Medical image processing method, computer equipment and storage medium

By extracting and fusing the features in MRI imaging and treatment plan data, machine learning algorithms are used to evaluate whether patients with nasopharyngeal carcinoma need adaptive radiotherapy, solving the problem of difficult to accurately distinguish patients with ART in the prior art, and improving the efficacy of radiotherapy and efficiency.

CN119006382APending Publication Date: 2024-11-22SUN YAT SEN UNIVERSITY CANCER CENTER (CANCER HOSPITAL AFFILIATED TO SUN YAT SEN UNIVERSITY CANCER RESEARCH INSTITUTE OF SUN YAT SEN UNIVERSITY)

Patent Information

Application Number
CN202411022059.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-29
Publication Date
2024-11-22

AI Technical Summary

Technical Problem

The prior art is difficult to accurately distinguish patients with nasopharyngeal carcinoma who need to perform adaptive radiation therapy, which affects the efficacy of radiotherapy and is expensive to perform ART.

Method used

By obtaining MRI imaging data and treatment plan data, the target imaging and target dose-omic features are extracted, and the target imaging and dosage-omic features are performed, and the prediction evaluation is carried out based on the machine learning algorithm to determine whether an adaptive radiotherapy regimen is needed.

Benefits of technology

It has achieved rapid and accurate assessment of whether the subjects need adaptive radiotherapy, helping medical workers accurately distinguish the situations where ART is required, and improving the efficacy and efficiency of radiotherapy for nasopharyngeal carcinoma.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119006382A_ABST
    Figure CN119006382A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of image processing, and discloses a medical image processing method, computer equipment and a storage medium. The method comprises the following steps: firstly, acquiring MRI image data and treatment plan data of a subject; wherein the MRI image data is obtained by performing MRI examination on the nasopharynx part of the tested object; the treatment plan data comprises three-dimensional dose distribution data; determining target radiomics characteristics of the MRI image data and target dosimics characteristics of the three-dimensional dose distribution data; fusing the target image omics features and the target dose omics features to obtain target omics features; and predicting and evaluating the adaptive radiotherapy scheme of the tested object based on the target omics features to obtain scheme evaluation data. Therefore, an auxiliary means is realized to quickly, accurately and effectively predict and evaluate whether the self-adaptive radiotherapy scheme needs to be carried out on the tested object, so that medical workers are helped to accurately distinguish the condition that self-adaptive radiotherapy of nasopharynx cancer needs to be carried out.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of image processing technology, and in particular to a medical image processing method, a computer device and a storage medium. Background Art

[0002] Nasopharyngeal carcinoma is an epithelial malignant tumor originating from the nasopharyngeal mucosal area. Because the location of nasopharyngeal carcinoma is somewhat hidden, often adjacent to important organs and structures, and highly sensitive to radiation, radiotherapy (abbreviated as "radiotherapy") is often used to prevent and treat nasopharyngeal carcinoma in clinical practice. Doctors outline the tumor target area and surrounding important organs at risk based on the patient's pre-treatment CT images, and tailor-make the radiotherapy plan. However, during the entire radiotherapy process, nasopharyngeal carcinoma patients may experience deviations in the radiation dose received by the tumor target area, or the radiation dose received by important organs at risk may exceed the limit, thereby affecting the patient's radiotherapy efficacy. Adaptive radiotherapy (ART) can solve the above problems well. However, because not every patient needs to undergo ART, although ART can bring great advantages to patients in theory, it is actually difficult to implement in clinical applications.

[0003] Therefore, in order to identify NPC patients who need ART, related technologies use image processing and machine learning to mine deep quantitative features from images, arrange and combine the extracted features, and build prediction models by grouping for verification and testing, thereby revealing the heterogeneity of tumors to assist in diagnosis, radiotherapy sensitivity, prognosis and efficacy evaluation, etc. However, the accuracy of this process in predicting the necessity of ART for NPC needs to be improved. Summary of the invention

[0004] The present application aims to solve at least one of the technical problems in the related art to a certain extent. To this end, the present application proposes a medical image processing method, a computer device and a storage medium. The main technical solutions adopted by the present application include:

[0005] In a first aspect, an embodiment of the present application provides a medical image processing method, the method comprising: obtaining MRI image data and treatment plan data of a subject; wherein the MRI image data is obtained by performing an MRI examination on the nasopharyngeal area of ​​the subject; the treatment plan data comprises three-dimensional dose distribution data; determining target imaging omics features of the MRI image data and target dose omics features of the three-dimensional dose distribution data; fusing the target imaging omics features and the target dose omics features to obtain target omics features; predicting and evaluating the adaptive radiotherapy plan of the subject based on the target omics features to obtain plan evaluation data.

