocarcinoma lymph node metastasis imaging detection system, and the collected CT / MRI image data to be detected is imported into the cholangiocarcinoma lymph node metastasis imaging detection system to complete the detection of suspicious lymph node candidate regions.

[0014] A second aspect of the present invention provides an imaging detection system for lymph node metastasis in cholangiocarcinoma based on frequency-aware and semantic constraint fusion, comprising:

[0015] The image acquisition module is used to acquire and annotate CT / MRI images of patients with cholangiocarcinoma, and obtain a dataset of CT / MRI images of cholangiocarcinoma lymph node metastases.

[0016] The model pre-training module is used to train the YOLO-World network using a public dataset to obtain the YOLO-World pre-trained network model.

[0017] The model structure modification module is used to modify the structure of the YOLO-World pre-trained network model: adding a frequency-aware progressive refinement module to the image branch and a semantic constraint context attention module to the image-text fusion stage;

[0018] The model retraining module is used to input the CT / MRI image dataset of cholangiocarcinoma lymph node metastasis into the modified YOLO-World pre-trained network model and retrain it to obtain the target network model.

[0019] The image detection module is used to load the target network model into the cholangiocarcinoma lymph node metastasis image detection system, import the acquired CT / MRI image data to be detected into the cholangiocarcinoma lymph node metastasis image detection system, and complete the detection of suspicious lymph node candidate regions.

[0020] In a third aspect, the present invention 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 in the above-described method for detecting lymph node metastasis in cholangiocarcinoma based on frequency perception and semantic constraint fusion.

[0021] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps in the above-described method for detecting lymph node metastasis in cholangiocarcinoma based on frequency-aware and semantic constraint fusion.

[0022] The technical solution provided by this invention has the following beneficial effects:

[0023] This invention modifies the image branch and image-text fusion module of the YOLO-World network. A frequency-aware progressive refinement module is added to the image branch, enhancing the weak boundary and texture response of small-sized suspicious lymph nodes in CT / MRI images through joint modeling of spatial convolution and frequency domain residual branches. A stage-guided feature fusion unit progressively fuses deep semantic features and shallow detail features, enabling the model to obtain stable contextual information while preserving local details. In the image-text fusion stage, a semantically constrained contextual attention module is added, aggregating multi-scale visual features in a coarse-to-fine order and using text features to perform semantic relevance screening of visual features. This suppresses interference from non-target anatomical structures such as blood vessels, bile ducts, and soft tissue boundaries, reducing false detections and improving the detection capability and localization accuracy of suspicious lymph node candidate regions in CT / MRI images of cholangiocarcinoma lymph node metastasis. Attached Figure Description

[0024] Figure 1 A flowchart illustrating the imaging detection system for cholangiocarcinoma lymph node metastasis provided in this application embodiment;

[0025] Figure 2 This is a modified YOLO-World network structure diagram provided in the embodiments of this application;

[0026] Figure 3 A structural diagram of the semantic constraint context attention module provided in the embodiments of this application;

[0027] Figure 4 This is a structural diagram of the frequency-aware progressive refinement module provided in an embodiment of this application;

[0028] Figure 5 This paper compares the detection performance of the embodiments of this application with that of the benchmark network on a CT / MRI image dataset of cholangiocarcinoma lymph node metastasis. Detailed Implementation

[0029] To describe the present invention in more detail, the technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0030] See Figure 2 This application provides an image detection method for cholangiocarcinoma lymph node metastasis based on frequency awareness and semantic constraint fusion. The method includes the following steps:

[0031] Step S1: Obtain CT and / or MRI images of patients with cholangiocarcinoma. Radiology professionals will mark the bounding boxes of suspicious lymph node areas in the images to obtain a CT / MRI image dataset of cholangiocarcinoma lymph node metastasis.

[0032] Step S2: Train the YOLO-World network on the Objects365v1 large object detection dataset to obtain the pre-trained YOLO-World network model.

