Eyelid tumor image recognition method and device, computer device and storage medium
Patent Information
- Application Number
- CN202410482443.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-04-22
- Publication Date
- 2026-10-27
AI Technical Summary
但是，由于眼睑肿瘤的形状各异，导致分割难度较大，现有的肿瘤分割方法不考虑器官结构且分割精度较低，影响后续的识别
[0055]经由上述的技术方案可知，与现有技术相比，本发明公开提供了一种眼睑肿瘤图像识别方法、装置、计算机设备和计算机可读存储介质，通过对CT扫描图像进行处理，得到对应的三维眼睑模型，并通过三维眼睑模型进行肿瘤分割，同时对CT扫描图像进行肿瘤分割处理，最后将两个分割结果进行融合处理，得到最终的眼睑肿瘤分割结果，利用医学图像分割技术实现了器官与肿瘤的分割，能够提高肿瘤分割的精度，来辅助医生进行后续的处理。
Smart Images
Figure CN122989505A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of image processing technology, and in particular to a method, apparatus, computer device, and computer-readable storage medium for eyelid tumor image recognition. Background Technology
[0002] Eyelid tissue originates from the ectoderm and mesoderm, with a complex origin. Its surface skin is the thinnest skin tissue in the body and is directly exposed to the external environment. Therefore, the eyelid is the most common site for tumors in the entire eye and its adnexa. Unlike tumors in other locations, eyelid tumors can damage eye tissues and their functions. Given the limited area and structural complexity of the eyelid, accurate identification of eyelid tumors is crucial for clinicians.
[0003] When identifying eyelid tumors, CT scan images of the eyelids are first acquired, and the shape of the eyelid tumor is then identified within these images. However, the varied shapes of eyelid tumors make segmentation challenging. Existing tumor segmentation methods do not consider organ structure and have low segmentation accuracy, affecting subsequent identification. Therefore, providing a method that can accurately identify the precise location of eyelid tumors within CT scan images is a problem urgently needing to be solved by those skilled in the art. Summary of the Invention
[0004] Therefore, it is necessary to provide an eyelid tumor image recognition method, device, computer equipment, and storage medium to address the aforementioned technical problems, which can effectively segment tumor images from CT scan images for subsequent auxiliary diagnosis.
[0005] On the one hand, a method for recognizing eyelid tumor images is provided, including:
[0006] Acquire CT scan images of the eyelids and preprocess the CT scan images;
[0007] The eyelid organ structure is extracted from the preprocessed CT scan image, the eyelid organ structure is segmented into a tumor, and the segmented tumor image is reconstructed in three dimensions to obtain the first segmentation result.
[0008] A segmentation model was constructed and trained using an eyelid tumor segmentation dataset;
[0009] The preprocessed CT scan image is input into the trained segmentation model to perform tumor segmentation and obtain a second segmentation result.
[0010] The first segmentation result and the second segmentation result are fused together to obtain the final eyelid tumor segmentation result.
[0011] In one embodiment, after obtaining the final eyelid tumor segmentation result, the method further includes:
[0012] A tumor recognition model is constructed using a ResNet convolutional neural network;
[0013] The final eyelid tumor segmentation result is input into the tumor recognition model for processing to obtain the eyelid tumor image recognition result.
[0014] In one embodiment, the steps of extracting eyelid organ structures from the preprocessed CT scan image, performing tumor segmentation on the eyelid organ structures, and reconstructing the segmented tumor image in three dimensions to obtain the first segmentation result include:
[0015] Obtain a clean eyelid organ image dataset, and determine the standard boundary contour of the eyelid organ and the corresponding feature point dataset based on the clean eyelid organ image dataset;
[0016] The CT scan image is converted into depth image data, feature points are extracted from the depth image data, and the distance between the feature points and their corresponding neighboring points is calculated. Feature points with a distance greater than a first distance change threshold are used to form a basic contour.
[0017] The eyelid organ contour is extracted from the depth image data according to the standard boundary contour, and the eyelid organ contour and the basic contour are fused to form the eyelid organ structure, and the eyelid organ structure is displayed with AR enhancement.
[0018] The eyelid organ structure is segmented into a tumor, and the segmented tumor image is reconstructed in three dimensions to obtain the first segmentation result.
[0019] In one embodiment, the step of constructing a segmentation model and training the segmentation model using an eyelid tumor segmentation dataset includes:
[0020] A segmentation model is constructed, which includes a CNN network and a multi-level pyramid feature extraction network connected in sequence, wherein the multi-level pyramid feature extraction network includes a four-layer pyramid convolutional model.
[0021] Obtain an eyelid tumor segmentation dataset, and use the eyelid tumor segmentation dataset to train, test, and validate the segmentation model.
[0022] In one embodiment, the step of fusing the first segmentation result and the second segmentation result to obtain the final eyelid tumor segmentation result includes:
[0023] A three-dimensional rectangular coordinate system and a rectangular coordinate system are constructed based on the first segmentation result and the second segmentation result, respectively.