[0006] In a second aspect, an embodiment of the present application provides a medical image processing device, comprising:

[0007] A data acquisition module is used to acquire MRI image data and treatment plan data of the subject; wherein the MRI image data is obtained by performing an MRI examination on the nasopharyngeal area of ​​the subject; the treatment plan data includes three-dimensional dose distribution data; a feature determination module is used to determine the target imaging omics features of the MRI image data and the target dose omics features of the three-dimensional dose distribution data; a feature fusion module is used to fuse the target imaging omics features and the target dose omics features to obtain the target omics features; a prediction and evaluation module is used to predict and evaluate the adaptive radiotherapy plan of the subject based on the target omics features to obtain plan evaluation data.

[0008] The present application also provides a computer device including a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any of the above methods when executing the computer program.

[0009] The present application also provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of any of the above methods are implemented.

[0010] In the above embodiment, a medical image processing method is provided, and the MRI image data and treatment plan data of the subject are first obtained, wherein the treatment plan data includes three-dimensional dose distribution data. Feature extraction is performed based on the MRI image data to obtain the target imaging omics features, and feature extraction is performed based on the three-dimensional dose distribution data to obtain the target dose omics features. The extracted target imaging omics features and target dose omics features are merged and fused to obtain the fused omics features. Finally, the adaptive radiotherapy regimen of the subject is predicted and evaluated based on the fused omics features, and the regimen evaluation data is obtained using the prediction and evaluation model constructed based on the machine learning algorithm. At this point, an auxiliary means is realized to quickly, accurately and effectively predict and evaluate whether the subject needs to carry out an adaptive radiotherapy regimen, thereby helping medical workers to accurately distinguish situations where it is necessary to carry out adaptive radiotherapy for nasopharyngeal carcinoma. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the specific implementation methods of the present application or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0012] Figure 1 A flowchart of a medical image processing method provided according to an embodiment of the present application;

[0013] Figure 2 A flowchart of a medical image processing method provided according to another embodiment of the present application;

[0014] Figure 3 A flowchart of a method for determining target imaging omics features and target dose omics features according to one embodiment of the present application;

[0015] Figure 4 A flow chart of a method for determining target imaging omics features and target dose omics features according to another embodiment of the present application;

[0016] Figure 5 A flowchart of a screening process of a feature set to be screened according to an embodiment of the present application;

[0017] Figure 6 is a flowchart of a medical image processing method according to yet another embodiment of the present application;

[0018] Figure 7 is a structural block diagram of a medical image processing device according to an embodiment of the present application;

[0019] Figure 8 The figure is a diagram of the internal structure of a computer device according to one embodiment of the present application. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without making creative work are within the scope of protection of this application.

[0021] Nasopharyngeal carcinoma is an epithelial malignant tumor that originates from the nasopharyngeal mucosa and is relatively common in head and neck tumors. Since the location of nasopharyngeal carcinoma is somewhat hidden, often adjacent to important organs and structures, and highly sensitive to radiation, radiotherapy (abbreviated as "radiotherapy") has become the most important treatment for the prevention and treatment of nasopharyngeal carcinoma. In particular, the rise of high-precision technologies such as intensity-modulated radiotherapy and image-guided radiotherapy in recent years has further improved the efficacy of radiotherapy for nasopharyngeal carcinoma and reduced side effects. Currently, the 5-year survival rate of early nasopharyngeal carcinoma after radiotherapy can be as high as 85%, and the local control rate can even be as high as 90%.

