Automatic sketching method for rectal cancer radiotherapy target area and related equipment

Abstract

The invention provides an automatic sketching method for a rectal cancer radiotherapy target area and related equipment, and is applied to the technical field of data processing. The method comprises the following steps: acquiring target user information, target region label information, to-be-trained target region label information, a training data set matched with the to-be-trained target region label information, and a preset target region automatic sketching model based on nnUNET; processing the training data set based on the to-be-trained target region label information to generate a preset sketching rule; processing a preset target area automatic sketching model based on nnUNET to generate a target area automatic sketching model; processing the target user information, and generating multi-modal CT image information of the target user and clinical priori knowledge information matched with the target user; and processing the multi-modal CT image information of the target user, the target region label information and the clinical priori knowledge information matched with the target user based on the target region automatic sketching model to generate target region sketching information.

Description

Technical Field

[0001] The invention relates to the technical field of data processing, and in particular to an automatic delineation method for a rectal cancer radiotherapy target area and related equipment. Background Art

[0002] Radiotherapy is an important treatment for rectal cancer. Tumor radiotherapy is a local treatment method that uses radiation to treat tumors. About 70% of cancer patients need radiotherapy during the treatment of cancer, and about 40% of cancers can be cured by radiotherapy. The role and status of radiotherapy in tumor treatment are becoming increasingly prominent, and it has become one of the main means of treating malignant tumors.

[0003] The automatic delineation model of rectal cancer target area medical images based on deep learning can greatly shorten the time for doctors to modify the target area during clinical treatment. However, due to the poor quality of CT imaging of rectal cancer target areas, the current automatic delineation model of rectal cancer target areas is often based on MRI. There are very few automatic delineation models of rectal cancer target areas based on CT images, and the model delineation effect is not ideal and cannot meet the requirements of clinical use. At present, there are many scenarios in the clinical treatment of rectal cancer using CT-linac (radiotherapy linear accelerator guided by CT images), and the delineation styles of rectal cancer targets are inconsistent in various medical centers. Therefore, it is urgent to develop a multi-target automatic delineation model for rectal cancer based on CT images to assist in precise treatment and accelerate the user's treatment process.

[0004] It should be noted that the information disclosed in the above background technology section is only used to enhance the understanding of the background of the present disclosure, and therefore may include information that does not constitute the prior art known to ordinary technicians in the field. Summary of the invention

[0005] The purpose of this application is to provide a method and related equipment for automatically outlining the target area for radiotherapy of rectal cancer, which at least to a certain extent overcomes the problems existing in the prior art, by integrating and learning the user's previous images (multimodal images such as plain CT and enhanced CT), the anatomical structure limitations of the surrounding organs and the clinical prior knowledge of the center, and fully learning the internal information of the custom label (that is, the target area contour that the doctor needs to train), thereby achieving the goal of significantly shortening the manual outlining time during the implementation of clinical radiotherapy, while reducing the subjective differences between doctors due to different experiences, and meeting the customized needs of single-center automatic outlining.

[0006] Other features and advantages of the present application will become apparent from the following detailed description, or may be learned in part by the practice of the present invention.

[0007] According to one aspect of the present application, a method for automatically delineating a target area for radiotherapy of rectal cancer is provided, comprising: obtaining target user information, target area label information, target area label information to be trained, a training data set matching the target area label information to be trained, and a preset target area automatic delineation model based on nnUNET, wherein the training data set comprises multimodal CT image information of other users and clinical prior knowledge information matching the other users, and the multimodal CT images comprise plain scan CT images and enhanced CT images; based on the target area label information to be trained, the training data set matching the target area label information to be trained is processed to generate a preset delineation rule, wherein the preset delineation rule is used to characterize the delineation labels of different plain scan CT images and different enhanced CT images. The method comprises the following steps: obtaining the anatomical structure restriction information of organs adjacent to the target area label; preprocessing the training data set that matches the target area label information to be trained to generate a training data set with target feature data; processing the preset target area automatic delineation model based on nnUNET based on the training data set with target feature data and the preset delineation rules to generate a target target area automatic delineation model; processing the target user information to generate the target user's multimodal CT image information and the clinical prior knowledge information matching the target user; processing the target user's multimodal CT image information, the target target area label information and the clinical prior knowledge information matching the target user based on the target target area automatic delineation model to generate the target target area delineation information.

[0008] Another aspect of the present application is an automatic delineation device for rectal cancer radiotherapy target area, characterized in that the device includes: an acquisition module, used to acquire target user information, target area label information, target area label information to be trained, a training data set matching the target area label information to be trained, and a preset target area automatic delineation model based on nnUNET, wherein the training data set includes multimodal CT image information of other users and clinical prior knowledge information matching other users, and the multimodal CT images include plain scan CT images and enhanced CT images; a processing module, used to process the training data set matching the target area label information to be trained based on the target area label information to be trained, and generate a preset delineation rule, wherein the preset delineation rule is used to characterize different plain scan CT images, different Enhance the delineation standards of CT images and the anatomical structure restriction information of organs adjacent to the target area label; pre-process the training data set that matches the target area label information to be trained to generate a training data set with target feature data; process the preset target area automatic delineation model based on nnUNET based on the training data set with target feature data and the preset delineation rules to generate a target target area automatic delineation model; process the target user information to generate the target user's multimodal CT image information and the clinical prior knowledge information matching the target user; based on the target target area automatic delineation model, process the target user's multimodal CT image information, the target target area label information and the clinical prior knowledge information matching the target user to generate the target target area delineation information.

[0009] According to another aspect of the present application, an electronic device is characterized in that it includes: a first processor; and a memory for storing executable instructions of the first processor; wherein the first processor is configured to implement the above-mentioned method of automatically delineating the target area for radiotherapy of rectal cancer by executing the executable instructions.

[0010] According to another aspect of the present application, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a second processor, the method for automatically delineating a target area for radiotherapy of rectal cancer is implemented.

[0011] According to another aspect of the present application, a computer program product is provided, comprising a computer program, wherein the computer program implements the above-mentioned method for automatically delineating a target area for radiotherapy of rectal cancer when executed by a third processor.

[0012] The present application provides an automatic delineation method and related equipment for the target area of ​​rectal cancer radiotherapy, wherein the server obtains target user information, target area label information, target area label information to be trained, training data set and preset target area automatic delineation model based on nnUNET, and the training data set contains multimodal CT images of other users and clinical prior knowledge. Based on the target area label information to be trained, the training data set is processed to generate preset delineation rules, including generating preset delineation feature points, and then processing the inverse mapping and target interpolation to obtain the initial delineation information, and finally generating rules based on the preset loss function. Then, the training data set is preprocessed, and a training data set with target feature data is generated through the steps of feature extraction, data feature analysis, adjacent feature generation, sampling ratio determination, sampling feature generation and data group construction. Then, the preset target area automatic delineation model is trained using the data set and the preset delineation rules, including generating clinical prior knowledge and anatomical structure restriction information, dividing the training set and the verification set, extracting the data set and processing the target data set for training, and finally determining the target area automatic delineation model based on the verification set test.

[0013] For the target user, its information is processed to generate multimodal CT image information and clinical prior knowledge information, including generating test report information, generating clinical prior knowledge based on it, and extracting plain scan and enhanced CT images to generate multimodal CT image information. Finally, based on the target area automatic delineation model, the target area label information generates the area to be delineated information, the clinical prior knowledge information generates delineation restriction information, the multimodal CT image information generates channel information, the initial target area delineation information is generated based on the area to be delineated information and channel information, and then the target area delineation information is generated in combination with the delineation restriction information to assist clinical precision radiotherapy.