[0033] Step S3, as follows Figure 2 , Figure 3 and Figure 4 As shown, the structure of the pre-trained network model of YOLO-World is modified.

[0034] In this application embodiment, a frequency-aware progressive refinement module is added to the image branch of YOLO-World to enhance weak boundary and texture features; a semantically constrained contextual attention module is added to the image-text fusion stage of YOLO-World to enhance the effective matching between text prompts and visual features of suspicious lymph nodes.

[0035] Furthermore, the frequency-aware progressive refinement module is used to enhance weak boundary and subtle texture information before the detection backbone network extracts deep features, avoiding the weakening of small target evidence of suspicious lymph nodes during multiple convolutions and downsampling processes; the semantic constraint context attention module is used to constrain the aggregation process between visual features and text features in the image-text fusion stage, making the model pay more attention to visual regions related to lymph node transfer detection targets and suppressing false responses caused by non-target anatomical structures.

[0036] Specifically, the image features in the improved YOLO-World pre-trained network model structure are processed according to the following steps:

[0037] (1) When the input CT image and / or MRI image is fed into the improved YOLO-World pre-trained network model, the input image feature X is first weakly enhanced in terms of boundary and texture through the frequency-aware progressive refinement module FAPR.

[0038] like Figure 4 As shown, the FAPR module includes a frequency-driven convolutional unit FDConv and a stage-guided feature fusion unit SGFF. The frequency-driven convolutional unit includes a spatial convolutional branch and a frequency domain residual branch, which extracts features from both the spatial and frequency domains. The stage-guided feature fusion unit is used to fuse deep semantic features and shallow detail features step by step.

[0039] Furthermore, the frequency-driven convolutional unit performs feature processing according to the following steps:

[0040] The input features are fed into the spatial convolution branch to obtain local features in the spatial domain; the input features are then fed into the frequency domain residual branch to perform a two-dimensional frequency domain transformation, resulting in frequency domain features; the real and imaginary parts of the frequency domain features are convolved separately, and the processed frequency domain features are then inversely transformed to obtain frequency domain residual features; the local features in the spatial domain and the frequency domain residual features are fused to obtain the enhanced frequency-driven convolution output features.

[0041]

[0042]

[0043] in, Represents two-dimensional frequency domain transformation. F represents the inverse frequency domain transform. real (X) and F imag (X) represent the real and imaginary parts of the frequency domain feature, respectively, F out This step enhances the output of the frequency-driven convolutional unit. Through this process, the model is able to enhance the weak boundaries, fine textures, and low-contrast responses of suspicious lymph node regions while preserving local anatomical information from CT / MRI.

[0044] (2) The frequency-driven convolutional unit FDConv is combined with downsampling to construct multi-level features from shallow to deep, and the feature processing is completed by the stage-guided feature fusion unit SGFF. The stage-guided feature fusion unit SGFF performs feature processing according to the following steps:

[0045] First, SGFF upsamples deep semantic features level by level, fusing deep semantic features and shallow detail features step by step along a top-down path. Let the input feature be X, the multi-level feature calculation is as follows:

[0046]

[0047] in, The downsampling operation results in {X1, X2, X3, X4, X5} representing multi-level features from shallow to deep.

[0048] Subsequently, for the current stage feature X k Stage modulation weights are generated based on visual features and then combined with stage guidance information. i The fusion yields stage fusion characteristics:

[0049]

[0050] in, represents the Sigmoid function, and ⊙ represents element-wise multiplication. Through this step, deep contextual information can be progressively passed to shallow localization features, and shallow boundary and texture information can also supplement deep semantic expression, thereby enhancing the target evidence of small-sized suspicious lymph nodes.

[0051] Finally, the stage fusion features are divided into multiple channel groups, and each group is further refined by lightweighting the self-attention unit. The refined features of each channel group are then spliced ​​together to obtain the frequency-aware progressively refined output features.