[0022] In the clinical practice of radiotherapy for nasopharyngeal carcinoma, the patient's treatment plan is tailored according to the anatomical structure information before treatment. The doctor outlines the tumor target area and the surrounding important organs at risk based on the patient's pre-treatment CT images. The physicist then designs the intensity-modulated radiotherapy plan based on this to maximize the dose coverage of the tumor target area while protecting normal tissue as much as possible. Considering the inconsistent sensitivity of the tumor target area and normal tissue to radiation, a multi-fraction treatment model is required when implementing the radiotherapy plan. For example, nasopharyngeal carcinoma usually has a prescribed dose of about 66-70Gy, which needs to be performed in 30-33 times according to the conventional fractionation mode, usually once a day, and 5 times a week. This can gradually increase the cumulative damage of radiation to tumor cells and increase the sensitivity of tumors to radiation. At the same time, it can allow normal tissue cells to have sufficient time to repair and reduce the chance of side effects, thereby improving the overall radiotherapy effect of nasopharyngeal carcinoma.

[0023] However, during the entire radiotherapy process, NPC patients may experience anatomical changes such as weight loss, shrinkage of the tumor target area, and changes in the edema area, which may cause the actual radiation dose received by the tumor target area to deviate from the planned dose, or the radiation dose of important organs at risk may exceed the limit, thus affecting the patient's radiotherapy efficacy. Adaptive radiotherapy (ART) can solve the above problems well and compensate for the dose deviation that may occur during the treatment process. Adaptive radiotherapy ART uses medical images re-acquired during the treatment process (such as re-scanned CT, CBCT, or magnetic resonance imaging MRI, etc.) to re-delineate and locate the tumor target area and organs at risk, and then redesign and optimize the new treatment plan to ensure that the patient receives accurate radiotherapy. Although ART can bring great advantages to patients in theory, it is actually difficult to implement in clinical applications. The reason is that the human cost, time cost, and economic cost of implementing ART are too high, especially for those radiotherapy centers that treat a large number of patients every day. Furthermore, although nasopharyngeal carcinoma is sensitive to radiation, the sensitivity of different patients varies and is easily affected by individual differences or tumor staging, so not every patient needs to undergo ART. It is very necessary to accurately identify nasopharyngeal carcinoma patients who need ART.

[0024] In the related technologies, the method used in clinical judgment of whether NPC patients need ART is based on the images before fractionated treatment, such as CT / PET-CT, CBCT, X-ray or EPID images. This method has the following problems to a greater or lesser extent:

[0025] 1) Considering the long radiotherapy cycle for NPC, taking radiographs at each fraction will greatly increase the cost of treatment for patients, while also bringing additional radiation burden and increasing the chance of side effects for patients.

[0026] 2) Traditional X-ray-based imaging has poor resolution for soft tissue imaging, which leads to greater uncertainty in distinguishing the boundaries of NPC lesions.

[0027] 3) Relying solely on imaging judgment is not accurate enough. The response of nasopharyngeal carcinoma tumor lesions to radiation and the probability of side effects in surrounding normal tissues depend largely on the distribution of radiation dose received by the lesion site.

[0028] In fact, in addition to reflecting the anatomical structure of lesions and organs at risk, medical images of patients before treatment also contain a lot of information about diagnosis and prognosis that is difficult to distinguish with the naked eye. This is the concept of radiomics. Through image processing and machine learning technologies, deep quantitative features can be mined from images, thereby effectively revealing the heterogeneity of tumors to assist in diagnosis, radiotherapy sensitivity, prognosis and efficacy evaluation. Therefore, radiomics technology has shown great potential in the differential diagnosis, efficacy evaluation and prognosis prediction of nasopharyngeal carcinoma.

[0029] Compared with human judgment, imaging omics methods can deeply quantify the internal heterogeneity of NPC from multiple dimensions and directions, revealing the tissue proliferation and necrosis in the lesion area, as well as the changes in the internal tumor microenvironment. In related technologies, imaging omics research and application for NPC have involved tumor differential diagnosis, staging, efficacy and prognosis prediction, but few studies have applied it to the prediction of adaptive radiotherapy. Moreover, most imaging omics studies are based on CT or PET-CT images. Considering the poor resolution of CT / PET-CT in the nasopharyngeal soft tissue, coupled with the complexity of the anatomical structure around the nasopharynx, the use of magnetic resonance imaging (MRI) can more clearly distinguish the soft tissue boundaries, thereby more clearly showing the heterogeneity of NPC lesions and the invasion of surrounding tissue structures. The National Comprehensive Cancer Network (NCCN) guidelines of the United States clearly recommend MRI as the preferred imaging examination for NPC. Furthermore, the magnetic resonance scanning process does not involve ionizing radiation and will not bring additional radiation burden to patients. However, unlike CT images, MRI images have a variety of sequences to choose from. Different MRI sequences can highlight different anatomical details and contain different microscopic heterogeneity characteristics, providing richer information for analyzing and evaluating tumor lesion areas.