[0024] The first segmentation result is mapped to the three-dimensional Cartesian coordinate system to obtain the corresponding first mapping result, and the second segmentation result is mapped to the Cartesian coordinate system to obtain the corresponding second mapping result.
[0025] Establish the association between the first mapping result and the second mapping result, and complete the fusion to obtain the corresponding fusion result. Use the fusion result as the final eyelid tumor segmentation result.
[0026] In one embodiment, the step of establishing the association between the first mapping result and the second mapping result, and completing the fusion to obtain the corresponding fusion result, and using the fusion result as the final eyelid tumor segmentation result includes:
[0027] The first sampling point of the first segmentation result and the second segmentation result are selected as the center, and a three-dimensional rectangular coordinate system and a rectangular coordinate system are constructed respectively.
[0028] Determine the target point coordinates (x, y, z) in the three-dimensional rectangular coordinate system, and project the target point into the rectangular coordinate system using homogeneous coordinates. The specific expression is as follows:
[0029]
[0030] In the formula, s is the scaling factor, K is the projection matrix, f is the acquisition field size of the CT scan image, and (X, Y) are the corresponding two-dimensional coordinates;
[0031] Feature points are collected again from the first segmentation result and the second segmentation result to obtain the corresponding first feature point set and second feature point set. The first feature point set is then transformed into a rectangular coordinate system using the expression.
[0032] The transformed first feature point set and the second feature point set are fused using a weighted average fusion method, where the expression for the weighted average fusion method is:
[0033] F(x, y) = w A ·A(x,y)+w B ·B(x, y);
[0034] In the formula, F(x, y) represents the fusion result, and w A w B Let A(x, y) represent the weighting coefficients, and B(x, y) represent the two feature points to be fused.
[0035] On the other hand, an eyelid tumor image recognition device is provided, comprising:
[0036] The scan image acquisition module is used to acquire CT scan images of the eyelids and preprocess the CT scan images;
[0037] The first segmentation module is used to extract eyelid organ structures from the preprocessed CT scan image, perform tumor segmentation on the eyelid organ structures, and perform three-dimensional reconstruction on the segmented tumor image to obtain the first segmentation result.
[0038] A segmentation model construction module is used to construct a segmentation model, and the segmentation model is trained using an eyelid tumor segmentation dataset;
[0039] The second segmentation module is used to input the preprocessed CT scan image into the trained segmentation model to perform tumor segmentation and obtain a second segmentation result.
[0040] The fusion module is used to fuse the first segmentation result and the second segmentation result to obtain the final eyelid tumor segmentation result.
[0041] In one embodiment, it further includes:
[0042] The recognition module is used to construct a tumor recognition model using a ResNet convolutional neural network; the final eyelid tumor segmentation result is input into the tumor recognition model for processing to obtain the eyelid tumor image recognition result.
[0043] In another aspect, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the following steps:
[0044] Acquire CT scan images of the eyelids and preprocess the CT scan images;
[0045] The eyelid organ structure is extracted from the preprocessed CT scan image, the eyelid organ structure is segmented into a tumor, and the segmented tumor image is reconstructed in three dimensions to obtain the first segmentation result.
[0046] A segmentation model was constructed and trained using an eyelid tumor segmentation dataset;
[0047] The preprocessed CT scan image is input into the trained segmentation model to perform tumor segmentation and obtain a second segmentation result.
[0048] The first segmentation result and the second segmentation result are fused together to obtain the final eyelid tumor segmentation result.
[0049] In another aspect, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, performs the following steps:
[0050] Acquire CT scan images of the eyelids and preprocess the CT scan images;
[0051] The eyelid organ structure is extracted from the preprocessed CT scan image, the eyelid organ structure is segmented into a tumor, and the segmented tumor image is reconstructed in three dimensions to obtain the first segmentation result.
[0052] A segmentation model was constructed and trained using an eyelid tumor segmentation dataset;
[0053] The preprocessed CT scan image is input into the trained segmentation model to perform tumor segmentation and obtain a second segmentation result.
[0054] The first segmentation result and the second segmentation result are fused together to obtain the final eyelid tumor segmentation result.
[0055] As can be seen from the above technical solution, compared with the prior art, the present invention discloses an eyelid tumor image recognition method, device, computer equipment, and computer-readable storage medium. By processing CT scan images, a corresponding three-dimensional eyelid model is obtained, and tumor segmentation is performed using the three-dimensional eyelid model. At the same time, tumor segmentation processing is performed on the CT scan images. Finally, the two segmentation results are fused to obtain the final eyelid tumor segmentation result. By using medical image segmentation technology, the separation of organs and tumors is realized, which can improve the accuracy of tumor segmentation and assist doctors in subsequent processing. Attached Figure Description
[0056] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0057] Figure 1 A flowchart of the eyelid tumor image recognition method provided in this application;
[0058] Figure 2 A flowchart of the method for obtaining the first segmentation result provided in this application;
[0059] Figure 3 A flowchart illustrating the method for training the segmentation model provided in this application;
[0060] Figure 4A flowchart illustrating the method for obtaining the final eyelid tumor segmentation result provided in this application;
[0061] Figure 5 This is a structural block diagram of an eyelid tumor image recognition device in one embodiment of this application;
[0062] Figure 6 This is an internal structural diagram of a computer device in one embodiment of this application. Detailed Implementation
[0063] 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.
