Tumor evolution process analysis method and system based on CT image and storage medium

Abstract

The application relates to the field of intelligent medicine, and provides a tumor evolution process analysis method and system based on CT images and a storage medium. The original CT image of a tumor to be detected is subjected to noise reduction pretreatment to obtain a tumor image matrix; the tumor image matrix is subjected to positioning labeling to obtain an image matrix containing tumor position information; the image matrix containing the tumor position information is input into a tumor image deep network model to obtain a tumor feature vector; the obtained tumor feature vector is projected onto a tumor evolution trajectory obtained in advance through training to obtain the position of the tumor to be detected on the tumor evolution trajectory; and the score of the tumor evolution process of the tumor to be detected is obtained and output according to the position of the tumor to be detected on the tumor evolution trajectory. The technical effect of accurately providing the stage of the tumor in the evolution process and assisting doctors in accurately judging the clinical characteristics of the tumor lesion is achieved.

Description

Technical Field

[0001] This invention relates to the field of intelligent medical technology, and specifically to a method for analyzing tumor evolution based on CT images. Background Technology

[0002] Cancer is a complex disease threatening the health of the Chinese population, primarily manifesting as malignant tumors. After normal cells transform into cancer cells or develop into pathologically or clinically observable tumors, the changes in cancer cells do not cease but rather continue to evolve. Tumor evolution is the phenomenon of malignant tumors becoming increasingly invasive during their growth, including accelerated growth, infiltration of surrounding tissues, and distant metastasis. During tumor evolution, the genotype and composition of cells within the tumor continuously change, leading to changes in the tumor's morphology and characteristics. Furthermore, because the growth and division of cancer cells are unregulated, different tumors may develop in the same direction or differentiate into different subtypes. Different subtypes of tumors may exhibit differences in phenotypic aspects such as immune characteristics, growth rate, and invasiveness, ultimately leading to differences in sensitivity to different anti-tumor drugs or radiotherapy. Therefore, clarifying the possible directions of tumor evolution and the stage the tumor is in during its evolution helps to accurately determine the tumor subtype and characteristics, thus providing a basis for clinical diagnosis and treatment selection.

[0003] Currently, doctors use biopsies to diagnose lesions through histopathological examination, determining the stage of tumor progression based on the distribution of tumor cells. Biopsies can cause some damage to the body and may carry risks of cancer cell spread and metastasis due to the use of equipment or bleeding. Furthermore, biopsies have a limited sampling range, only providing localized information about the tumor and failing to reflect the overall picture of the entire tumor tissue.

[0004] Computed tomography (CT) is a more convenient and less invasive diagnostic method. Since its invention in the 1970s, CT remains a core diagnostic tool for many diseases. CT provides high spatial resolution, offering high-quality image data for fully extracting tumor characteristics. With the development of computer vision, it has been discovered that CT images contain much information that is difficult to interpret with the naked eye. This information may be related to certain clinical manifestations of patients or even genetic factors, but machines can interpret this information.

[0005] In recent years, deep learning technology, widely used in image processing, has further advanced the breadth and depth of image information extraction, achieving excellent results in a series of downstream tasks such as disease classification and lesion segmentation. Because images can capture the entire lesion region without sampling bias, image analysis combining image processing and deep learning technologies provides a powerful tool for studying tumor progression in cancer. However, the research and application of CT images in tumor progression are still in the early exploratory stage, and mature methods and solutions are yet to be developed.

[0006] Therefore, there is an urgent need for a method that can quantify the tumor evolution process using CT images. Summary of the Invention

[0007] This invention provides a method for analyzing tumor evolution based on CT images. By constructing tumor evolution trajectories using CT images, the corresponding tumor evolution process scores are obtained. This solves the problem in existing technologies that can only analyze the categories in the tumor evolution process using CT images, but cannot refine discrete grades into continuous scores.