[0030] In terms of dosimetry, the relevant technologies mainly use three-dimensional dose distribution in clinical practice in combination with the dose volume histogram DVH of the outlined structure to evaluate the exposure of tumor lesions and surrounding organs at risk. This method loses the deep dosimetric features that cannot be observed by the naked eye, such as statistical features, texture features, and filtering features at different dose levels. The three-dimensional dose distribution is a collection of average dose values ​​in discrete voxels in space calculated by the treatment planning system, which can be regarded as an image file. Therefore, by drawing on the imaging omics method, dosimetry came into being. It can quantitatively calculate the high-flux dosimetric characteristics in the three-dimensional dose distribution, reflecting the subtle changes in the three-dimensional spatial dose distribution, thereby capturing additional information related to the patient's adaptive radiotherapy and guiding the formulation of adaptive radiotherapy plans. There are few studies on the application of dosimetry methods to adaptive radiotherapy prediction in the relevant technologies, but relevant studies have shown that the inclusion of dosimetry features can significantly improve the accuracy of radiotherapy efficacy prediction.

[0031] The research on dosimetrics combined with radiomics is mostly used for the prediction of survival after radiotherapy and the prediction of radiation side effects. So far, there has been no research on its use for adaptive radiotherapy prediction. Moreover, in the related technologies, almost all dosimetric features and radiomics features are arranged and combined, and prediction models are constructed by grouping for verification and testing. This method is cumbersome and inefficient. Redundant features may appear, resulting in a bloated model, which affects the performance of the model and makes it impossible to accurately predict the necessity of ART for NPC.

[0032] In summary, according to an embodiment of the present application, a medical image processing method, a computer device, and a storage medium embodiment are provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in an order different from that shown here.

[0033] Figure 1 is a flow chart of a medical image processing method according to an embodiment of the present application, such as Figure 1 As shown, the method includes:

[0034] S110, obtaining MRI image data and treatment plan data of the subject.

[0035] Among them, the MRI image data is obtained by performing MRI examination on the nasopharynx of the subject. The subject may be a patient who needs to undergo predictive evaluation of adaptive radiotherapy for nasopharyngeal carcinoma. The MRI image data is medical image data obtained through magnetic resonance imaging technology.

[0036] The treatment plan data includes three-dimensional dose distribution data, wherein the treatment plan data can be represented by a series of DICOM data reflecting the radiotherapy plan of the subject. Specifically, the treatment plan data can be obtained by obtaining multiple DICOM data reflecting the condition of the subject. Among them, the multiple DICOM data reflecting the condition of the subject include three-dimensional dose distribution data and other data. Among them, the three-dimensional dose distribution data is a collection of average dose values ​​in discrete voxels in three-dimensional space calculated by the treatment planning system, which can be equivalent to a kind of image file data, and is also represented by the DICOM format.

[0037] S120 , determining target imaging features of the MRI image data and target dose features of the three-dimensional dose distribution data.

[0038] Among them, the target radiomics features are a series of quantitative image features, which are features used to describe and characterize the morphology, structure, texture, signal intensity and other information of the tumor at the lesion location. Specifically, the radiomics features come from different imaging technologies, such as structured MRI and functional MRI. Exemplarily, the target radiomics features can be features that characterize the location of nasopharyngeal carcinoma lesions, and the target radiomics features can be determined by extracting features from MRI image data obtained by performing an MRI examination on the nasopharynx of the subject.

[0039] Similarly, the target dose omics feature is a feature that can quantitatively describe and evaluate the radiation therapy dose distribution, describing the distribution pattern of the dose in the entire treatment volume, including the uniformity and gradient of the dose. Specifically, the dose omics feature is derived from the dose distribution in the radiation therapy planning system and is represented in the form of a three-dimensional matrix. Dosimetric features include various types of features, such as dose intensity, dose uniformity, dose gradient, and hot spot areas. Exemplarily, the target dose omics feature can be a feature that characterizes the location of the nasopharyngeal carcinoma lesion and the dose situation of multiple important adjacent organs at risk, and the target dose omics feature that determines the three-dimensional dose distribution data can be obtained from the three-dimensional dose distribution data in multiple DICOM data reflecting the subject's condition.