[0064] In one embodiment, such as Figure 1 As shown, an image recognition method for eyelid tumors is provided, including the following steps:
[0065] Step S1: Acquire CT scan images of the eyelids and preprocess the CT scan images;
[0066] Step S2: Extract eyelid organ structures from the preprocessed CT scan images, perform tumor segmentation on the eyelid organ structures, and reconstruct the segmented tumor images in three dimensions to obtain the first segmentation result.
[0067] Step S3: Construct a segmentation model and train it using an eyelid tumor segmentation dataset;
[0068] Step S4: Input the preprocessed CT scan image into the trained segmentation model to segment the tumor and obtain the second segmentation result;
[0069] Step S5: The first segmentation result and the second segmentation result are fused to obtain the final eyelid tumor segmentation result.
[0070] Among them, CT (Computed Tomography) uses precisely collimated X-ray beams, gamma rays, ultrasound, etc., together with highly sensitive detectors to scan a part of the human body one section after another. It has the characteristics of fast scanning time and clear images, and can be used to examine a variety of diseases.
[0071] like Figure 1 As shown, in this embodiment, after obtaining the final eyelid tumor segmentation result, the process further includes:
[0072] Step S6: A tumor recognition model is constructed using a ResNet convolutional neural network; the final eyelid tumor segmentation result is input into the tumor recognition model for processing to obtain the eyelid tumor image recognition result.
[0073] Specifically, the tumor identification model can be a ResNet convolutional neural network or a network with different convolutional forms of ResNet convolutional neural networks, which can perform relevant detection and classification based on the eyelid tumor segmentation results to assist doctors in making judgments.
[0074] like Figure 2 As shown, in this embodiment, the steps of extracting eyelid organ structures from preprocessed CT scan images, performing tumor segmentation on the eyelid organ structures, and reconstructing the segmented tumor images in three dimensions to obtain the first segmentation result include:
[0075] Step S21: Obtain a clean eyelid organ image dataset, and determine the standard boundary contour of the eyelid organ and the corresponding feature point dataset based on the clean eyelid organ image dataset.
[0076] Step S22: Convert the CT scan image into depth image data, extract feature points from the depth image data, calculate the distance between the feature points and their corresponding neighboring points, and form a basic contour for feature points that are greater than the first distance change threshold.
[0077] Step S23: Extract the eyelid organ contour from the depth image data according to the standard boundary contour, and fuse the eyelid organ contour and the basic contour to form the eyelid organ structure, and perform AR enhancement display on the eyelid organ structure.
[0078] Step S24: Perform tumor segmentation on the eyelid organ structure, and reconstruct the obtained tumor image in three dimensions to obtain the first segmentation result.
[0079] Augmented Reality (AR) is a relatively new technology that integrates information from the real world and the virtual world. It uses computer technology to simulate and overlay virtual information onto physical entities that are difficult to experience in the real world, making the virtual information perceptible to human senses and creating a sensory experience that transcends reality.
[0080] In the process of extracting eyelid organ structures from preprocessed CT scan images, a three-dimensional eyelid model corresponding to the eyelid organ structures is obtained by processing the CT scan images.
[0081] like Figure 3As shown, in this embodiment, the steps of constructing a segmentation model and training the segmentation model using an eyelid tumor segmentation dataset include:
[0082] Step S31: Construct a segmentation model. The segmentation model includes a CNN network and a multi-level pyramid feature extraction network connected in sequence. The multi-level pyramid feature extraction network includes a four-layer pyramid convolutional model.
[0083] Step S32: Obtain the eyelid tumor segmentation dataset, and use the eyelid tumor segmentation dataset to train, test, and validate the segmentation model.
[0084] like Figure 4 As shown, in this embodiment, the steps of fusing the first segmentation result and the second segmentation result to obtain the final eyelid tumor segmentation result include:
[0085] Step S51: Construct a three-dimensional rectangular coordinate system and a rectangular coordinate system based on the first segmentation result and the second segmentation result, respectively;
[0086] Step S52: Map the first segmentation result to a three-dimensional Cartesian coordinate system to obtain the corresponding first mapping result; at the same time, map the second segmentation result to a Cartesian coordinate system to obtain the corresponding second mapping result.
[0087] Step S53: Establish the association between the first mapping result and the second mapping result, and complete the fusion to obtain the corresponding fusion result. Use the fusion result as the final eyelid tumor segmentation result.