[0008] To achieve the above objectives, this invention provides a method for analyzing tumor progression based on CT images, the method comprising:

[0009] S1. The original CT images of the tumor to be detected are denoised and preprocessed to obtain the tumor image matrix;

[0010] S2. Localize and label the tumor image matrix to obtain an image matrix containing tumor location information;

[0011] S3. Input the image matrix containing tumor location information into the tumor image depth network model to obtain the tumor feature vector;

[0012] S4. Project the obtained tumor feature vector onto the tumor evolution trajectory obtained in advance through training to obtain the position of the tumor to be detected on the tumor evolution trajectory;

[0013] S5. Based on the position of the tumor to be detected on the tumor evolution trajectory, obtain and output the score of the tumor evolution process of the tumor to be detected.

[0014] Furthermore, preferably, methods for training pre-acquired tumor evolution trajectories include:

[0015] The tumor image deep network model is trained using a tumor image dataset to obtain tumor feature vectors; wherein, the tumor image dataset includes tumor images containing evolution stage labels;

[0016] Principal component analysis transformation is performed on the tumor feature vector to calculate the center point of the sample points corresponding to each evolution stage; wherein, the tumor feature vector is a feature vector with less than 100 dimensions, and each dimension of the feature vector is a floating-point number between 0 and 1;

[0017] The sample points and center points are mapped onto the principal component analysis space, and the tumor evolution trajectory is inferred using the minimum spanning tree algorithm.

[0018] Further, preferably, projecting the obtained tumor feature vector onto a pre-acquired tumor evolution trajectory obtained through training to obtain the position of the tumor to be detected on the tumor evolution trajectory also includes,

[0019] Obtain the tumor evolution trajectory and coordinates of the tumor to be detected;

[0020] Based on the tumor evolution trajectory and its coordinates, an evolution trajectory curve is plotted to form a visualization of tumor evolution.

[0021] Furthermore, preferably, the location of the input tumor image is marked on the tumor evolution visualization map;

[0022] The score of the tumor progression is displayed at the location of the input tumor image.

[0023] Further, preferably, in step S2, the method for locating and labeling the tumor image matrix to obtain an image matrix containing tumor location information includes:

[0024] The tumor image matrix is ​​input into an image segmentation algorithm to obtain a tumor image matrix containing tumor contour annotations;

[0025] Manual annotation and adjustment are performed on tumor images containing tumor contour annotations to obtain an image matrix containing tumor location information.

[0026] Furthermore, preferably, the starting point of the tumor evolution trajectory is 0, the ending point is 1, and each point of the tumor evolution trajectory corresponds to a tumor evolution process score.

[0027] Further, preferably, the method for obtaining a tumor image matrix by performing noise reduction preprocessing on the original CT images includes:

[0028] Read the image matrix and image information of the original CT images, and obtain the three-dimensional image matrix and image information list;

[0029] By changing the dimensions of the 3D image matrix according to the preset pixel spacing threshold, a 3D image with uniform pixel spacing can be obtained.

[0030] The 3D image is denoised and saved to obtain a preprocessed tumor image matrix.

[0031] To address the aforementioned problems, this invention also provides a tumor evolution process analysis system based on CT images, comprising a CT image preprocessing unit, an image localization unit, a tumor feature vector acquisition unit, and a tumor evolution process scoring unit; wherein,

[0032] The CT image preprocessing unit is used to perform noise reduction preprocessing on the original CT images of the tumor to be detected to obtain a tumor image matrix;

[0033] The image localization unit is used to locate and annotate the tumor image matrix to obtain an image matrix containing tumor location information;

[0034] The tumor feature vector acquisition unit is used to input an image matrix containing tumor location information into the tumor image depth network model to obtain tumor feature vectors.

[0035] The tumor evolution process scoring unit is used to project the obtained tumor feature vector onto the tumor evolution trajectory obtained in advance through training to obtain the position of the tumor to be detected on the tumor evolution trajectory; based on the position of the tumor to be detected on the tumor evolution trajectory, the unit obtains and outputs the score of the tumor evolution process of the tumor to be detected.