[0052] (3) Input the image features enhanced by FAPR into the YOLO-World detection backbone network to obtain three multi-scale visual features X1, X2, and X3. Among them, X1 represents coarse-scale semantic features, X2 represents meso-scale fusion features, and X3 represents fine-scale localization features.

[0053] like Figure 3 As shown, in the image-text fusion stage, the Semantic Constraint Context Attention (SCCA) module performs feature processing according to the following steps:

[0054] First, text conditional features X2 are obtained using the text conditional feature interaction layer. ′ and X3 ′ The three scale features are then spliced ​​and expanded in order from coarse to fine:

[0055]

[0056]

[0057] Where Z represents the unfolded visual token sequence, and the visual tokens are arranged in order from coarse scale to fine scale.

[0058] Then, the text prompt information related to the target of cholangiocarcinoma lymph node metastasis detection is input into the text encoder to obtain text features t. A text query vector q is generated based on the text features, and a visual key vector k is generated based on the visual token sequence. j and visual value vector v j :

[0059]

[0060]

[0061]

[0062] In order to enable fine-scale localization features to aggregate coarse-scale semantic context information, SCCA constructs a directional constraint mask M based on a coarse-to-fine scale order. dir This enables fine-scale localization features to aggregate coarse-scale semantic context information:

[0063]

[0064] Meanwhile, in order to suppress non-target anatomical structures that are weakly related to text cues, SCCA constructs a semantic relevance mask M based on the distance between the text query vector and the visual key vector. sem :

[0065]

[0066] Among them, M dir Used to constrain visual tokens to aggregate in a coarse-to-fine order, ε is a minimal constant to prevent the denominator from being zero, and m sem (j) represents the semantic relevance between the j-th visual token and the text prompt. The semantic relevance score is broadcast along the query dimension to obtain M. sem (i, j).

[0067] Next, the directional constraint mask and the semantic relevance mask are combined to form the guiding mask:

[0068]

[0069] Subsequently, image and text feature aggregation is performed on the visual token based on the guiding mask to obtain the updated visual token representation:

[0070]

[0071] Among them, A ij This represents the attention weights modulated and normalized by the guiding mask.

[0072] Finally, the updated visual token is remapped into a multi-scale feature map and fed into the YOLO-World detection head, outputting suspicious lymph node candidate boxes, confidence scores, and text alignment scores.

[0073] Step S4: Input the CT / MRI images of cholangiocarcinoma lymph node metastasis into the modified YOLO-World pre-trained network model on the cholangiocarcinoma lymph node metastasis CT / MRI image dataset, and retrain the modified model to obtain the target network model.

[0074] Step S5: Load the target network model into the bile duct cancer lymph node metastasis imaging detection system, and load the image data to be detected acquired by CT equipment and / or MRI equipment into the system to realize the detection of suspicious lymph node candidate regions.

[0075] The workflow of the bile duct cancer lymph node metastasis imaging detection system is as follows: Figure 1 As shown, the specific steps include the following:

[0076] (1) The cholangiocarcinoma lymph node metastasis imaging detection system reads the CT and / or MRI images to be detected.

[0077] (2) The CT and / or MRI images to be detected, along with text prompts, are input into the target network model for forward inference.

[0078] (3) Determine whether there are suspicious lymph node candidate regions in the CT image and / or MRI image to be detected that meet the confidence threshold. If there are suspicious lymph node candidate regions, proceed to step (4); otherwise, proceed to step (5).

[0079] (4) The bile duct cancer lymph node metastasis imaging detection system marks the candidate regions of suspicious lymph nodes and outputs the candidate box, confidence score and text alignment score, indicating that there are candidate regions of suspicious lymph nodes in the image.

[0080] (5) Determine if there are any unread images. If there are, return to step (1); otherwise, end the detection.

[0081] Experimental results show that, compared with the benchmark network YOLO-World, the present invention significantly improves mAP50, mAP50-95, and recall, as detailed in Table 1.

[0082] Table 1. Comparison of detection performance between the present invention and the benchmark network.