[0088] In one embodiment, the steps of establishing the association between the first mapping result and the second mapping result, and completing the fusion to obtain the corresponding fusion result, and using the fusion result as the final eyelid tumor segmentation result include:
[0089] The first sampling point of the first segmentation result and the second segmentation result are selected as the center, and a three-dimensional rectangular coordinate system and a rectangular coordinate system are constructed respectively.
[0090] Determine the target point's coordinates (x, y, z) in a 3D Cartesian coordinate system, and project the target point into the Cartesian coordinate system using homogeneous coordinates. The specific expression is as follows:
[0091]
[0092] In the formula, s is the scaling factor, K is the projection matrix, f is the acquisition field size of the CT scan image, and (X, Y) are the corresponding two-dimensional coordinates;
[0093] Feature points are collected again from the first segmentation result and the second segmentation result to obtain the corresponding first feature point set and second feature point set. The first feature point set is then transformed into a rectangular coordinate system using the above expression.
[0094] The transformed first feature point set and the second feature point set are fused using a weighted average fusion method, where the expression for the weighted average fusion method is:
[0095] F(x, y) = w A ·A(x,y)+w B ·B(x, y);
[0096] In the formula, F(x, y) represents the fusion result, and w A w B Let A(x, y) represent the weighting coefficients, and B(x, y) represent the two feature points to be fused.
[0097] The above process combines three-dimensional and two-dimensional eyelid tumor segmentation results to achieve tumor segmentation and localization, making full use of the various information contained in the segmentation results and improving the accuracy of tumor segmentation.
[0098] In the above-mentioned eyelid tumor image recognition method, a corresponding three-dimensional eyelid model is obtained by processing the CT scan image, and the tumor is segmented using the three-dimensional eyelid model. At the same time, the CT scan image is also processed for tumor segmentation. Finally, the two segmentation results are fused to obtain the final eyelid tumor segmentation result. The medical image segmentation technology is used to segment organs and tumors, which can improve the accuracy of tumor segmentation and assist doctors in subsequent processing.
[0099] It should be understood that, although Figures 1-4 The steps in the flowchart are shown sequentially as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order in which these steps are executed, and they can be performed in other orders. Figures 1-4 At least some of the steps in the process may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least some of the sub-steps or stages of other steps.
[0100] In one embodiment, such as Figure 5 As shown, an eyelid tumor image recognition device is provided, including: a scanned image acquisition module 1, a first segmentation module 2, a segmentation model construction module 3, a second segmentation module 4, and a fusion module 5.
[0101] The image acquisition module 1 is used to acquire CT scan images of the eyelids and preprocess the CT scan images.
[0102] The first segmentation module 2 is used to extract eyelid organ structures from preprocessed CT scan images, perform tumor segmentation on the eyelid organ structures, and reconstruct the segmented tumor images in three dimensions to obtain the first segmentation result.
[0103] Module 3 for building the segmentation model is used to build the segmentation model and train it using the eyelid tumor segmentation dataset.
[0104] The second segmentation module 4 is used to input the preprocessed CT scan image into the trained segmentation model to perform tumor segmentation and obtain the second segmentation result.
[0105] The fusion module 5 is used to fuse the first segmentation result and the second segmentation result to obtain the final eyelid tumor segmentation result.
[0106] like Figure 5 As shown, the eyelid tumor image recognition device also includes a recognition module 6.
[0107] After obtaining the final eyelid tumor segmentation result, the recognition module 6 is used to: construct a tumor recognition model using a ResNet convolutional neural network; input the final eyelid tumor segmentation result into the tumor recognition model for processing, and obtain the eyelid tumor image recognition result.
[0108] In this embodiment, the first segmentation module 2 is used to extract eyelid organ structures from the preprocessed CT scan image, perform tumor segmentation on the eyelid organ structures, and reconstruct the segmented tumor image in three dimensions to obtain the first segmentation result, including:
[0109] Obtain a clean eyelid organ image dataset, and determine the standard boundary contour of the eyelid organ and the corresponding feature point dataset based on the clean eyelid organ image dataset;
[0110] The CT scan image is converted into depth image data, feature points are extracted from the depth image data, and the distance between the feature points and their corresponding neighboring points is calculated. Feature points with a distance greater than a first distance change threshold are used to form a basic contour.
[0111] The eyelid organ contour is extracted from the depth image data based on the standard boundary contour, and the eyelid organ contour and the basic contour are fused to form the eyelid organ structure, and the eyelid organ structure is displayed with AR enhancement.
[0112] The eyelid organ structure is segmented into a tumor, and the segmented tumor image is reconstructed in three dimensions to obtain the first segmentation result.
[0113] In this embodiment, the steps of segmenting the eyelid organ structure into a tumor and then reconstructing the segmented tumor image in three dimensions to obtain the first segmentation result include:
[0114] The eyelid organ structure is segmented into a tumor, and the segmented tumor image is reconstructed in three dimensions to obtain the first segmentation result.