[0036] To address the above problems, the present invention also provides an electronic device, the electronic device comprising:

[0037] Memory, storing at least one instruction; and

[0038] The processor executes instructions stored in memory to implement the steps in the above-described method for analyzing tumor evolution based on CT images.

[0039] To address the aforementioned problems, the present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the aforementioned method for analyzing tumor evolution based on CT images.

[0040] This invention provides a method, system, electronic device, and storage medium for analyzing tumor evolution based on CT images, which accurately indicates the stage of tumor evolution, thereby assisting doctors in accurately determining the characteristics of tumor lesions. Attached Figure Description

[0041] Figure 1 This is a flowchart illustrating a method for analyzing tumor evolution based on CT images, provided in an embodiment of the present invention.

[0042] Figure 2 This is a schematic diagram of the structure of a tumor evolution process analysis system based on CT images provided in an embodiment of the present invention;

[0043] Figure 3This is a schematic diagram of the internal structure of an electronic device for a tumor evolution process analysis method based on CT images provided in an embodiment of the present invention.

[0044] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0045] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0046] This invention provides a method for analyzing tumor evolution based on CT images. By constructing a tumor evolution trajectory using CT images, a score of the tumor evolution process can be obtained. This solves the problem in the prior art that can only use pathological slides for diagnosis or CT images to predict the grading of tumor evolution, but cannot refine discrete grading into continuous scores.

[0047] Reference Figure 1 The diagram shown is a flowchart illustrating a method for analyzing tumor evolution based on CT images according to an embodiment of the present invention. This method can be executed by a device, which can be implemented in software and / or hardware.

[0048] In this embodiment, the tumor evolution process analysis method based on CT images includes steps S1 to S5:

[0049] S1. The original CT images of the tumor to be detected are denoised and preprocessed to obtain the tumor image matrix.

[0050] A method for obtaining a tumor image matrix by denoising and preprocessing the original CT images includes steps S11 to S13:

[0051] S11. Read the image matrix and image information of the original CT images to obtain a 3D image matrix and image information list. Specifically, load and read the image matrix data and information of the original CT images; the original images can be in formats such as DICOM, MHD, or NITFI. During the loading process, the following operations are performed: If the file format is DICOM, a series of 2D images are read in, and the coordinates of each 2D image are extracted. Based on the coordinates, the 2D image matrix is ​​rearranged into a 3D image matrix in memory; if the file format is MHD or NITFI, the 3D images are read directly to obtain the 3D image matrix. In addition, key information of the images is also read during the loading process, including the image starting coordinates, image matrix dimensions, and 3D pixel spacing. Furthermore, when reading image files through a user-specified path, the system automatically identifies the file format. If the system fails to automatically identify the format, the user must manually select and modify the format. The system reads the images into memory and displays the 3D image data and key information to the user in the display unit. The user can use the mouse to scroll through each layer of 2D images and use drop-down functions to view the image information of each layer.

[0052] S12. The dimensions of the 3D image matrix are changed according to a preset pixel spacing threshold to obtain a 3D image with uniform pixel spacing. After the loading process, the tumor image matrix is ​​interpolated using an interpolation algorithm to obtain the image matrix. The interpolation algorithm can be linear interpolation, nearest neighbor interpolation, etc. The pixel spacing after interpolation can be set, and an algorithm can be selected to interpolate the original image matrix. After interpolation is completed, the system generates the processed image matrix, which the user can view in the display unit.

[0053] S13. Denoise the 3D image and save it to obtain a preprocessed tumor image matrix. Denoising algorithms, such as moving average, are used. One or more denoising algorithms are selected to denoise the original image matrix, and the denoising effect can be viewed in real time. After denoising, a preprocessed image matrix is ​​generated and saved locally for further processing. In practice, the preprocessed image matrix can be viewed in the display unit. Supported formats include DICOM, MHD, and NIFI.