[0014] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 A flow chart showing a method for automatically delineating a target area for radiotherapy of rectal cancer provided by an embodiment of the present application; Figure 2 A schematic structural diagram of an automatic delineation device for a rectal cancer radiotherapy target area provided in one embodiment of the present application is shown; Figure 3 A schematic diagram of the nnUNET network structure of an automatic target area delineation model provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0016] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0017] Combine the following Figure 1 To describe the automatic delineation method of the target area for radiotherapy of rectal cancer according to the exemplary embodiment of the present application. It should be noted that the following application scenarios are only shown to facilitate understanding of the spirit and principle of the present application, and the embodiments of the present application are not limited in this respect. On the contrary, the embodiments of the present application can be applied to any applicable scenario.

[0018] In one embodiment, the present application also proposes an automatic delineation method for rectal cancer radiotherapy target area and related equipment. Figure 1 The following schematically shows a flow chart of a method for automatically delineating a target area for radiotherapy of rectal cancer according to an embodiment of the present application. Figure 1 As shown, the method is applied to a server, comprising: S101, obtaining target user information, target target area label information, target area label information to be trained, a training data set matching the target area label information to be trained, and a preset target area automatic delineation model based on nnUNET.

[0019] In one embodiment, the training data set includes multimodal CT image information of other users and clinical prior knowledge information matching other users, and the multimodal CT images include plain scan CT images and enhanced CT images. Specifically, the target user information may include the user's basic identity information, such as name, age, gender, medical record number, etc., which is used to uniquely identify the user. For example, user Zhang San, male, 55 years old, medical record number is 20230801001. It may also include the user's clinical medical history information, such as whether there are other underlying diseases (such as hypertension, diabetes, etc.), which may have a certain impact on the radiotherapy process and target area delineation. For example, the user has a history of hypertension and has been taking antihypertensive drugs for a long time, which may affect his physical response during radiotherapy, and thus needs to be considered when delineating the target area and formulating the radiotherapy plan.

[0020] For rectal cancer radiotherapy, the target target label information can be specific to the rectal part where the tumor is located (such as the upper, middle and lower rectum), the size of the tumor, the shape and other characteristics. For example, the target target is a tumor in the middle rectum, which is about 3cm×2cm in size and irregular in shape. It can also include preliminary markings of areas around the tumor that may be affected, such as tissues within a certain range from the edge of the tumor. These areas also need to be focused on during radiotherapy to ensure that the tumor is treated as thoroughly as possible and reduce the risk of recurrence. The target label information to be trained can be annotated examples of different types of rectal cancer targets collected from a large number of clinical cases. For example, for early rectal cancer, the target label may focus on the primary tumor and its adjacent tissues; for mid- and late-stage rectal cancer, the target label may also include surrounding lymph node metastasis areas, etc. Take a set of training data as an example, which contains target labeling of rectal cancer cases of different stages, locations and sizes. These diverse annotations will be used to train the model so that it can adapt to target delineation in various situations.

[0021] The training data set includes multimodal CT image information of other users (i.e. previous users) and clinical prior knowledge information matching these users. In terms of multimodal CT image information, plain scan CT images can clearly display the anatomical structure of the rectum and surrounding tissues, while enhanced CT images (venous phase and arterial phase) can better display the blood supply of the tumor. For example, a user's plain scan CT showed thickening of the rectal wall, and the enhanced CT venous phase image showed obvious enhancement of the tumor site, indicating rich blood supply. Clinical prior knowledge information can be the user's previous treatment (such as whether he has undergone surgery, chemotherapy, etc.), family history, etc. If there is a history of colorectal cancer in the user's family, this may increase their genetic risk factors for cancer, which can be used as a reference factor when training the model so that the model can learn the impact of different factors on the characteristics of the target area.

[0022] like Figure 3 As shown in the figure, nnUNET is an advanced deep learning network structure. Its preset target area automatic delineation model plays a core role in this system. It has the ability to automatically extract medical image features. For example, it can accurately identify key information such as the contour of the rectum and the approximate location of the tumor from multimodal CT images. The network structure of the model (as shown in the document) Figure 1) contains multiple convolutional layers (such as 3x3×3.Conv, etc.), which can extract features from images layer by layer. The downsampling and upsampling processes help capture information at different scales, and the jump connection can fuse feature information at different levels to more comprehensively understand the image content. During the training process, the model continuously adjusts internal parameters based on the input training data set (multimodal CT images and related knowledge) to improve the accuracy of target delineation. For example, by learning a large amount of image data with accurate target area annotations, the model gradually learns to distinguish the characteristic differences between tumor tissue and normal tissue, so that it can accurately delineate the target area on new user images. In actual applications, after inputting the multimodal CT images and related information of the target user, the model will output the predicted target area delineation results based on the knowledge and patterns it has learned, providing a reference for doctors and assisting them in making more accurate radiotherapy plans.

[0023] S102: Based on the target area label information to be trained, a training data set matching the target area label information to be trained is processed to generate a preset delineation rule.

[0024] In one embodiment, the target area label information to be trained is processed to generate preset delineation feature points, wherein the preset delineation feature points are used to characterize the delineation feature points corresponding to the target area label. The preset delineation feature points are key feature points corresponding to the target area label. Taking the rectal tumor target area as an example, possible feature points include the center position of the tumor, the most prominent point of the tumor edge, and the key point at the junction with the adjacent organs. Assuming that in a case, the tumor is located on the posterior wall of the middle rectum, its center coordinates (in the CT image coordinate system) can be used as an important preset delineation feature point, which can roughly determine the location center of the tumor and help to determine the overall range of the tumor in the subsequent process. The most prominent point of the tumor edge close to the mesorectum is also a key feature point, because this may be the key part of the tumor invading the surrounding tissues, and special attention should be paid to it when delineating. The feature points at the border adjacent to the bladder are equally important, which defines the boundary between the target area and the bladder to avoid mis-irradiation of the bladder during radiotherapy. By processing these preset delineation feature points, the morphology and position of the target area can be described more accurately, providing a basis for subsequent model training and delineation operations.

[0025] Based on the preset processing rules, the preset delineation feature points and the training data set matching the target area label information to be trained are processed to generate the inverse mapping and target interpolation. The inverse mapping and target interpolation are obtained by processing the preset delineation feature points and the training data set. For example, in the channel information processing of multimodal CT images, it is assumed that the training data set contains plain CT and enhanced CT images of multiple users, and each image has different channel levels (such as different grayscale and other feature levels). Based on the preset delineation feature points (such as the tumor center position mentioned above), the matching feature points are found in the plain CT image, and the corresponding matching feature points are found in the enhanced CT image. Then, the affine transformation matrix is ​​calculated through these matching feature points to obtain the inverse mapping and target interpolation. The inverse mapping can accurately correspond the spatial positions of the feature points in different modal images, such as mapping a feature point in the plain CT image to the corresponding position in the enhanced CT image, so as to better integrate the multimodal image information. The target interpolation is used to estimate the data value of the unknown position between the known feature points. For example, between two adjacent tumor edge feature points, more continuous tumor boundary information can be obtained through target interpolation, which supplements the details of the imaging data and enables the model to more accurately understand the shape and range of the target area. Especially for irregularly shaped tumor targets, target interpolation can provide more precise boundary description and improve the accuracy of delineation.