[0083]

[0084] Meanwhile, this invention significantly outperforms the benchmark network YOLO-World on the validation set, with a significantly improved detection rate, as detailed in [link to details]. Figure 5 .

[0085] This application also discloses an imaging detection system for cholangiocarcinoma lymph node metastasis based on frequency-aware and semantic constraint fusion, including:

[0086] The image acquisition module is used to acquire and annotate CT / MRI images of patients with cholangiocarcinoma, and obtain a dataset of CT / MRI images of cholangiocarcinoma lymph node metastases.

[0087] The model pre-training module is used to train the YOLO-World network using a public dataset to obtain the YOLO-World pre-trained network model.

[0088] The model structure modification module is used to modify the structure of the YOLO-World pre-trained network model: adding a frequency-aware progressive refinement module to the image branch and a semantic constraint context attention module to the image-text fusion stage;

[0089] The model retraining module is used to input the CT / MRI image dataset of cholangiocarcinoma lymph node metastasis into the modified YOLO-World pre-trained network model and retrain it to obtain the target network model.

[0090] The image detection module is used to load the target network model into the cholangiocarcinoma lymph node metastasis image detection system, import the acquired CT / MRI image data to be detected into the cholangiocarcinoma lymph node metastasis image detection system, and complete the detection of suspicious lymph node candidate regions.

[0091] This application also discloses a computer device, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method for detecting lymph node metastases in cholangiocarcinoma based on frequency-aware and semantic constraint fusion. The memory may include main memory, such as high-speed random access memory, or it may also include non-volatile memory, such as at least one disk storage device. The processor, network interface, and memory are interconnected via an internal bus, which may be an industry-standard architecture bus, a peripheral component interconnection standard bus, an extended industry-standard architecture bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. The memory is used to store the program; specifically, the program may include program code, which includes computer operation instructions. The memory may include main memory and non-volatile memory, and provides instructions and data to the processor.

[0092] This application also discloses a computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements the steps in the above-described method for detecting lymph node metastases in cholangiocarcinoma based on frequency-aware and semantic constraint fusion. Specifically, the computer-readable storage medium includes, but is not limited to, volatile memory and / or non-volatile memory. The volatile memory may include random access memory (RAM) and / or cache memory, etc. The non-volatile memory may include read-only memory (ROM), hard disk, flash memory, optical disk, magnetic disk, etc.

[0093] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. An imaging method for detecting lymph node metastasis in cholangiocarcinoma based on frequency-aware and semantic constraint fusion, characterized in that, Includes the following steps: CT / MRI images of patients with cholangiocarcinoma were acquired and annotated to obtain a dataset of CT / MRI images of lymph node metastases in cholangiocarcinoma. The YOLO-World network is trained using a public dataset to obtain a pre-trained YOLO-World network model. The YOLO-World pre-trained network model structure was modified as follows: a frequency-aware progressive refinement module was added to the image branch, and a semantic constraint context attention module was added to the image-text fusion stage; The CT / MRI image dataset of cholangiocarcinoma lymph node metastasis was input into the modified YOLO-World pre-trained network model and retrained to obtain the target network model. The target network model is loaded into the cholangiocarcinoma lymph node metastasis imaging detection system, and the collected CT / MRI image data to be detected is imported into the cholangiocarcinoma lymph node metastasis imaging detection system to complete the detection of suspicious lymph node candidate regions.

2. The method for detecting lymph node metastasis in cholangiocarcinoma based on frequency perception and semantic constraint fusion according to claim 1, characterized in that, The frequency-aware progressive refinement module includes a frequency-driven convolutional unit and a stage-guided feature fusion unit.

3. The method for detecting lymph node metastasis in cholangiocarcinoma based on frequency perception and semantic constraint fusion according to claim 2, characterized in that, The frequency-driven convolutional unit includes a spatial convolution branch and a frequency domain residual branch, and the feature processing is completed according to the following steps: The input features are fed into the spatial convolution branch to obtain local features in the spatial domain, and into the frequency domain residual branch, which undergoes two-dimensional frequency domain transformation, convolution of the real and imaginary parts, and inverse transformation to obtain frequency domain residual features. By fusing local features in the spatial domain with residual features in the frequency domain, enhanced frequency-driven convolution output features are obtained.