[0054] S2. The tumor image matrix is ​​localized and labeled to obtain an image matrix containing tumor location information.

[0055] In step S2, the method for locating and annotating the tumor image matrix to obtain an image matrix containing tumor location information includes: inputting the tumor image matrix into an image segmentation algorithm to obtain a tumor image matrix containing tumor contour annotations; and performing manual annotation and adjustment on the tumor image containing tumor contour annotations to obtain an image matrix containing tumor location information.

[0056] It should be noted that in the specific implementation process, tumor contour annotation can be done manually or with algorithm-assisted annotation. Manual annotation involves the user directly using the mouse to draw the contour or mark contour points around the lesion to obtain pixel-level location information of the lesion. Algorithm-assisted annotation involves adding or deleting annotation points using the mouse based on the lesion area automatically generated by the algorithm provided by the system to obtain pixel-level location information of the lesion. As an improvement to this embodiment, tumor image segmentation algorithms (such as U-Net, Mask-RCNN) can also be used to automatically generate pre-annotations of the tumor contour in the algorithm-assisted annotation stage. Then, annotation points can be added or deleted using the mouse on the pre-annotated tumor contour to generate the annotation result.

[0057] S3. Input the image matrix containing tumor location information into the tumor image depth network model to obtain the tumor feature vector.

[0058] The tumor imaging deep network model includes convolutional layers, pooling layers, fully connected layers, and custom layers that can be added by the user. It also includes a series of predefined loss functions to measure image similarity, as well as user-defined loss functions, for training the tumor imaging deep network model.

[0059] Specifically, deep learning networks for tumor imaging typically consist of three layers: convolutional layers, pooling layers, and fully connected layers. Users can also add new layers. Convolutional layers extract features from the previous level input image, pooling layers fuse local features and reduce dimensionality, and fully connected layers fuse global features. The deep learning network outputs features of a fixed dimension as a low-dimensional representation of the image; the feature dimension is set by the user. Furthermore, the similarity between the low-dimensional features output by the deep learning network is calculated using a predefined metric, serving as the model's loss function. In practice, the loss function can be one of the triplet loss, contrastive loss, or a custom loss function. The loss function is typically calculated based on the tumor type roughly labeled by the doctor and propagated back to the deep learning network via backpropagation to update network parameters for weakly supervised learning. After convergence, the network is fixed, and the last low-dimensional vector is extracted as the input to the next evolution trajectory construction module.

[0060] S4. Project the obtained tumor feature vector onto the tumor evolution trajectory obtained in advance through training to obtain the position of the tumor to be detected on the tumor evolution trajectory.

[0061] By using low-dimensional feature vectors output by a tumor imaging deep learning network, the trajectory of tumor evolution is inferred based on the distribution of categories in the low-dimensional feature space, and the trajectory coordinate matrix is ​​saved locally. The specific method of dimensionality reduction is not limited here; it can be, but is not limited to, PCA, LDA, SVD, etc. Dimensionality reduction achieves the following technical effects: reduced data dimensionality and storage space; reduced computation and model training time; removal of useless or irrelevant features, minimizing their impact on the model and improving model usability; removal of redundant features; and easier data visualization.

[0062] Methods that train pre-acquired tumor evolution trajectories include:

[0063] The tumor image deep network model is trained using a tumor image dataset to obtain tumor feature vectors; wherein the tumor image dataset includes tumor images with evolution stage labels; principal component analysis is performed on the tumor feature vectors to calculate the center points of sample points corresponding to each evolution stage; wherein the tumor feature vectors are feature vectors with less than 100 dimensions, and each dimension of the feature vector is a floating-point number between 0 and 1; the sample points and center points are mapped onto the principal component analysis space, and the tumor evolution trajectory is inferred using the minimum spanning tree algorithm.