[0115] In this embodiment, the segmentation model construction module 3 is used to construct a segmentation model. Training the segmentation model using the eyelid tumor segmentation dataset includes:
[0116] A segmentation model is constructed, which includes a CNN network and a multi-level pyramid feature extraction network connected sequentially. The multi-level pyramid feature extraction network includes a four-layer pyramid convolutional model.
[0117] Obtain an eyelid tumor segmentation dataset, and use the eyelid tumor segmentation dataset to train, test, and validate the segmentation model.
[0118] In this embodiment, the fusion module 5 is used to fuse the first segmentation result and the second segmentation result to obtain the final eyelid tumor segmentation result, including:
[0119] A three-dimensional rectangular coordinate system and a rectangular coordinate system are constructed based on the first segmentation result and the second segmentation result, respectively.
[0120] The first segmentation result is mapped to a three-dimensional Cartesian coordinate system to obtain the corresponding first mapping result. At the same time, the second segmentation result is mapped to a Cartesian coordinate system to obtain the corresponding second mapping result.
[0121] Establish the association between the first mapping result and the second mapping result, and complete the fusion to obtain the corresponding fusion result. The fusion result is used as the final eyelid tumor segmentation result.
[0122] In this embodiment, the steps of establishing the association between the first mapping result and the second mapping result, and completing the fusion to obtain the corresponding fusion result, and using the fusion result as the final eyelid tumor segmentation result include:
[0123] The first sampling point of the first segmentation result and the second segmentation result are selected as the center, and a three-dimensional rectangular coordinate system and a rectangular coordinate system are constructed respectively.
[0124] Determine the target point's coordinates (x, y, z) in a 3D Cartesian coordinate system, and project the target point into the Cartesian coordinate system using homogeneous coordinates. The specific expression is as follows:
[0125]
[0126] In the formula, s is the scaling factor, K is the projection matrix, f is the acquisition field size of the CT scan image, and (X, Y) are the corresponding two-dimensional coordinates;
[0127] Feature points are collected again from the first segmentation result and the second segmentation result to obtain the corresponding first feature point set and second feature point set. The first feature point set is then transformed into a rectangular coordinate system using the above expression.
[0128] The transformed first feature point set and the second feature point set are fused using a weighted average fusion method, where the expression for the weighted average fusion method is:
[0129] F(x, y) = w A ·A(x,y)+w B ·B(x, y);
[0130] In the formula, F(x, y) represents the fusion result, and w A w B Let A(x, y) represent the weighting coefficients, and B(x, y) represent the two feature points to be fused.
[0131] In the aforementioned eyelid tumor image recognition device, a corresponding three-dimensional eyelid model is obtained by processing the CT scan image, and the tumor is segmented using the three-dimensional eyelid model. At the same time, the CT scan image is also processed for tumor segmentation. Finally, the two segmentation results are fused to obtain the final eyelid tumor segmentation result. By using medical image segmentation technology, the separation of organs and tumors is realized, which can improve the accuracy of tumor segmentation and assist doctors in subsequent processing.
[0132] Specific limitations regarding the eyelid tumor image recognition device can be found in the limitations of the eyelid tumor image recognition method described above, and will not be repeated here. Each module in the aforementioned eyelid tumor image recognition device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.
[0133] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 6As shown, the computer device includes a processor, memory, network interface, and database 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, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database stores eyelid tumor image recognition data. The network interface communicates with external terminals via a network connection. When the processor executes the computer program, it implements an eyelid tumor image recognition method.
[0134] Those skilled in the art will understand that Figure 6 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.
[0135] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the following steps:
[0136] Acquire CT scan images of the eyelids and preprocess the CT scan images;
[0137] The eyelid organ structure is extracted from the preprocessed CT scan image, and the eyelid organ structure is segmented into a tumor. The segmented tumor image is then reconstructed in three dimensions to obtain the first segmentation result.
[0138] A segmentation model was constructed and trained using an eyelid tumor segmentation dataset;
[0139] The preprocessed CT scan image is input into the trained segmentation model to segment the tumor and obtain the second segmentation result;
[0140] The first and second segmentation results are fused to obtain the final segmentation result of the eyelid tumor.
[0141] In one embodiment, the processor, when executing a computer program, also performs the following steps:
[0142] After obtaining the final eyelid tumor segmentation results, the following is also included:
[0143] Construct a tumor identification model;
[0144] The final eyelid tumor segmentation results are input into the tumor recognition model for processing to obtain the eyelid tumor image recognition results.
[0145] In one embodiment, the processor, when executing a computer program, also performs the following steps:
[0146] The steps to build a tumor recognition model include:
[0147] A tumor recognition model was constructed using a ResNet convolutional neural network.