[0040] S130, fusing the target imaging omics features and the target dose omics features to obtain the target omics features.

[0041] Among them, the target omics feature can be a composite feature that includes a series of target imaging omics features and target dosimetric features. Specifically, by fusing the target imaging omics features and the target dosimetric features, the selected target imaging omics features and the target dosimetric features are spliced ​​and merged into a large feature matrix to obtain the target omics features that include both the target imaging omics features and the target dosimetric features. Exemplarily, the target imaging omics features and the target dosimetric features are fused, and the process can be shown as follows:

[0042]

[0043] Among them, F t (x) is the target omics feature that includes both the target imaging omics feature and the target dose omics feature; F MRI (x i ) is the target imaging features obtained from different sequence MRI image data; i is the weight of radiomics features of different sequences; F DS (x) is the target dose-omics feature obtained from the three-dimensional dose distribution data.

[0044] S140. Predictively evaluate the subject's adaptive radiotherapy regimen based on the target omics features to obtain regimen evaluation data.

[0045] Among them, the program evaluation data is used to determine whether the subject needs to carry out an adaptive radiotherapy program. It can be understood that the program evaluation data is used to describe whether the subject needs to carry out an adaptive radiotherapy program, thereby helping medical workers to accurately distinguish situations where it is necessary to carry out adaptive radiotherapy for nasopharyngeal carcinoma.

[0046] Specifically, the constructed predictive evaluation model is used to predict and evaluate the adaptive radiotherapy regimen of the subjects based on the target omics characteristics and obtain regimen evaluation data.

[0047] Exemplarily, a logistic regression algorithm and a Sigmoid function are used for predictive evaluation. Taking the target omics feature as the input variable, for each input target omics feature, this prediction model will output a probability value in the range of [0,1]. It can be understood that this probability value is the predictive evaluation data used to determine whether the subject needs to carry out an adaptive radiotherapy program, that is, the program evaluation data. Furthermore, if this probability value is closer to 1, it means that the patient who provides the fusion omics feature is more likely to need adaptive radiotherapy. Similarly, if this probability value is closer to 0, it means that the patient who provides the fusion omics feature is less likely to need adaptive radiotherapy.

[0048] Optionally, when constructing a predictive evaluation model, the adaptive radiotherapy data evaluated in clinical treatment is combined with a machine learning algorithm to construct a predictive evaluation model. The adaptive radiotherapy data evaluated in clinical treatment is the data used by doctors to evaluate whether the radiotherapy plan (i.e., adaptive radiotherapy, ART) needs to be adjusted according to the patient's specific conditions and treatment response during the actual clinical treatment process. The data can be represented by a single numerical value or by a probability value.

[0049] Exemplarily, a suitable machine algorithm can be used to construct a predictive evaluation model. For example, a predictive evaluation model is constructed using algorithms such as Lasso logistic regression, support vector machine SVM, decision tree, random forest, or AdaBoost, with target omics features as variables, and the parameters of the model are determined by fitting through maximum likelihood estimation, and the adaptive radiotherapy data evaluated in clinical treatment is used as the prediction endpoint. If the value of the adaptive radiotherapy data evaluated in clinical treatment is 1, it means that ART needs to be carried out, and the value of the adaptive radiotherapy data evaluated in clinical treatment is 0, which can represent a situation where ART is not needed.

[0050] In the above-mentioned embodiment, the MRI image data and treatment plan data of the subject are first obtained, wherein the treatment plan data includes three-dimensional dose distribution data. Feature extraction is performed based on the MRI image data to obtain the target imaging omics features, and feature extraction is performed based on the three-dimensional dose distribution data to obtain the target dose omics features. The extracted target imaging omics features and target dose omics features are merged and fused to obtain the target omics features. Finally, the adaptive radiotherapy regimen of the subject is predicted and evaluated based on the target omics features, and the prediction and evaluation model constructed based on the machine learning algorithm is used to obtain the regimen evaluation data. At this point, an auxiliary means is realized to quickly, accurately and effectively predict and evaluate whether the subject needs to carry out an adaptive radiotherapy regimen, thereby helping medical workers to accurately distinguish situations where it is necessary to carry out adaptive radiotherapy for nasopharyngeal carcinoma.