[0064] Specifically, the method for obtaining the tumor evolution trajectory includes: training a tumor image deep network model using a tumor image dataset to obtain tumor feature vectors; wherein the tumor image dataset includes tumor images with evolution stage labels; performing principal component analysis on the tumor feature vectors to obtain the low-dimensional spatial coordinates of sample points corresponding to each tumor evolution stage; wherein the tumor feature vectors are feature vectors with less than 100 dimensions, and each dimension of the feature vector is a floating-point number between 0 and 1; based on the distribution of sample points and center points corresponding to each evolution stage in the low-dimensional feature space, using the minimum spanning tree algorithm to fit the distribution of sample points corresponding to each evolution stage in the low-dimensional feature space to obtain the tumor evolution trajectory and the tumor evolution trajectory coordinates.

[0065] In summary, the minimum spanning tree algorithm is used to fit the distribution of sample points at each evolution stage in a low-dimensional feature space, inferring the tumor evolution trajectory and saving the trajectory coordinate matrix locally. During the testing phase, the tumor sample to be detected is simply processed and projected onto the existing tumor evolution trajectory to obtain a score for the tumor evolution process.

[0066] S5. Based on the position of the tumor to be detected on the tumor evolution trajectory, obtain and output the score of the tumor evolution process of the tumor to be detected.

[0067] It should be noted that the tumor evolution trajectory starts at 0 and ends at 1, with each point on the trajectory corresponding to a score for a specific tumor evolution process. The corresponding score is obtained based on the point on the trajectory. In other words, the tumor evolution trajectory is divided into several small segments, with one end designated as the starting point (score 0) and the other end as the ending point (score 1). When the segments are sufficiently fine, the score corresponding to each point on the trajectory can be obtained. Then, the low-dimensional features of each tumor image are projected onto the trajectory to obtain a quantitative score for the tumor evolution process, and the corresponding score is saved locally.

[0068] In one specific embodiment, the process further includes a visualization process of the evolution trajectory of the tumor to be detected. After obtaining the tumor evolution trajectory and the tumor evolution trajectory coordinates, the evolution trajectory curve is drawn based on the tumor evolution trajectory and the tumor evolution trajectory coordinates to form a tumor evolution visualization map.

[0069] As an improvement to this embodiment, the location of the input tumor image is marked on the tumor evolution visualization map; a score for the corresponding tumor evolution process is displayed at the location of the input tumor image. Specifically, the location corresponding to the input tumor image is highlighted on the tumor evolution trajectory curve, and the score of the tumor evolution stage output by the quantification module is displayed. Furthermore, the visualization map can be made more interactive with the user, allowing the user to click on points on the two-dimensional scatter plot to obtain tumor images in the corresponding process, facilitating direct comparison of differences between tumor images at different positions on the evolution trajectory.

[0070] Furthermore, the system may include a feature visualization process, visualizing the convolutional kernels used to extract image features, which are obtained during the training of the tumor image deep network. Visualizing the convolutional kernels allows users to intuitively understand the features extracted by the deep image network, improving the interpretability of the model. In summary, visualizing the intermediate layer convolutional kernels enhances the interpretability of the deep image network; visualizing the tumor evolution trajectory allows users to intuitively see the stage of tumor evolution and adjacent tumor images, thereby better determining the characteristics of the tumor.

[0071] Taking lung adenocarcinoma, a major type of lung cancer, as an example, the progression from a lung lesion to lung cancer involves the following stages: atypical adenomatous hyperplasia—adenocarcinoma in situ—microinvasive adenocarcinoma—invasive adenocarcinoma. Invasive adenocarcinoma further progresses through three main stages: invasive adenocarcinoma grade I—invasive adenocarcinoma grade II—invasive adenocarcinoma grade III. This is the tumor progression process of lung adenocarcinoma.