[0026] Based on inverse mapping and target interpolation, the training data set that matches the target area label information to be trained is processed to obtain initial delineation information, wherein the initial delineation information includes the organ image located at the target area label position. The initial delineation information obtained after processing the training data set based on inverse mapping and target interpolation includes the organ image located at the target area label position. Taking the rectal cancer target area as an example, the initial delineation information will display a rough image of the rectal tumor and its surrounding organs (such as male organs such as the bladder and seminal vesicle or female organs such as the uterus and vagina). In the initial delineation image, it can be seen that the rectal tumor presents a certain shape (such as round, irregular, etc.), and its density has different manifestations in plain CT and enhanced CT images, and the outlines of the surrounding organs are also clearly visible. For example, in male users, the relative position relationship between the seminal vesicle and the rectal tumor is reflected in the initial delineation information, which helps doctors and models to further determine whether the tumor invades the surrounding organs. The initial delineation information provides a basic framework for subsequent precise delineation. Based on this initial information, the model can combine more knowledge and algorithms to gradually optimize and refine the target area delineation to make it more consistent with the actual pathological situation.

[0027] The initial delineation information is processed based on the preset loss function to generate preset delineation rules. The preset delineation rules are used to characterize the delineation standards of different plain scan CT images and different enhanced CT images and the anatomical structure restriction information of the organs adjacent to the target area label. The preset delineation rules cover the standards for delineating the target area of ​​different plain scan CT images and enhanced CT images and the anatomical structure restriction information of the organs adjacent to the target area label. For example, in plain scan CT images, there is a certain range of delineation standards for the thickness, density and other characteristics of the rectal wall. The normal rectal wall may appear as a uniform medium-density shadow on plain scan CT, with a thickness within a certain range. If local thickening or density changes are found and meet certain numerical standards (such as thickness exceeding a certain threshold or density higher than normal tissue to a certain extent), it may indicate that the lesion area needs to be delineated. For enhanced CT images (venous phase and arterial phase), the enhancement characteristics of tumor tissue are different in different periods. For example, it may quickly enhance to a higher density in the arterial phase, and the degree of enhancement may vary in the venous phase. The tumor target area range is accurately delineated according to the standards of these enhancement characteristics. At the same time, the anatomical structures of organs adjacent to the target label, such as the bladder, prostate (male) or uterus, vagina (female), etc., limit the boundaries of the target area. For example, when outlining the target area of ​​a rectal tumor, it is not allowed to exceed the boundaries of the normal tissue around the rectum, and damage to adjacent organs must be avoided to ensure the safety and effectiveness of radiotherapy. These rules are summarized based on a large amount of clinical experience and anatomical knowledge. By incorporating these rules into the training process, the model learns the correct outline standards and boundary restrictions.

[0028] The preset loss function (such as a mixed loss function, including the cross entropy function and the Dice loss function) processes the initial delineation information to generate the preset delineation rules. The cross entropy function mainly measures the classification difference between the model prediction result and the true label (such as the accurate target area delineation marked by the doctor). In this scenario, it is to judge whether the model accurately classifies the target area and the non-target area. For example, if the model mistakenly divides the normal tissue into the target area or vice versa, the cross entropy function will give a large loss value, prompting the model to adjust the parameters to improve the classification accuracy. The Dice loss function focuses on the prediction error of the model at the boundary pixels and evaluates the accuracy of the delineation of the target area boundary. For example, whether the delineation of the tumor edge is accurate, whether there is over-delineation or under-delineation. By continuously adjusting the model parameters, the loss function value is minimized, and the optimized preset delineation rules are obtained. These rules not only include the accurate definition of the target area itself, but also consider multiple factors such as the anatomical relationship with adjacent organs, so that the final generated preset delineation rules are more scientific and reasonable, and can accurately guide the automatic delineation process of the target area in practical applications, improving the accuracy and safety of radiotherapy.

[0029] The method includes a calculation formula for obtaining a target interpolation value, and the calculation formula is: ;in, is the weight of the cross entropy loss function, is the weight of the Dice loss function; Among them, CE represents the cross entropy loss function, y represents the true label, and p represents the predicted probability of the model; Among them, Dice represents the loss function, y represents the true label, and p represents the predicted probability of the model. is a very small constant to avoid the denominator being zero.

[0030] This formula is used to obtain the target interpolation and plays an important role in the deep learning-based automatic delineation model for rectal cancer targets. To better understand the formula, we assume a simple example scenario: there is a small rectal cancer user image dataset containing plain CT and enhanced CT images of 10 users, and we want to train a model to accurately delineate the tumor target.

[0031] For the cross entropy loss function , assuming that in this data set, for a certain pixel, its true label (indicating whether the pixel belongs to the tumor target area, 1 for yes, 0 for no) is 1, and the model predicts that the probability p of the pixel belonging to the tumor target area is 0.8. Then calculate according to the formula: First calculate .

[0032] Then calculate .

[0033] So the cross entropy loss value of this pixel is 0.223. If this calculation is performed on each pixel of the entire image and the sum is calculated (i.e. Operation), we get the cross entropy loss value of the entire image. This value reflects the degree of difference between the model prediction and the true label in classification. The smaller the value, the more accurate the model classification. During the training process, the model will continuously adjust the parameters according to this loss value to reduce the cross entropy loss and improve the classification accuracy of tumor target and non-target areas.

[0034] For Dice loss function Assume that in a certain area in the image (containing multiple pixels), the sum of the true labels y (pixels belonging to the tumor target area are 1, otherwise they are 0) (Indicates that 50 pixels in this area belong to the real tumor target area), the sum of the model prediction probabilities (The model predicts that 60 pixels in this area belong to the tumor target area), and (Indicates that there are 45 pixels whose true label is 1 and whose model prediction probability is also high). (a very small constant).

[0035] First calculate the numerator part .

[0036] Then calculate the denominator .

[0037] The Dice coefficient of this area is 1-(90.001 / 110.001) = 0.182. The closer the Dice coefficient is to 1, the higher the overlap between the target area predicted by the model and the actual target area, and the more accurate the boundary delineation. During the training process, the model will adjust parameters to increase the Dice coefficient and reduce the Dice loss, thereby optimizing the delineation effect of the target area boundary.

[0038] Standard interpolation calculation formula Example. Suppose during model training, we set the weight of the cross entropy loss function , based on the cross entropy loss value (assuming CE=0.3) and Dice loss value (assuming Dice=0.2) calculated previously.

[0039] First calculate Then calculate . So the target interpolation L=0.15+0.1=0.25. This target interpolation value will be used to adjust the model parameters during model training to balance the impact of cross entropy loss and Dice loss on the model, so that the model can achieve better results in classification accuracy and boundary delineation accuracy. In actual training, such calculations and parameter adjustments will be performed on each sample (image) of the entire data set, and the model will be iteratively optimized until the model performance reaches a satisfactory level, so that the rectal cancer target area can be accurately and automatically delineated.

[0040] In another embodiment, a training data set matching the target area label information to be trained is processed to generate channel information of a multimodal CT image, wherein the channel information of the multimodal CT image includes feature information of the same CT image at different channel levels, and the channel information of the multimodal CT image is processed based on preset delineated feature points to generate plain scan CT image matching feature points and enhanced CT image matching feature points. The plain scan CT image matching feature points and enhanced CT image matching feature points are processed to generate an affine transformation matrix, and the affine transformation matrix is ​​processed to generate an inverse mapping and target interpolation.

[0041] The method includes a calculation formula for calculating the target interpolation value, which is:

[0042]

[0043] in, is the value corresponding to the adjacent feature points, Represent the weights for the x, y, and z directions respectively.