[0051] In some embodiments, please refer to the attached Figure 2 The treatment plan data also includes delineation structure data and CT image data. Before determining the target radiomics features of the MRI image data and the target dosimetric features of the three-dimensional dose distribution data, the method further includes:

[0052] S210: Determine a first distribution of the lesion region on the MRI image data and a second distribution of the lesion region on the three-dimensional dose distribution data based on the delineated structure data and the CT image data.

[0053] Accordingly, the target radiomics features of the MRI image data and the target dosimetric features of the three-dimensional dose distribution data are determined, including:

[0054] S220 , extracting features from the MRI image data according to the first distribution to obtain target imaging omics features.

[0055] S230 , extracting features from the three-dimensional dose distribution data according to the second distribution to obtain target dose omic features.

[0056] Among them, the outline structure data is the data of accurate description and digital representation of the patient's anatomical structure, and its scope includes tumor target area, normal tissue and potential organs at risk (OAR) represented by DICOM format. CT image data is cross-sectional image data used to provide detailed anatomical structure information inside the subject's body. The first distribution situation is used to characterize the mapping area of ​​lesion distribution in MRI image data corresponding to CT image data; the second distribution situation is used to characterize the dose mapping situation on the three-dimensional dose distribution data corresponding to CT image data.

[0057] Specifically, determining a first distribution of the lesion region on the MRI image data and a second distribution of the lesion region on the three-dimensional dose distribution data is achieved by migrating the tumor target area information delineated on the CT image data to the MRI image data and the three-dimensional dose distribution data, wherein migration refers to the process of copying or mapping the structure or region information in one image mode to another image mode.

[0058] Exemplarily, the tumor target area information on the CT image data is migrated to the MRI image data to obtain the distribution of the primary tumor area (GTVnx) and the cervical lymph node metastasis area (GTVnd) on the MRI image data, that is, to obtain the first distribution of the lesion area on the MRI image data. Similarly, the tumor target area information on the CT image data is migrated to the three-dimensional dose distribution data to obtain the dose distribution of the primary tumor area (GTVnx) and the cervical lymph node metastasis area (GTVnd) on the three-dimensional dose distribution data, that is, to obtain the second distribution of the lesion area on the three-dimensional dose distribution data.

[0059] Furthermore, feature extraction was performed based on the distribution of the primary tumor region (GTVnx) and the cervical lymph node metastasis region (GTVnd) on the MRI image data to obtain the target imaging features. Feature extraction was performed based on the dose distribution of the primary tumor region (GTVnx) and the cervical lymph node metastasis region (GTVnd) on the three-dimensional dose distribution data to obtain the target dose features.

[0060] In the above embodiment, by migrating the tumor target area information delineated on the CT image data to the MRI image data and the three-dimensional dose distribution data, a first distribution of the lesion area on the MRI image data and a second distribution of the lesion area on the three-dimensional dose distribution data are determined. Further, based on the first distribution and the second distribution, the target imaging features of the MRI image data and the target dose features of the three-dimensional dose distribution data are determined, so as to evaluate the detailed features of the tumor on the MRI image data, evaluate the dose characteristics of the radiotherapy on the three-dimensional dose distribution data, optimize the treatment plan, improve the treatment effect, and reduce the damage to normal tissues.

[0061] In some embodiments, a first distribution of the lesion area on the MRI image data and a second distribution of the lesion area on the three-dimensional dose distribution data are determined based on the delineated structure data and the CT image data, including: aligning the CT image data and the MRI image data to obtain the registered MRI image data aligned with the pixel position of the CT image data; determining the first distribution based on the delineated structure data and the registered MRI image data, and determining the second distribution based on the delineated structure data and the three-dimensional dose distribution data.

[0062] Wherein, registration can refer to the image data from the same subject but may be obtained by different imaging technologies, or the image data obtained by the same technology at different time points, and the operation of spatial alignment is performed. Specifically, the image elastic registration method based on mutual information is used to register CT image data and MRI image data, and the registered MRI image data aligned with the pixel position of the CT image data is obtained. Wherein, mutual information is a measure of the similarity between two image data, and the elastic registration criterion allows non-rigid deformation alignment between images. Exemplarily, the MRI image data of the subject are registered with the CT image data. By calculating the mutual information between the two images, the registration algorithm finds the best transformation (including rotation, translation, scaling and more complex deformation) so that each pixel position in the MRI image data is aligned with the corresponding position in the CT image data.