[0072] In clinical diagnosis, the stage of tumor progression is primarily determined through pathological diagnosis, but this information is difficult to obtain from CT images through visual observation. The CT image-based tumor progression analysis method of this invention can obtain a quantitative score of the lung adenocarcinoma progression process. Compared to classification, it can evaluate the location of lung adenocarcinoma in the progression process with more detailed and continuous numerical values. Being able to determine the stage of lung adenocarcinoma progression using CT images can avoid the harm and side effects of biopsy sampling to patients.

[0073] In summary, the tumor evolution analysis method based on CT images of this invention, through deep image networks and similarity measurement, and by constructing the trajectory of tumor image distribution in a low-dimensional space, realizes the quantitative scoring of tumor evolution process and stage of tumor evolution from images, thereby assisting doctors in accurately analyzing the characteristics of tumors and avoiding the harm caused to patients by biopsy.

[0074] like Figure 2 As shown, this invention provides a tumor evolution process analysis system 200 based on CT images, which can be installed in an electronic device. Depending on the functions implemented, the CT image-based tumor evolution process analysis system 200 may include a CT image preprocessing unit 210, an image localization unit 220, a tumor feature vector acquisition unit 230, and a tumor evolution process scoring unit 240. The module described in this invention can also be referred to as a unit, which refers to a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, stored in the memory of the electronic device.

[0075] In this embodiment, the functions of each module / unit are as follows:

[0076] The CT image preprocessing unit 210 is used to perform noise reduction preprocessing on the original CT image of the tumor to be detected to obtain a tumor image matrix; the image localization unit 220 is used to locate and label the tumor image matrix to obtain an image matrix containing tumor location information; the tumor feature vector acquisition unit 230 is used to input the image matrix containing tumor location information into a tumor image depth network model to obtain tumor feature vectors; the tumor evolution process scoring unit 240 is used to project the obtained tumor feature vectors onto a tumor evolution trajectory obtained in advance through training to obtain the position of the tumor to be detected on the tumor evolution trajectory; and obtain and output a score of the tumor evolution process of the tumor to be detected based on the position of the tumor to be detected on the tumor evolution trajectory.

[0077] The specific implementation of the CT image-based tumor evolution process analysis system 200 of the present invention refers to the CT image-based tumor evolution process analysis method; it can accurately provide the stage of tumor evolution, thereby assisting doctors in accurately judging the characteristics of tumor lesions.

[0078] like Figure 3 As shown, the present invention provides an electronic device 3 for analyzing tumor evolution based on CT images.

[0079] The electronic device 3 may include a processor 30, a memory 31, and a bus. It may also include computer programs stored in the memory 31 and executable on the processor 30, such as a tumor evolution analysis program 32 based on CT images. The memory 31 may include both internal storage units of the CT image-based tumor evolution analysis system for artificial intelligence-based medicine and external storage devices. The memory 31 can be used not only to store application software and various types of data installed on the artificial intelligence-based medical assistance device, such as the code of the artificial intelligence-based medical assistance program, but also to temporarily store data that has been output or will be output.

[0080] The memory 31 includes at least one type of readable storage medium, such as flash memory, portable hard drive, multimedia card, card-type memory (e.g., SD or DX memory), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 31 can be an internal storage unit of the electronic device 3, such as a portable hard drive. In other embodiments, the memory 31 can be an external storage device of the electronic device 3, such as a plug-in portable hard drive, smart media card (SMC), secure digital card (SD), flash card, etc., equipped on the electronic device 3. Furthermore, the memory 31 can include both internal and external storage units of the electronic device 3. The memory 31 can be used not only to store application software and various types of data installed on the electronic device 3, such as code for a tumor evolution analysis program based on CT images, but also to temporarily store data that has been output or will be output.

[0081] In some embodiments, the processor 30 may be composed of integrated circuits, such as a single packaged integrated circuit or multiple integrated circuits with the same or different functions, including combinations of one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and various control chips. The processor 30 is the control unit of the electronic device, connecting various components of the entire electronic device via various interfaces and lines. It executes programs or modules stored in the memory 31 (e.g., tumor evolution analysis programs based on CT images) and calls data stored in the memory 31 to perform various functions and process data in the electronic device 3.