[0044] Suppose we have a rectal cancer training dataset containing 10 users, each of whom has plain CT and enhanced CT (venous phase, arterial phase) images. For one of the user's CT images, the channel information of its multimodal CT image can be understood as follows: the plain CT image itself can be regarded as a channel, which contains the basic anatomical structure information of the rectum and surrounding tissues when no contrast agent is injected, such as the thickness and density of the rectal wall and the morphology of surrounding organs. The enhanced CT venous phase image is another channel, which shows the distribution of the contrast agent in the tissue during the venous phase. Tumor tissue may have a specific enhancement pattern during this period, which helps to more clearly identify the tumor range. The enhanced CT arterial phase image is the third channel, which can provide information on the tumor blood supply in the early arterial phase. Combined with the venous phase image, it can provide a more comprehensive understanding of the vascular characteristics of the tumor. These feature information at different channel levels together constitute the channel information of the multimodal CT image, providing a rich data foundation for subsequent processing and model training.

[0045] Based on the preset delineation feature points (such as the center point and edge point of the tumor), the images of the above-mentioned users are processed. Assuming that the preset delineation feature point is the center point of the tumor, in the plain scan CT image, the point corresponding to the center point in spatial position is found through the image recognition algorithm. This point is the plain scan CT image matching feature point. Similarly, in the enhanced CT venous phase and arterial phase images, points corresponding to the preset delineation feature points are also found, that is, the enhanced CT image matching feature points. For example, in the plain scan CT image, the pixel coordinates corresponding to the center point of the tumor are, the corresponding coordinates in the enhanced CT venous phase image are, and the corresponding coordinates in the enhanced CT arterial phase image are. These matching feature points will be used for subsequent affine transformation matrix calculations to establish the spatial correspondence between images of different modalities.

[0046] The preset delineation feature points are key feature points corresponding to the target area labels, which are of great significance for the automatic delineation of rectal cancer tumors. For example, for the rectal tumor target area, its center position can be used as an important preset delineation feature point. This point can roughly determine the center of the tumor in three-dimensional space and provide a basic positioning reference for subsequent delineation. In addition, the most prominent point on the edge of the tumor is also a key feature point, because this may be the most obvious part where the tumor invades the surrounding tissues. Accurately identifying these points helps to determine the approximate range of the tumor. There are also key points at the junction with adjacent organs, such as feature points at the adjacent boundaries of the bladder, prostate (male) or uterus, vagina (female), etc. These points define the boundaries between the target area and adjacent organs, which is crucial to avoid accidental irradiation of adjacent organs during radiotherapy.

[0047] First, in multimodal CT images (plain CT and enhanced CT venous phase, arterial phase), based on these preset outline feature points, through image recognition and analysis algorithms, find their corresponding positions in different modal images. For example, determine the coordinates of the tumor center point in the plain CT image , and then find the corresponding point coordinates of the same anatomical position in the enhanced CT venous phase and arterial phase images Then, the affine transformation matrix is ​​calculated using these feature points matched in different modal images. This matrix can describe the spatial transformation relationship between different modal images, including rotation, translation, scaling, etc. Through the affine transformation matrix, the feature points in different modal images can be accurately matched in spatial position, for example, a feature point in a plain scan CT image can be accurately mapped to the corresponding position in the enhanced CT image, thus realizing multi-modal imaging.

[0048] Based on the calculated affine transformation matrix and target interpolation (calculated by a preset calculation formula using information such as adjacent feature points and weights), the training data set is processed to obtain initial delineation information. The initial delineation information includes organ images located at the location of the target area label. For example, a rough image of the rectal tumor and its surrounding organs (such as male organs such as the bladder and seminal vesicles or female organs such as the uterus and vagina) can be seen. In the initial delineation image, the shape of the rectal tumor (such as round, irregular, etc.) can be presented, and its density has different manifestations in plain CT and enhanced CT images, and the outlines of the surrounding organs are also clearly visible.

[0049] Based on the initial delineation information, it is input into a deep learning model (such as a preset target area automatic delineation model based on nnUNET) for training. During the training process, the model processes the initial delineation information according to the preset loss function (including the cross entropy function and the Dice loss function, etc.) and continuously adjusts the model parameters. The cross entropy function measures the classification difference between the model prediction results and the true label (such as the accurate target area delineation marked by the doctor), prompting the model to accurately distinguish between tumor target areas and non-target areas. The Dice loss function focuses on the prediction errors of the model at the boundary pixels and optimizes the delineation accuracy of the tumor edge. Through continuous iterative training to minimize the loss function value, the model gradually learns the accurate tumor delineation mode, so that it can accurately automatically delineate the tumor for new user images according to the preset delineation feature points and related processing procedures, output more accurate target area delineation results, and assist doctors in radiotherapy planning.

[0050] The affine transformation matrix is ​​calculated using the plain scan CT image matching feature points and enhanced CT image matching feature points obtained above. Suppose we select three non-collinear matching feature points in the plain scan CT image: And the corresponding three matching feature points in the enhanced CT image (taking the venous phase as an example) . Through a series of mathematical calculations (involving matrix operations in linear algebra), an affine transformation matrix A can be obtained. This matrix can describe the transformation relationship from the plain CT image space to the enhanced CT image space (taking the venous phase as an example), including rotation, translation, scaling and other operations. The same method can be used to calculate the affine transformation matrix between the plain CT image and the enhanced CT arterial phase image. These matrices will be used for subsequent inverse mapping and target interpolation calculations to achieve accurate registration and information fusion between different modality images.

[0051] For the inverse mapping, assuming that we have obtained the affine transformation matrix from the plain scan CT image to the enhanced CT venous phase image, if there is a point in the enhanced CT venous phase image , through inverse mapping calculation (using mathematical operations such as matrix inversion), we can get the corresponding point of this point in the plain scan CT image , which helps to accurately map the information in the enhanced CT image back to the plain scan CT image space and realize the integration of multimodal imaging information.

[0052] From the formula of affine transformation To start, we need to calculate the inverse mapping, that is, Find P.

[0053] First rewrite the formula as , then multiply both sides by ,get .

[0054] calculate , for a 3x3 matrix Its inverse matrix in is the determinant of A, yes The algebraic cofactor of yes The cofactor of , that is, the determinant of the matrix after removing the i rows and j columns).

[0055] Finally, Substitution In the formula, we can get the inverse mapping formula, that is, for a given point after affine transformation , we can calculate its corresponding point P in the original space. For example, for the x coordinate: Similarly, the calculation formulas for x and z coordinates can be obtained. In actual calculations, it is necessary to optimize and numerically process matrix operations according to specific image data and calculation accuracy requirements to ensure the accuracy and stability of the calculation results. At the same time, this formula can also be expanded according to actual conditions, such as considering more image dimensions or processing more transformation parameters.

[0056] For target interpolation, suppose we have two adjacent feature points and , their corresponding values ​​(such as CT values) are For the points to be interpolated , first calculate the direction weight , and calculate the y direction weight in the same way and z-direction weight Then according to the target interpolation calculation formula Calculate the interpolation result V. For example, CT value Calculated Substituting into the formula, we get V = 53.2. Target interpolation can estimate the data value of unknown positions between known feature points, making the image data more continuous and complete, helping the model to understand and process image information more accurately, and improving the accuracy of target area delineation.

[0057] S103, preprocessing the training data set that matches the target area label information to be trained to generate a training data set with target feature data.