[0063] Specifically, the tumor target area information on the registered CT image data is migrated to the MRI image data to obtain a first distribution of the lesion area on the MRI image data. Similarly, the tumor target area information on the registered CT image data is migrated to the three-dimensional dose distribution data to obtain a second distribution of the lesion area on the three-dimensional dose distribution data.

[0064] In the above embodiment, by registering the CT image data and the MRI image data, the registered MRI image data is obtained, which is aligned with the pixel position of the CT image data, so that they overlap in space or are as close as possible, ensuring that the same anatomical structure in different image data can be identified and analyzed in the same coordinate system. Furthermore, the detailed characteristics of the tumor are evaluated on the MRI image data, and the dose characteristics of the radiotherapy are evaluated on the three-dimensional dose distribution data. The accuracy of predicting and evaluating the adaptive radiotherapy plan for the subject can be improved by accurate target area positioning.

[0065] In some embodiments, please refer to the attached Figure 3 , determine the target radiomics features of MRI image data and the target dosimetric features of 3D dose distribution data, including:

[0066] S310 , resample the three-dimensional dose distribution data to obtain intermediate dose distribution data having the same voxel dimension as the MRI image data.

[0067] Specifically, the intermediate dose distribution data is obtained by resampling the three-dimensional dose distribution data and performing resolution conversion on the three-dimensional dose distribution data to obtain intermediate dose distribution data having the same pixel size and voxel dimension as the MRI image data. Optionally, if the dimension of the three-dimensional dose distribution data is smaller than that of the MRI image data, the MRI image data can be three-dimensionally cropped to obtain intermediate image data having the same pixel size and voxel dimension as the three-dimensional dose distribution data.

[0068] S320, performing bias field correction and image enhancement operations on the MRI image data to obtain enhanced image data.

[0069] Among them, the bias field correction is a non-parametric non-uniformity correction algorithm. Exemplarily, the N4 bias field correction is used to estimate and correct the non-uniformity of the MRI image data. Further, the visual effect of the image data processed by the bias field correction is improved by an image enhancement operation to obtain enhanced image data. Exemplarily, the image enhancement operation performs image denoising and edge detection on the image data processed by the bias field correction to finally obtain the enhanced image data. Optionally, the processed image data is subjected to intensity standardization processing to convert the intensity values ​​of different image data into a unified metric.

[0070] S330, extracting features based on the enhanced image data to obtain target imaging omics features; extracting features based on the intermediate dose distribution data to obtain target dose omics features.

[0071] Specifically, various quantitative features can be extracted from the enhanced image data to obtain target imaging features. Similarly, various quantitative features can be extracted based on the intermediate dose distribution data to obtain target dose features.

[0072] In the above embodiment, the three-dimensional dose distribution data is resampled so that the three-dimensional dose distribution data and the MRI image data have the same pixel size as the CT image data. Bias field correction and image enhancement operations are performed based on the MRI image data. The bias field correction can obtain image data after eliminating the unevenness of the signal intensity, and the image enhancement operation can make the specific features in the image more prominent. Furthermore, the processed image data can be subjected to intensity standardization processing to eliminate the influence caused by the different magnitudes of intensity values ​​in different image data, so that the intensity values ​​of different sequence image data are comparable.

[0073] In some embodiments, please refer to the attached Figure 4 , determine the target radiomics features of MRI image data and the target dosimetric features of 3D dose distribution data, including:

[0074] S410 , extracting first-order features, morphological features, texture features, and filtering features based on the MRI image data to obtain intermediate imaging omics features.

[0075] Among them, the first-order features are used to describe the basic statistics of the image intensity distribution, such as mean, variance, skewness and kurtosis; morphological features are used to describe the characteristics of the shape and size of objects in the image, such as surface area, volume and sphericity; texture features are used to reflect the complexity of the relationship between pixels in the image; filter features are used to identify specific patterns or structures in the image. Exemplarily, the open source Pyradiomics toolkit can be used to analyze MRI image data and extract first-order features, morphological features, texture features and filter features to obtain intermediate imaging omics features.