[0148] In one embodiment, the processor, when executing a computer program, also performs the following steps:
[0149] The steps of extracting eyelid organ structures from preprocessed CT scan images, performing tumor segmentation on the eyelid organ structures, and reconstructing the segmented tumor images in three dimensions to obtain the first segmentation result include:
[0150] Obtain a clean eyelid organ image dataset, and determine the standard boundary contour of the eyelid organ and the corresponding feature point dataset based on the clean eyelid organ image dataset;
[0151] The CT scan image is converted into depth image data, feature points are extracted from the depth image data, and the distance between the feature points and their corresponding neighboring points is calculated. Feature points with a distance greater than a first distance change threshold are used to form a basic contour.
[0152] The eyelid organ contour is extracted from the depth image data based on the standard boundary contour, and the eyelid organ contour and the basic contour are fused to form the eyelid organ structure, and the eyelid organ structure is displayed with AR enhancement.
[0153] The eyelid organ structure is segmented into a tumor, and the segmented tumor image is reconstructed in three dimensions to obtain the first segmentation result.
[0154] In one embodiment, the processor, when executing a computer program, also performs the following steps:
[0155] The steps for segmenting the eyelid organ structure into a tumor and then reconstructing the segmented tumor image in three dimensions to obtain the first segmentation result include:
[0156] The eyelid organ structure is segmented into a tumor, and the segmented tumor image is reconstructed in three dimensions to obtain the first segmentation result.
[0157] In one embodiment, the processor, when executing a computer program, also performs the following steps:
[0158] The steps for constructing a segmentation model and training it using an eyelid tumor segmentation dataset include:
[0159] A segmentation model is constructed, which includes a CNN network and a multi-level pyramid feature extraction network connected sequentially. The multi-level pyramid feature extraction network includes a four-layer pyramid convolutional model.
[0160] Obtain an eyelid tumor segmentation dataset, and use the eyelid tumor segmentation dataset to train, test, and validate the segmentation model.
[0161] In one embodiment, the processor, when executing a computer program, also performs the following steps:
[0162] The steps to fuse the first and second segmentation results to obtain the final eyelid tumor segmentation result include:
[0163] A three-dimensional rectangular coordinate system and a rectangular coordinate system are constructed based on the first segmentation result and the second segmentation result, respectively.
[0164] The first segmentation result is mapped to a three-dimensional Cartesian coordinate system to obtain the corresponding first mapping result. At the same time, the second segmentation result is mapped to a Cartesian coordinate system to obtain the corresponding second mapping result.
[0165] Establish the association between the first mapping result and the second mapping result, and complete the fusion to obtain the corresponding fusion result. The fusion result is used as the final eyelid tumor segmentation result.
[0166] In one embodiment, the steps of establishing the association between the first mapping result and the second mapping result, and completing the fusion to obtain the corresponding fusion result, and using the fusion result as the final eyelid tumor segmentation result include:
[0167] The first sampling point of the first segmentation result and the second segmentation result are selected as the center, and a three-dimensional rectangular coordinate system and a rectangular coordinate system are constructed respectively.
[0168] Determine the target point's coordinates (x, y, z) in a 3D Cartesian coordinate system, and project the target point into the Cartesian coordinate system using homogeneous coordinates. The specific expression is as follows:
[0169]
[0170] In the formula, s is the scaling factor, K is the projection matrix, f is the acquisition field size of the CT scan image, and (X, Y) are the corresponding two-dimensional coordinates;
[0171] Feature points are collected again from the first segmentation result and the second segmentation result to obtain the corresponding first feature point set and second feature point set. The first feature point set is then transformed into a rectangular coordinate system using the above expression.
[0172] The transformed first feature point set and the second feature point set are fused using a weighted average fusion method, where the expression for the weighted average fusion method is:
[0173] F(x, y) = w A ·A(x,y)+w B ·B(x, y);
[0174] In the formula, F(x, y) represents the fusion result, and w A w B Let A(x, y) represent the weighting coefficients, and B(x, y) represent the two feature points to be fused.
[0175] For specific limitations on the steps implemented by the processor when executing a computer program, please refer to the limitations on the method for eyelid tumor image recognition mentioned above, which will not be repeated here.
[0176] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor:
[0177] Acquire CT scan images of the eyelids and preprocess the CT scan images;
[0178] The eyelid organ structure is extracted from the preprocessed CT scan image, and the eyelid organ structure is segmented into a tumor. The segmented tumor image is then reconstructed in three dimensions to obtain the first segmentation result.
[0179] A segmentation model was constructed and trained using an eyelid tumor segmentation dataset;
[0180] The preprocessed CT scan image is input into the trained segmentation model to segment the tumor and obtain the second segmentation result;
[0181] The first and second segmentation results are fused to obtain the final segmentation result of the eyelid tumor.
[0182] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:
[0183] After obtaining the final eyelid tumor segmentation results, the following is also included:
[0184] Construct a tumor identification model;
[0185] The final eyelid tumor segmentation results are input into the tumor recognition model for processing to obtain the eyelid tumor image recognition results.