[0058] In one embodiment, feature extraction is performed on a training data set that matches the target area label information to be trained to generate a feature data set. Assume that our training data set contains multimodal CT image data (plain CT and enhanced CT venous phase, arterial phase) of 50 users with rectal cancer and the corresponding target area label information. For each user's image data, we perform feature extraction through a specific image processing algorithm. For example, the thickness features and texture features (such as grayscale co-occurrence matrix features, etc.) of the rectal wall are extracted from the plain CT image, and the enhancement features (such as enhancement degree, enhancement pattern, etc.) of the tumor are extracted from the enhanced CT image. By combining these extracted features, a feature data set is generated. This feature data set contains rich information and can be used for subsequent model training and analysis.

[0059] Obtain any number of data features in the feature data set. In the generated feature data set, we randomly obtain 10 data features. These data features can be a certain feature value of different users. For example, one data feature may be the average thickness of the rectal wall in the plain scan CT image of user A, and another data feature may be the maximum enhancement value of the tumor in the enhanced CT venous phase image of user B. The purpose of obtaining these different data features is to further analyze the relationship and distribution between them.

[0060] Based on the distance between any number of data features and other number of data features of the same category, adjacent features are generated, where the adjacent features include a preset number of any number of data features. The distance between these 10 data features and other data features of the same category (such as the rectal wall thickness features in the plain scan CT images of other users, etc.) is calculated. The distance here can be measured by a suitable distance measurement method such as the Euclidean distance. Assuming that we set the preset number to 3, based on the distance calculation results, find the 3 other data features closest to each data feature, which constitute the adjacent features. For example, for the data feature of the rectal wall thickness feature in the plain scan CT image of user A, find the rectal wall thickness features in the plain scan CT images of the other two users (user C and user D), which are closest to the feature of user A, and these three features constitute a set of adjacent features.

[0061] Based on the number of data features in the training data set, determine the sampling ratio, and based on the sampling ratio, determine the sampling ratio. Count the total number of data features in the training set, assuming it is 200. Determine the sampling ratio based on the preset rules or algorithms. For example, if we want to select approximately 1 / 5 of the data features for subsequent operations, the sampling ratio is 0.2. Based on this sampling ratio, calculate the sampling ratio. If a simple proportional sampling method is used, the sampling ratio may be 0.2 (that is, 1 out of every 5 data features is selected). However, in actual applications, more complex sampling strategies may be adopted according to the distribution of data and specific needs, such as stratified sampling, and the calculation of the sampling ratio will be different.

[0062] Adjacent features are sampled based on the sampling ratio to generate a preset number of sampling features. Adjacent features are sampled based on the calculated sampling ratio (assuming it is 0.2). Since the preset number of adjacent features we set earlier is 3, data features are selected from each group of adjacent features according to the sampling ratio. For example, for a group of adjacent features (including the rectal wall thickness features of users A, C, and D), sampling at a sampling ratio of 0.2 may select one of the features (assuming that the feature of user A is selected). This sampling operation is performed on all adjacent feature groups, and finally a preset number of sampling features (here is the number determined according to the sampling results, which may be less than or equal to the initially set preset number of adjacent features of 3, assuming that 2 sampling features are finally generated) are generated.

[0063] Based on any data feature and each sampling feature, multiple data groups are generated, wherein each data group includes a preset number of data samples, and at least one data sample includes target feature data, which is used to characterize that the image area matching the target area label information has an abnormality. For any data feature among the 10 data features randomly obtained above (for example, the data feature of the maximum enhancement degree value of the tumor in the enhanced CT venous phase image of user B), combined with each sampling feature (assuming that the two sampling features generated above are from certain features of user E and user F respectively), multiple data groups are generated. Each data group includes a preset number of data samples (assuming that the preset number is 5). In these data groups, at least one data sample includes target feature data. For example, one of the data groups may be [the maximum enhancement degree value of the tumor of user B, the sampling feature value of user E, the sampling feature value of user F, the normal tissue feature value of user G, and the feature value of the abnormal area of ​​user H (the feature value of user H here is the target feature data, because the corresponding image area has an abnormality, such as the tumor invading the surrounding tissue, etc.)]. Generating multiple data sets in this way can be used for subsequent model training or analysis to help the model learn the relationship between data features and abnormal feature patterns related to target area label information, thereby improving the accuracy of automatic delineation of rectal cancer targets.

[0064] S104, processing the preset target area automatic delineation model based on nnUNET based on the training data set with the target feature data and the preset delineation rules to generate a target area automatic delineation model.

[0065] In one embodiment, the preset delineation rules are processed to generate preset clinical prior knowledge information and preset anatomical structure restriction information, wherein the preset clinical prior knowledge information is used to characterize the clinical prior knowledge information of different users, and the preset anatomical structure restriction information is used to characterize the regional information of different organs. For the preset clinical prior knowledge information, assuming that in the treatment of rectal cancer, the clinical prior knowledge information of different users includes whether the user has a family history of cancer, whether other cancer-related treatments (such as chemotherapy, surgery, etc.) have been received before, and the user's basic physical condition (such as whether he suffers from other chronic diseases). For example, user A has a family history of rectal cancer and has undergone intestinal surgery before, and user B has no family history but suffers from diabetes. These different information constitute the preset clinical prior knowledge information, which can reflect the potential risks of different users during radiotherapy and possible responses to treatment. For the preset anatomical structure restriction information, taking the rectum and its surrounding organs as an example, the anatomical position relationship between the rectum and organs such as the bladder, prostate (male) or uterus, vagina (female) is important restriction information. Clarifying the position of the rectum in the pelvic cavity and its boundaries with surrounding organs can avoid mistakenly including normal organs in the target area or missing areas invaded by tumors when outlining the target area. For example, under normal circumstances, there is a certain tissue interval between the rectum and the bladder. When outlining the target area of ​​the rectal tumor, it is necessary to ensure that it does not exceed this interval to prevent damage to the bladder.

[0066] The training dataset with target feature data is divided into a training set for training the preset target area automatic delineation model and a validation set for testing the preset target area automatic delineation model based on a preset ratio. Suppose we have a training dataset with target feature data (such as tumor area features, surrounding tissue abnormality features, etc.) containing 100 users with rectal cancer. Divide it into a training set and a validation set according to a preset ratio (such as the common 80:20 ratio). Then the training set will contain data from 80 users, and the validation set will contain data from 20 users. The purpose of this division is to use most of the data (training set) to train the preset target area automatic delineation model so that it can learn the features and rules in the data, and then use a smaller but independent part of the data (validation set) to test the performance of the model and evaluate the accuracy and generalization ability of the model on unseen data.

[0067] Extract multiple data groups from the training set, wherein each data group contains a preset number of data samples, wherein at least one data sample includes target feature data. Extract data groups from the training set, assuming that each data group contains 5 data samples, and at least one data sample includes target feature data. For example, for one of the data groups, data sample 1 may be the texture feature of the tumor area in the plain scan CT image of user C (this is the target feature data), data sample 2 is the density feature of the tissue around the tumor in the enhanced CT image of user D, data sample 3 is the clinical prior knowledge information of user E (such as chemotherapy before), data sample 4 is the thickness feature of the normal rectal wall of user J, and data sample 5 is the enhancement feature of the tumor in the enhanced CT image of user G (this is also the target feature data). Extracting multiple data groups in this combination can comprehensively reflect the image features, clinical information and other aspects of different users in the training set, and provide rich input data for subsequent model training.