[0082] The bus can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This bus can be divided into an address bus, a data bus, a control bus, etc. The bus is configured to enable communication between the memory 31 and at least one processor 30, etc.

[0083] Figure 3 Only electronic devices with components are shown; it will be understood by those skilled in the art that... Figure 3 The structure shown does not constitute a limitation on the electronic device 3, and may include fewer or more components than shown, or combine certain components, or have different component arrangements.

[0084] For example, although not shown, the electronic device 3 may also include a power supply (such as a battery) to power various components. Preferably, the power supply can be logically connected to the at least one processor 30 through a power management system, thereby enabling functions such as charging management, discharging management, and power consumption management through the power management system. The power supply may also include one or more DC or AC power supplies, recharging systems, power fault detection circuits, power converters or inverters, power status indicators, and other arbitrary components. The electronic device 3 may also include various sensors, Bluetooth modules, Wi-Fi modules, etc., which will not be described in detail here.

[0085] Furthermore, the electronic device 3 may also include a network interface. Optionally, the network interface may include a wired interface and / or a wireless interface (such as a Wi-Fi interface, a Bluetooth interface, etc.), which is typically used to establish communication connections between the electronic device 3 and other electronic devices.

[0086] Optionally, the electronic device 3 may further include a user interface, which may be a display, an input unit (such as a keyboard), or a standard wired or wireless interface. Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen. The display may also be appropriately referred to as a screen or display unit, used to display information processed in the electronic device 3 and to display a visual user interface.

[0087] It should be understood that the embodiments described are for illustrative purposes only and are not limited to this structure in the scope of the patent application.

[0088] The tumor evolution process analysis program 32 based on CT images stored in the memory 31 of the electronic device 3 is a combination of multiple instructions. When run in the processor 30, it can achieve the following: S1, perform noise reduction preprocessing on the original CT image of the tumor to be detected to obtain a tumor image matrix; S2, perform localization annotation on the tumor image matrix to obtain an image matrix containing tumor location information; S3, input the image matrix containing tumor location information into a tumor image depth network model to obtain a tumor feature vector; S4, project the obtained tumor feature vector onto a tumor evolution trajectory obtained in advance through training to obtain the position of the tumor to be detected on the tumor evolution trajectory; S5, obtain and output a score of the tumor evolution process of the tumor to be detected based on the position of the tumor to be detected on the tumor evolution trajectory.

[0089] Specifically, the specific implementation method of the computer program when executed by the processor can be referred to the description of the relevant steps in the embodiment of the tumor evolution process analysis method based on CT images, which will not be repeated here.

[0090] In the several embodiments provided by this invention, it should be understood that the disclosed devices, systems, and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0091] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0092] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0093] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0094] Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be embraced within the invention. No appended diagram markings in the claims should be construed as limiting the scope of the claims.

[0095] Furthermore, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or systems stated in a system claim may also be implemented by a single unit or system through software or hardware. The term "second class" is used to indicate names and does not indicate any specific order.

[0096] 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. A method for analyzing tumor progression based on CT images, characterized in that, The methods include: S1. The original CT images of the tumor to be detected are denoised and preprocessed to obtain the tumor image matrix; S2. The tumor image matrix is ​​localized and labeled to obtain an image matrix containing tumor location information; S3. Input the image matrix containing tumor location information into the tumor image depth network model to obtain the tumor feature vector; S4. Project the obtained tumor feature vector onto the tumor evolution trajectory obtained in advance through training to obtain the position of the tumor to be detected on the tumor evolution trajectory; S5. Based on the position of the tumor to be detected on the tumor evolution trajectory, obtain and output a score of the tumor evolution process of the tumor to be detected; The method for training a pre-acquired tumor evolution trajectory includes: training a tumor image deep network model using a tumor image dataset to obtain tumor feature vectors; wherein the tumor image dataset includes tumor images containing evolution stage labels; performing principal component analysis transformation on the tumor feature vectors to calculate the center points of sample points corresponding to each evolution stage; wherein the tumor feature vectors are feature vectors with less than 100 dimensions, and each dimension of the feature vector is a floating-point number between 0 and 1; mapping the sample points and center points to the principal component analysis space, and using the minimum spanning tree algorithm to infer the tumor evolution trajectory.