[0186] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:
[0187] The steps to build a tumor recognition model include:
[0188] A tumor recognition model was constructed using a ResNet convolutional neural network.
[0189] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:
[0190] The steps of extracting eyelid organ structures from preprocessed CT scan images, performing tumor segmentation on the eyelid organ structures, and reconstructing the segmented tumor images in three dimensions to obtain the first segmentation result include:
[0191] Obtain a clean eyelid organ image dataset, and determine the standard boundary contour of the eyelid organ and the corresponding feature point dataset based on the clean eyelid organ image dataset;
[0192] The CT scan image is converted into depth image data, feature points are extracted from the depth image data, and the distance between the feature points and their corresponding neighboring points is calculated. Feature points with a distance greater than a first distance change threshold are used to form a basic contour.
[0193] The eyelid organ contour is extracted from the depth image data based on the standard boundary contour, and the eyelid organ contour and the basic contour are fused to form the eyelid organ structure, and the eyelid organ structure is displayed with AR enhancement.
[0194] The eyelid organ structure is segmented into a tumor, and the segmented tumor image is reconstructed in three dimensions to obtain the first segmentation result.
[0195] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:
[0196] The steps for segmenting the eyelid organ structure into a tumor and then reconstructing the segmented tumor image in three dimensions to obtain the first segmentation result include:
[0197] The eyelid organ structure is segmented into a tumor, and the segmented tumor image is reconstructed in three dimensions to obtain the first segmentation result.
[0198] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:
[0199] The steps for constructing a segmentation model and training it using an eyelid tumor segmentation dataset include:
[0200] A segmentation model is constructed, which includes a CNN network and a multi-level pyramid feature extraction network connected sequentially. The multi-level pyramid feature extraction network includes a four-layer pyramid convolutional model.
[0201] Obtain an eyelid tumor segmentation dataset, and use the eyelid tumor segmentation dataset to train, test, and validate the segmentation model.
[0202] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:
[0203] The steps to fuse the first and second segmentation results to obtain the final eyelid tumor segmentation result include:
[0204] A three-dimensional rectangular coordinate system and a rectangular coordinate system are constructed based on the first segmentation result and the second segmentation result, respectively.
[0205] The first segmentation result is mapped to a three-dimensional Cartesian coordinate system to obtain the corresponding first mapping result. At the same time, the second segmentation result is mapped to a Cartesian coordinate system to obtain the corresponding second mapping result.
[0206] Establish the association between the first mapping result and the second mapping result, and complete the fusion to obtain the corresponding fusion result. The fusion result is used as the final eyelid tumor segmentation result.
[0207] In one embodiment, the steps of establishing the association between the first mapping result and the second mapping result, and completing the fusion to obtain the corresponding fusion result, and using the fusion result as the final eyelid tumor segmentation result include:
[0208] The first sampling point of the first segmentation result and the second segmentation result are selected as the center, and a three-dimensional rectangular coordinate system and a rectangular coordinate system are constructed respectively.
[0209] Determine the target point's coordinates (x, y, z) in a 3D Cartesian coordinate system, and project the target point into the Cartesian coordinate system using homogeneous coordinates. The specific expression is as follows:
[0210]
[0211] In the formula, s is the scaling factor, K is the projection matrix, f is the acquisition field size of the CT scan image, and (X, Y) are the corresponding two-dimensional coordinates;
[0212] Feature points are collected again from the first segmentation result and the second segmentation result to obtain the corresponding first feature point set and second feature point set. The first feature point set is then converted to a rectangular coordinate system using the above formula.
[0213] The transformed first feature point set and the second feature point set are fused using a weighted average fusion method, where the expression for the weighted average fusion method is:
[0214] F(x, y) = w A ·A(x,y)+w B ·B(x, y);
[0215] In the formula, F(x, y) represents the fusion result, and wA w B Let A(x, y) represent the weighting coefficients, and B(x, y) represent the two feature points to be fused.
[0216] For specific limitations on the steps implemented when a computer program is executed by a processor, please refer to the limitations on the method for eyelid tumor image recognition mentioned above, which will not be repeated here.
[0217] 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. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.
[0218] 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.
[0219] The above embodiments merely illustrate several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. 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 patent application should be determined by the appended claims.
Claims
1. A method for recognizing eyelid tumor images, characterized in that, include: Acquire CT scan images of the eyelids and preprocess the CT scan images; The eyelid organ structure is extracted from the preprocessed CT scan image, the eyelid organ structure is segmented into a tumor, and the segmented tumor image is reconstructed in three dimensions to obtain the first segmentation result. A segmentation model was constructed and trained using an eyelid tumor segmentation dataset; The preprocessed CT scan image is input into the trained segmentation model to perform tumor segmentation and obtain a second segmentation result. The first segmentation result and the second segmentation result are fused together to obtain the final eyelid tumor segmentation result.