[0068] Based on the preset clinical prior knowledge information and the preset anatomical structure restriction information, multiple data sets are processed to generate a target data set. Based on the preset clinical prior knowledge information (such as the user's medical history, treatment history, etc.) and the preset anatomical structure restriction information (such as the boundary relationship of organs) generated previously, multiple data sets are processed. For example, if the preset clinical prior knowledge information indicates that the user has a history of intestinal surgery, it may affect the anatomical structure and tissue characteristics of the intestine. When processing the data set, the relevant data samples (such as image feature data) will be adjusted or weighted according to this information. For the preset anatomical structure restriction information, if the boundary relationship between the rectum and the bladder is known, when the image feature data near the bladder area is involved in the data set, the data will be processed according to this boundary restriction to ensure that the data conforms to the actual anatomical situation. After such processing, a target data set that better meets the actual clinical and anatomical requirements is generated, so that the model can learn more accurate and useful information.

[0069] The preset target area automatic delineation model is trained based on the target data set to generate a trained target area automatic delineation model. The trained target area automatic delineation model is processed based on the validation set to generate a test result. If the data sample containing the target feature data in the test result is a data sample representing that the image area matching the target area label information has an abnormality, the trained target area automatic delineation model is used as the target target area automatic delineation model. The preset target area automatic delineation model (for example, a model based on nnUNET) is trained using the target data set. During the training process, the model learns the feature patterns in the data set and adjusts the internal parameters to improve the accuracy of the target area delineation. After multiple iterations of training, the trained target area automatic delineation model is obtained. Then the data in the validation set is input into the trained model for processing to generate a test result. Assuming that there is image data of user Xin in the validation set, after the model predicts the target area delineation, the test result obtained may be a delineation contour and related probability values ​​and other information. If in this test result, it is found that the image area that matches the target area label information is abnormal (such as inaccurate tumor delineation range, missing some tumor invasion areas, etc.), it is necessary to further analyze the cause, which may be insufficient model training, inaccurate data group processing, etc. If there is no such problem in the test result, and the model performs well in other evaluation indicators (such as Dice coefficient, Hausdorff distance, etc.), then this trained target area automatic delineation model can be used as the target area automatic delineation model for the actual rectal cancer radiotherapy target area automatic delineation task, to assist doctors in more accurate radiotherapy treatment.

[0070] S105: Process the target user information to generate multimodal CT image information of the target user and clinical prior knowledge information matching the target user.

[0071] In one embodiment, the target user information is processed to generate the user's test report information, wherein the test report information includes the area where the user's rectal cancer primary lesion is located, the coverage area of ​​the primary lesion, the target area delineation influencing parameters of the test center, and the user's physical examination information. Assume that the target user is user Li. For the area where the user's rectal cancer primary lesion is located, after a series of examinations (such as colonoscopy, rectal digital examination, imaging examination, etc.), it is determined that the primary lesion is located in the middle rectum. The coverage area of ​​the primary lesion is evaluated by imaging examinations such as CT and MRI. Assuming that the tumor involves part of the rectal wall but does not completely penetrate, its coverage area is mainly concentrated in a side wall and part of the posterior wall of the middle rectum. In terms of the target area delineation influencing parameters of the test center, the test center is accustomed to using the dose irradiation range as a reference for the target area delineation range when performing radiotherapy, and based on past experience, for tumors located in the middle rectum, a certain range of delineation areas will be appropriately expanded to ensure sufficient irradiation doses. The user's physical examination information includes basic indicators such as height, weight, blood routine, liver and kidney function, etc. For example, if the height is 170cm, weight is 65kg, the blood routine indicators are basically normal, and there is no obvious abnormality in liver and kidney function. User Li did not undergo surgery before radiotherapy, but received a course of chemotherapy. This information will be included in the test report information.

[0072] Generate clinical prior knowledge information that matches the target user based on the user's test report information. Generate clinical prior knowledge information that matches Mr. Li based on the above test report information. Since the user's primary lesion is in the middle rectum and no surgery has been performed, this means that the morphology and position of the tumor may be relatively complex, and the target area needs to be located more accurately during radiotherapy to avoid damage to surrounding normal tissues. The user has undergone chemotherapy, which may have a certain impact on the morphology and blood supply of the tumor tissue. Changes after chemotherapy need to be considered when delineating the target area. The testing center's habit of using the dose irradiation range as a reference will also affect the formation of clinical prior knowledge. In subsequent model processing, the strategy for delineating Mr. Li's target area will be adjusted according to this feature to adapt to the treatment specifications of the testing center.

[0073] The target user information is subjected to image extraction and processing to generate the target user's plain CT image, enhanced CT venous phase image, and enhanced CT arterial phase image. The target user information of Li is subjected to image extraction and processing. During the plain CT examination, a series of plain CT images are obtained, which clearly show the anatomical structure of the rectum and surrounding tissues, such as the thickness of the rectal wall and the distribution of surrounding adipose tissue. In the enhanced CT venous phase examination, the enhanced CT venous phase image is obtained by scanning at a specific time point after the injection of contrast agent. At this time, it can be seen that the tumor tissue has different degrees of enhancement relative to the surrounding normal tissue, which helps to more accurately identify the scope and boundary of the tumor. The enhanced CT arterial phase image can reflect the blood supply of the tumor in the early arterial stage, which is helpful for judging the activity and invasiveness of the tumor. For example, in the enhanced CT arterial phase image, it is found that some areas of the tumor have obvious early enhancement, indicating that the area is rich in blood supply and may be the site of active tumor growth.

[0074] The multimodal CT impact information of the target user is generated based on the target user's plain CT image, enhanced CT venous phase image, and enhanced CT arterial phase image. Li's plain CT image, enhanced CT venous phase image, and enhanced CT arterial phase image are combined to generate his multimodal CT image information. In this multimodal CT image information, doctors and models can comprehensively analyze the characteristics of images of different modalities. For example, plain CT images provide a basic anatomical structure framework, enhanced CT venous phase images further clarify the approximate range of the tumor (by enhancing regional judgment), and enhanced CT arterial phase images supplement the details of the tumor blood supply. Through the comprehensive interpretation of these multimodal CT image information, we can have a more comprehensive understanding of Li's rectal tumor and provide a more accurate basis for subsequent automatic target area delineation and radiotherapy planning. For example, in multimodal CT imaging information, it can be seen that the tumor appears as an isodense shadow on plain CT, has uneven enhancement in the venous phase of enhanced CT, and obvious early enhancement in the arterial phase of enhanced CT. The relationship with the surrounding tissues (such as the relative position with the bladder, seminal vesicles, etc.) can be clearly presented in different modality images. This information is crucial for accurately delineating the target area.

[0075] S106: Process the multimodal CT image information of the target user, the target area label information, and the clinical prior knowledge information matching the target user based on the target area automatic delineation model to generate target area delineation information.

[0076] In one embodiment, the target target area label information is processed based on the target target area automatic delineation model to generate the area information to be delineated. Assume that the target target area label information specifies that the area to be delineated is a rectal tumor and the area that may be invaded. The target target area automatic delineation model will process these label information according to the knowledge and algorithm it has learned. For example, the model determines the approximate area range where the tumor may exist in the new target user image by analyzing the image features corresponding to similar target area labels in a large number of previous cases. If the model learns that the area with a certain specific density and morphological changes in the middle of the rectum may be the location of the tumor, then in the image of the target user, once a similar feature area is found, it is marked as the area information to be delineated. For example, in the plain scan CT image of the target user, it is found that the density of the local area in the middle of the rectum is slightly higher than the surrounding normal tissue, and the morphology is irregular. The model preliminarily determines this area as part of the area to be delineated, and combines the enhanced features of the area in the enhanced CT venous phase and arterial phase images. Information such as further refines the area information to be delineated so that it more accurately covers the tumor and its possible invasion range.