2. The method for analyzing tumor progression based on CT images as described in claim 1, characterized in that, The obtained tumor feature vector is projected onto a tumor evolution trajectory obtained in advance through training to obtain the position of the tumor to be detected on the tumor evolution trajectory. The method also includes obtaining the tumor evolution trajectory and the tumor evolution trajectory coordinates of the tumor to be detected. Based on the tumor evolution trajectory and its coordinates, an evolution trajectory curve is plotted to form a visualization of tumor evolution.

3. The method for analyzing tumor progression based on CT images as described in claim 2, characterized in that, Mark the location of the input tumor image on the tumor evolution visualization map; The score of the tumor evolution process is displayed at the location of the input tumor image.

4. The method for analyzing tumor evolution based on CT images as described in claim 1, characterized in that, In step S2, the method for locating and labeling the tumor image matrix to obtain an image matrix containing tumor location information includes: The tumor image matrix is ​​input into an image segmentation algorithm to obtain a tumor image matrix containing tumor contour annotations; Manual annotation and adjustment are performed on the tumor image containing tumor contour annotations to obtain an image matrix containing tumor location information.

5. The method for analyzing tumor progression based on CT images as described in claim 1, characterized in that, Each point in the tumor evolution trajectory corresponds to a score in the tumor evolution process, with the starting score being 0 and the ending score being 1.

6. The method for analyzing tumor evolution based on CT images as described in claim 1, characterized in that, The method for obtaining a tumor image matrix by performing noise reduction preprocessing on the original CT images includes: Read the image matrix and image information of the original CT image to obtain a three-dimensional image matrix and image information list; The dimensions of the three-dimensional image matrix are changed according to a preset pixel spacing threshold to obtain a three-dimensional image with uniform pixel spacing. The three-dimensional image is denoised and saved to obtain a preprocessed tumor image matrix.

7. A tumor evolution process analysis system based on CT images, characterized in that, It includes a CT image preprocessing unit, an image localization unit, a tumor feature vector acquisition unit, and a tumor evolution scoring unit; among which, The CT image preprocessing unit is used to perform noise reduction preprocessing on the original CT image of the tumor to be detected to obtain a tumor image matrix. The image localization unit is used to locate and annotate the tumor image matrix to obtain an image matrix containing tumor location information; The tumor feature vector acquisition unit is used to input an image matrix containing tumor location information into a tumor image depth network model to obtain tumor feature vectors. The tumor evolution process scoring unit is used to project the obtained tumor feature vector onto a pre-acquired tumor evolution trajectory obtained through training, and obtain the position of the tumor to be detected on the tumor evolution trajectory; based on the position of the tumor to be detected on the tumor evolution trajectory, obtain and output a score of the tumor evolution process of the tumor to be detected; the method of training the pre-acquired tumor evolution trajectory includes: training the tumor image deep network model using a tumor image dataset to obtain tumor feature vectors; wherein, the tumor image dataset includes tumor images containing evolution stage labels; performing principal component analysis transformation on the tumor feature vectors to calculate the center point of the sample points corresponding to each evolution stage; wherein, the tumor feature vector is a feature vector with less than 100 dimensions, and each dimension of the feature vector is a floating-point number between 0 and 1; mapping the sample points and center points to the principal component analysis space, and using the minimum spanning tree algorithm to infer the tumor evolution trajectory.

8. An electronic device, characterized in that, The electronic device includes: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform the steps in the CT image-based tumor evolution analysis method as described in any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the tumor evolution process analysis method based on CT images as described in any one of claims 1 to 6.

Images