2. The method for recognizing eyelid tumor images according to claim 1, characterized in that, After obtaining the final eyelid tumor segmentation results, the following is also included: A tumor recognition model is constructed using a ResNet convolutional neural network; The final eyelid tumor segmentation result is input into the tumor recognition model for processing to obtain the eyelid tumor image recognition result.
3. The method for recognizing eyelid tumor images according to claim 1, characterized in that, The steps of extracting eyelid organ structures from the preprocessed CT scan image, performing tumor segmentation on the eyelid organ structures, and reconstructing the segmented tumor image in three dimensions to obtain a first segmentation result include: Obtain a clean eyelid organ image dataset, and determine the standard boundary contour of the eyelid organ and the corresponding feature point dataset based on the clean eyelid organ image dataset; The CT scan image is converted into depth image data, feature points are extracted from the depth image data, and the distance between the feature points and their corresponding neighboring points is calculated. Feature points with a distance greater than a first distance change threshold are used to form a basic contour. The eyelid organ contour is extracted from the depth image data according to the standard boundary contour, and the eyelid organ contour and the basic contour are fused to form the eyelid organ structure, and the eyelid organ structure is displayed with AR enhancement. The eyelid organ structure is segmented into a tumor, and the segmented tumor image is reconstructed in three dimensions to obtain the first segmentation result.
4. The method for recognizing eyelid tumor images according to claim 1, characterized in that, The steps of constructing the segmentation model and training the segmentation model using the eyelid tumor segmentation dataset include: A segmentation model is constructed, which includes a CNN network and a multi-level pyramid feature extraction network connected in sequence, wherein the multi-level pyramid feature extraction network includes a four-layer pyramid convolutional model. Obtain an eyelid tumor segmentation dataset, and use the eyelid tumor segmentation dataset to train, test, and validate the segmentation model.
5. The method for recognizing eyelid tumor images according to claim 1, characterized in that, The step of fusing the first segmentation result and the second segmentation result to obtain the final eyelid tumor segmentation result includes: A three-dimensional rectangular coordinate system and a rectangular coordinate system are constructed based on the first segmentation result and the second segmentation result, respectively. The first segmentation result is mapped to the three-dimensional Cartesian coordinate system to obtain the corresponding first mapping result, and the second segmentation result is mapped to the Cartesian coordinate system to obtain the corresponding second mapping result. Establish the association between the first mapping result and the second mapping result, and complete the fusion to obtain the corresponding fusion result. Use the fusion result as the final eyelid tumor segmentation result.
6. The method for recognizing eyelid tumor images according to claim 5, characterized in that, The steps of establishing the association between the first mapping result and the second mapping result, completing the fusion, obtaining the corresponding fusion result, and using the fusion result as the final eyelid tumor segmentation result include: The first sampling point of the first segmentation result and the second segmentation result are selected as the center, and a three-dimensional rectangular coordinate system and a rectangular coordinate system are constructed respectively. Determine the target point coordinates (x, y, z) in the three-dimensional rectangular coordinate system, and project the target point into the rectangular coordinate system using homogeneous coordinates. The specific expression is as follows: In the formula, s is the scaling factor, K is the projection matrix, f is the acquisition field size of the CT scan image, and (X, Y) are the corresponding two-dimensional coordinates; Feature points are collected again from the first segmentation result and the second segmentation result to obtain the corresponding first feature point set and second feature point set. The first feature point set is then transformed into a rectangular coordinate system using the expression. The transformed first feature point set and the second feature point set are fused using a weighted average fusion method, where the expression for the weighted average fusion method is: F(x,y)=w A ·A(x,y)+w B ·B(x,y); In the formula, F(x, y) represents the fusion result, and w A w B Let A(x, y) represent the weighting coefficients, and B(x, y) represent the two feature points to be fused.
7. An eyelid tumor image recognition device, characterized in that, include: The scan image acquisition module is used to acquire CT scan images of the eyelids and preprocess the CT scan images; The first segmentation module is used to extract eyelid organ structures from the preprocessed CT scan image, perform tumor segmentation on the eyelid organ structures, and perform three-dimensional reconstruction on the segmented tumor image to obtain the first segmentation result. A segmentation model construction module is used to construct a segmentation model, and the segmentation model is trained using an eyelid tumor segmentation dataset; The second segmentation module is used to input the preprocessed CT scan image into the trained segmentation model to perform tumor segmentation and obtain a second segmentation result. The fusion module is used to fuse the first segmentation result and the second segmentation result to obtain the final eyelid tumor segmentation result.
8. The eyelid tumor image recognition device according to claim 7, characterized in that, Also includes: The recognition module is used to construct a tumor recognition model using a ResNet convolutional neural network; the final eyelid tumor segmentation result is input into the tumor recognition model for processing to obtain the eyelid tumor image recognition result.
9. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, 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 6.
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 6.