[0077] The automatic delineation model based on the target area processes the clinical prior knowledge information that matches the target user and generates delineation restriction information. For the clinical prior knowledge information that matches the target user, such as if the user has received chemotherapy before, the model will generate delineation restriction information based on this information. Because chemotherapy may cause changes such as necrosis and fibrosis of tumor tissue, affecting its performance on the image, the model will take into account the blurring of boundaries caused by these changes during delineation, and adjust the delineation strategy to avoid excessive or insufficient delineation. If the user's clinical prior knowledge information shows that the primary lesion is located in the upper rectum and close to the bladder, the model will generate corresponding delineation restriction information based on the anatomical relationship between the bladder and the upper rectum and the target area delineation habits of the testing center (such as avoiding damage to important organs such as the bladder). For example, it is stipulated that when delineating the target area, the boundary close to the bladder side must be strictly delineated according to a certain safety distance, and cannot exceed the boundary between normal tissue and the bladder to ensure the safety of the bladder during radiotherapy.

[0078] The target target area automatic delineation model processes the target user's multimodal CT image information to generate the target user's multimodal CT image channel information. The target target area automatic delineation model processes the target user's multimodal CT image information (including plain CT, enhanced CT venous phase and arterial phase images). The plain CT image itself is a channel, and the information it contains, such as the thickness and density of the rectal wall and the basic anatomical structure of the surrounding tissue, constitutes the characteristic information of the channel. The enhanced CT venous phase image is another channel, and the distribution of the contrast agent in the tissue during the venous phase, such as the degree of enhancement and enhancement pattern of the tumor tissue, becomes the unique characteristic information of the channel. The enhanced CT arterial phase image provides information on the tumor blood supply in the early arterial phase, such as the early enhancement characteristics of the tumor, the display of the blood supply vessels, etc., forming the characteristic information of the third channel. The characteristic information of these different channels is combined to generate the channel information of the target user's multimodal CT image. The model can fully understand the situation of the tumor and the surrounding tissue by analyzing these channel information.

[0079] Based on the information of the area to be delineated, the channel information of the target user's multimodal CT image is processed to generate initial target area delineation information. For example, in the area to be delineated, the model will focus on the characteristic performance of the area in different channel information. If the area to be delineated appears as an abnormal density area in the plain scan CT channel and presents a specific enhancement pattern in the enhanced CT venous phase channel, the model will preliminarily outline a rough target area contour on the image based on these features to form initial target area delineation information. This initial delineation may be a relatively rough outline that includes the main suspicious parts in the area to be delineated, but it may need further optimization. For example, in the initial target area delineation information, the tumor and the surrounding tissues within a certain range may be included, but the boundaries may not be accurate enough and need to be adjusted later based on more information.

[0080] The initial target area delineation information is processed based on the delineation restriction information to generate the target area delineation information. For example, based on the aforementioned delineation restriction information to avoid damaging the bladder, the boundary close to the bladder side in the initial target area delineation information is precisely adjusted to ensure that it does not exceed the safety range. If clinical prior knowledge suggests that the tumor may have shrinking boundaries after chemotherapy, then based on the initial target area delineation information, the target area range is appropriately narrowed to remove false positive areas that may be caused by chemotherapy. After such a series of processing, accurate target area delineation information is finally generated. This information will provide doctors with an accurate target area range for formulating radiotherapy plans, improving the accuracy and effectiveness of radiotherapy, while minimizing damage to surrounding normal tissues.

[0081] In this application, the server obtains target user information, target target area label information, target area label information to be trained, training data set and preset target area automatic delineation model based on nnUNET, and the training data set contains multimodal CT images of other users and clinical prior knowledge. Based on the target area label information to be trained, the training data set is processed to generate preset delineation rules, including generating preset delineation feature points, and then processing the inverse mapping and target interpolation to obtain the initial delineation information, and finally generating rules based on the preset loss function. Then the training data set is preprocessed, and a training data set with target feature data is generated through the steps of feature extraction, data feature analysis, adjacent feature generation, sampling ratio determination, sampling feature generation and data group construction. Then the preset target area automatic delineation model is trained using the data set and the preset delineation rules, including generating clinical prior knowledge and anatomical structure restriction information, dividing the training set and the verification set, extracting the data set and processing the target data set for training, and finally determining the target target area automatic delineation model based on the verification set test.

[0082] For the target user, its information is processed to generate multimodal CT image information and clinical prior knowledge information, including generating test report information, generating clinical prior knowledge based on it, and extracting plain scan and enhanced CT images to generate multimodal CT image information. Finally, based on the target area automatic delineation model, the target area label information generates the area to be delineated information, the clinical prior knowledge information generates delineation restriction information, the multimodal CT image information generates channel information, the initial target area delineation information is generated based on the area to be delineated information and channel information, and then the target area delineation information is generated in combination with the delineation restriction information to assist clinical precision radiotherapy.

[0083] In one embodiment, if Figure 2 As shown, the present application also provides an automatic delineation device for rectal cancer radiotherapy target area, comprising: An acquisition module 201 is used to acquire target user information, target area label information, target area label information to be trained, a training data set matching the target area label information to be trained, and a preset target area automatic delineation model based on nnUNET, wherein the training data set includes multimodal CT image information of other users and clinical prior knowledge information matching with other users, and the multimodal CT images include plain scan CT images and enhanced CT images; The processing module 202 is used to process the training data set matching the target area label information to be trained based on the target area label information to be trained, and generate a preset delineation rule, wherein the preset delineation rule is used to characterize the delineation standards of different plain scan CT images, different enhanced CT images and the anatomical structure restriction information of the target area label adjacent organs; pre-process the training data set matching the target area label information to be trained to generate a training data set with target feature data; process the preset target area automatic delineation model based on nnUNET based on the training data set with target feature data and the preset delineation rule to generate a target target area automatic delineation model; process the target user information to generate the target user's multimodal CT image information and the clinical prior knowledge information matching the target user; process the target user's multimodal CT image information, the target target area label information and the clinical prior knowledge information matching the target user based on the target target area automatic delineation model to generate the target target area delineation information.

[0084] The technical solution of the embodiment of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for enabling an electronic device (which may be an air conditioner, a refrigeration device, a personal computer, a server, or a network device, etc.) or a processor to execute all or part of the steps of the method described in the embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a ROM, a RAM, a magnetic disk, or an optical disk.

[0085] The computer program product provided in the above-mentioned embodiments of the present application and the method for automatically delineating the target area for radiotherapy of rectal cancer provided in the embodiments of the present application are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented by the application programs stored therein.

[0086] Each embodiment in the present application is described in a related manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the automatic delineation method, electronic device, electronic device, and readable storage medium embodiment for evaluating the target area for radiotherapy of rectal cancer, since it is basically similar to the above-mentioned automatic delineation method embodiment for radiotherapy of rectal cancer, the description is relatively simple, and the relevant parts can be referred to the partial description of the above-mentioned automatic delineation method embodiment for radiotherapy of rectal cancer.