4. The method for detecting lymph node metastasis in cholangiocarcinoma based on frequency perception and semantic constraint fusion according to claim 2, characterized in that, Based on the multi-level features obtained through multi-stage downsampling and frequency-driven convolution, the stage-guided feature fusion unit completes feature processing according to the following steps: Deep features are upsampled step by step, while deep semantic features and shallow detail features are fused together. Based on the current stage features, stage modulation weights are generated according to visual features and fused with stage guidance information to obtain stage fusion features. The stage fusion features are divided into different channel groups, which are then refined using lightweight self-attention units. The refined features of each channel group are then spliced ​​together to obtain frequency-aware progressively refined output features.

5. The method for detecting lymph node metastasis in cholangiocarcinoma based on frequency perception and semantic constraint fusion according to claim 2 or 4, characterized in that, The semantic constraint context attention module completes feature processing according to the following steps: The multi-scale visual features extracted by the YOLO-World network are arranged in order from coarse to fine and expanded into a sequence of visual tokens; Input text prompts related to the target of cholangiocarcinoma lymph node metastasis detection into a text encoder to obtain text features; A text query vector is generated based on the text features, and a key vector and a value vector are generated based on the visual token sequence. Construct directional constraint masks based on a scale order from coarse to fine; Construct a semantic relevance mask based on the distance between the text query vector and the key vector; The directional constraint mask and the semantic relevance mask are combined to form a guiding mask, which is then used to aggregate image and text features. After mapping, a multi-scale feature map is obtained.

6. The method for detecting lymph node metastasis in cholangiocarcinoma based on frequency perception and semantic constraint fusion according to claim 1, characterized in that, The bile duct cancer lymph node metastasis imaging detection system reads the CT / MRI images to be detected in real time, inputs them along with text prompts into the target network model for forward inference, marks suspicious lymph node candidate regions, and outputs candidate boxes, confidence scores, and text alignment scores.

7. The method for detecting lymph node metastasis in cholangiocarcinoma based on frequency perception and semantic constraint fusion according to claim 6, characterized in that, When the bile duct cancer lymph node metastasis imaging detection system marks suspicious lymph node candidate regions, it performs the following steps: a. Determine whether there are suspicious lymph node candidate regions in the CT / MRI image to be detected that meet the confidence threshold. If they exist, proceed to b; otherwise, proceed to c. b. The cholangiocarcinoma lymph node metastasis imaging detection system marks suspicious lymph node candidate areas, outputs candidate boxes, and indicates that there are suspicious lymph node candidate areas in the image, and then executes c. c. Determine if there are any unread images. If so, proceed with step a; otherwise, end the detection.

8. An imaging detection system for lymph node metastasis in cholangiocarcinoma based on frequency perception and semantic constraint fusion, characterized in that, include: The image acquisition module is used to acquire and annotate CT / MRI images of patients with cholangiocarcinoma, and obtain a dataset of CT / MRI images of cholangiocarcinoma lymph node metastases. The model pre-training module is used to train the YOLO-World network using a public dataset to obtain the YOLO-World pre-trained network model. The model structure modification module is used to modify the structure of the YOLO-World pre-trained network model: adding a frequency-aware progressive refinement module to the image branch and a semantic constraint context attention module to the image-text fusion stage; The model retraining module is used to input the CT / MRI image dataset of cholangiocarcinoma lymph node metastasis into the modified YOLO-World pre-trained network model and retrain it to obtain the target network model. The image detection module is used to load the target network model into the cholangiocarcinoma lymph node metastasis image detection system, import the acquired CT / MRI image data to be detected into the cholangiocarcinoma lymph node metastasis image detection system, and complete the detection of suspicious lymph node candidate regions.

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 7.

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 7.

Images