Claims

1. A method for automatically delineating a target area for radiotherapy of rectal cancer, characterized in that: include: Obtaining target user information, target target area label information, target area label information to be trained, a training data set matching the target area label information to be trained, and a preset target area automatic delineation model based on nnUNET, wherein the training data set includes multimodal CT image information of other users and clinical prior knowledge information matching with other users, and the multimodal CT images include plain scan CT images and enhanced CT images; Based on the target area label information to be trained, a training data set matching the target area label information to be trained is processed to generate a preset delineation rule, wherein the preset delineation rule is used to characterize delineation standards of different plain scan CT images and different enhanced CT images and anatomical structure restriction information of organs adjacent to the target area label; Preprocessing a training data set that matches the target area label information to be trained to generate a training data set with target feature data; Processing the preset target area automatic delineation model based on nnUNET based on the training data set with target feature data and the preset delineation rule to generate a target area automatic delineation model; Processing the target user information to generate multimodal CT image information of the target user and clinical prior knowledge information matching the target user; The multimodal CT image information of the target user, the target target area label information and the clinical prior knowledge information matching the target user are processed based on the target area automatic delineation model to generate target area delineation information.

2. The method according to claim 1, characterized in that Based on the target area label information to be trained, a training data set matching the target area label information to be trained is processed to generate a preset delineation rule, including: Processing the target area label information to be trained to generate preset delineation feature points, wherein the preset delineation feature points are used to represent the delineation feature points corresponding to the target area label; Processing the preset delineated feature points and a training data set matching the target area label information to be trained based on preset processing rules to generate an inverse mapping and target interpolation; Processing a training data set matching the target region label information to be trained based on the inverse mapping and the target interpolation to initially outline information, wherein the initial outline information includes an organ image located at a position where the target region label is located; Processing the initial delineation information based on a preset loss function to generate a preset delineation rule; The method includes a calculation formula for obtaining a target interpolation value, and the calculation formula is: ； in, is the weight of the cross entropy loss function, is the weight of the Dice loss function; ； Among them, CE represents the cross entropy loss function, y represents the true label, and p represents the predicted probability of the model; ； Among them, Dice represents the loss function, y represents the true label, and p represents the predicted probability of the model. is a very small constant to avoid the denominator being zero.

3. The method according to claim 2, characterized in that The preset delineated feature points and the training data set matching the target area label information to be trained are processed based on preset processing rules to generate inverse mapping and target interpolation, including: Processing a training data set that matches the target area label information to be trained to generate channel information of a multimodal CT image, wherein the channel information of the multimodal CT image includes feature information of the same CT image at different channel levels; Processing the channel information of the multimodal CT image based on the preset delineated feature points to generate plain scan CT image matching feature points and enhanced CT image matching feature points; Processing the plain scan CT image matching feature points and the enhanced CT image matching feature points to generate an affine transformation matrix; Processing the affine transformation matrix to generate an inverse mapping and a target interpolation; The method includes a calculation formula for calculating a target interpolation value, and the calculation formula is: ； ； ； ； in, and is the value corresponding to the adjacent feature points, Represent the weights for the x, y, and z directions respectively.

4. The method according to claim 1, preprocessing the training data set matching the target area label information to be trained to generate a training data set with target feature data, comprising: Extracting features from a training data set that matches the target area label information to be trained to generate a feature data set; Obtaining any number of data features in the feature data set; Generate adjacent features based on the distance between the any number of data features and other numbers of data features of the same category, wherein the adjacent features include a preset number of the any number of data features; Determine the sampling ratio based on the number of each data feature in the training data set; Based on the sampling ratio, determining a sampling ratio; Sampling the adjacent features based on the sampling ratio to generate a preset number of sampling features; Based on any data feature and each sampling feature, multiple data groups are generated, wherein each data group includes a preset number of data samples, and at least one data sample includes target feature data, and the target feature data is used to characterize the presence of an abnormality in an image area that matches the target area label information.

5. The method according to claim 4, characterized in that The preset target area automatic delineation model based on nnUNET is processed based on the training data set with target feature data and the preset delineation rule to generate the target area automatic delineation model, including: Processing the preset delineation rules to generate preset clinical prior knowledge information and preset anatomical structure restriction information, wherein the preset clinical prior knowledge information is used to characterize clinical prior knowledge information of different users, and the preset anatomical structure restriction information is used to characterize regional information of different organs; Dividing the training data set with the target feature data into a training set for training the preset target area automatic delineation model and a validation set for testing the preset target area automatic delineation model based on a preset ratio; Extracting a plurality of data groups from the training set, wherein each data group comprises a preset number of data samples, wherein at least one data sample comprises target feature data; Processing multiple data sets based on the preset clinical prior knowledge information and the preset anatomical structure restriction information to generate a target data set; Training the preset target area automatic delineation model based on the target data group to generate a trained target area automatic delineation model; Processing the trained target area automatic delineation model based on the validation set to generate a test result; If the data sample containing the target feature data in the test result indicates that an abnormality exists in the image region matching the target region label information, the trained target region automatic delineation model is used as the target region automatic delineation model.

6. The method according to claim 5, characterized in that The target user information is processed to generate multimodal CT image information of the target user and clinical prior knowledge information matching the target user, including: Processing the target user information to generate the user's test report information, wherein the test report information includes the area where the user's rectal cancer primary lesion is located, the coverage area of ​​the primary lesion, the target area delineation influencing parameters of the test center, and the user's physical examination information; Generate clinical prior knowledge information matching the target user based on the test report information of the user; Performing image extraction processing on the target user information to generate a plain scan CT image, an enhanced CT venous phase image, and an enhanced CT arterial phase image of the target user; Multimodal CT impact information of the target user is generated based on the plain scan CT image of the target user, the enhanced CT venous phase image, and the enhanced CT arterial phase image.

7. The method according to claim 6, processing the multimodal CT image information of the target user, the target area label information and the clinical prior knowledge information matching the target user based on the target area automatic delineation model to generate target area delineation information, comprising: Processing the target area label information based on the target area automatic delineation model to generate information of the area to be delineated; Processing the clinical prior knowledge information matching the target user based on the target area automatic delineation model to generate delineation restriction information; Processing the multimodal CT image information of the target user based on the target area automatic delineation model to generate channel information of the multimodal CT image of the target user; Processing the channel information of the multimodal CT image of the target user based on the information of the area to be delineated to generate initial target area delineation information; The initial target area delineation information is processed based on the delineation restriction information to generate target area delineation information.

8. An automatic delineation device for rectal cancer radiotherapy target area, characterized in that: The device comprises: An acquisition module, used to acquire target user information, target area label information, target area label information to be trained, a training data set matching the target area label information to be trained, and a preset target area automatic delineation model based on nnUNET, wherein the training data set includes multimodal CT image information of other users and clinical prior knowledge information matching with other users, and the multimodal CT images include plain scan CT images and enhanced CT images; A processing module is used to process a training data set that matches the target area label information to be trained based on the target area label information to be trained, and generate a preset delineation rule, wherein the preset delineation rule is used to characterize the delineation standards of different plain scan CT images, different enhanced CT images and the anatomical structure restriction information of organs adjacent to the target area label; pre-process the training data set that matches the target area label information to be trained to generate a training data set with target feature data; based on the training data set with target feature data and the preset delineation rule, process the preset target area automatic delineation model based on nnUNET to generate a target target area automatic delineation model; process the target user information to generate multimodal CT image information of the target user and clinical prior knowledge information matching the target user; based on the target target area automatic delineation model, process the multimodal CT image information of the target user, the target target area label information and the clinical prior knowledge information matching the target user to generate target target area delineation information.

9. An electronic device, characterized in that: include: a first processor; and a memory for storing executable instructions of the first processor; Wherein, the first processor is configured to execute the method for automatically delineating a target area for radiotherapy of rectal cancer according to any one of claims 1 to 7 by executing the executable instructions.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the second processor, the method for automatically delineating a target area for radiotherapy of rectal cancer as claimed in any one of claims 1 to 7 is implemented.

Images