[0076] S420, extracting first-order features, morphological features, texture features, and filtering features based on the three-dimensional dose distribution data to obtain intermediate dose-omics features.

[0077] Among them, the first-order features are used to describe the statistics of dose distribution, such as mean, variance, minimum and maximum values; morphological features are used to describe the spatial morphology of dose distribution, such as volume and surface area; texture features are used to reflect the texture information of dose distribution, involving the roughness or uniformity of dose distribution; filter features are used to identify specific patterns in dose distribution. Similarly, the Pyradiomics toolkit can be used to analyze three-dimensional dose distribution data and extract first-order features, morphological features, texture features, and filter features to obtain intermediate dose-omics features.

[0078] S430, obtaining target radiomics features by standardizing the intermediate radiomics features.

[0079] Exemplarily, the intermediate radiomics features are standardized using z-score standardization and ComBat compensation methods. Specifically, by calculating the difference between each eigenvalue of the intermediate radiomics feature and its mean, and then dividing the difference by the standard deviation, the eigenvalue is converted to a standard score with zero mean and unit variance to obtain the standardized radiomics feature. Furthermore, the ComBat compensation method is used to identify and adjust the systematic differences between radiomics features, and finally obtain the target radiomics feature.

[0080] S440. The target dose omic characteristics are obtained by standardizing the intermediate dose omic characteristics.

[0081] Similarly, the intermediate dose-omics features were standardized using z-score normalization and ComBat compensation method to obtain the target dose-omics features.

[0082] In the above embodiment, the Pyradiomics toolkit is used to analyze the MRI image data and perform feature extraction to obtain intermediate imaging omics features. Similarly, the Pyradiomics toolkit is used to analyze the three-dimensional dose distribution data and perform feature extraction to obtain intermediate dose omics features. Further, the intermediate imaging omics features are standardized using the z-score standardization and ComBat compensation method, so that the target imaging omics features obtained have a unified metric. Similarly, the intermediate dose omics features are standardized using the z-score standardization and ComBat compensation method, so that the target dose omics obtained has a unified metric, while reducing the impact caused by the data batch effect, and providing comparable feature data for predicting and evaluating the adaptive radiotherapy regimen of the subject.

[0083] In some embodiments, the MRI image data is multi-sequence MRI image data, and determining a target radiomics feature of the MRI image data comprises:

[0084] Obtain the imaging features of each sequence of multiple MRI image data.

[0085] Among them, the multi-sequence MRI image data is a series of MRI image data obtained using different imaging parameters or techniques. For example, the multi-sequence MRI image data includes T1-weighted imaging (T1w), T1-weighted enhanced imaging (T1wC), T1-weighted water-fat separation DIXON sequence (T1wDIXON) and T2-weighted imaging (T2w), and each sequence has its own unique imaging focus. Specifically, a series of parameters used in the imaging process are changed to change the contrast of the MRI image data and the displayed tissue characteristics to obtain multi-sequence MRI image data.

[0086] The radiomics features of multiple sequences are weighted averaged to obtain the target radiomics features.

[0087] Among them, a series of quantitative features that can describe and quantify image data extracted from a specific sequence of MRI are called sequential imaging omics features. Furthermore, the sequential imaging omics features are weighted and weighted combined to obtain target imaging omics features that integrate multiple sequence information data. Specifically, when combining the imaging omics features of multiple sequences to obtain the target imaging omics features, different levels of importance can be given to the judgment and predictive evaluation of the adaptive radiotherapy regimen of the subjects according to the different importance of different sequential imaging omics features. For example, if the T1w sequence is crucial for distinguishing the types of nasopharyngeal carcinoma tumors, it will be given a higher weight. If the T2w sequence is more helpful for evaluating edema around nasopharyngeal carcinoma tumors, it will also receive a higher weight.

[0088] The target imaging omics feature and the target dose omics feature are fused to obtain the target omics feature. The method further includes:

[0089] The target imaging omics features and target dose omics features are combined to obtain the target omics features.

[0090] Among them, splicing and combining can refer to a processing method of merging two different features into a unified feature set. Specifically, by splicing and combining the selected target imaging omics features and target dose omics features into a large feature matrix, a target omics feature that includes both the target imaging omics features and the target dose omics features is obtained. Exemplarily, the target imaging omics features and the target dose omics features are fused, and the process can be shown as follows:

[0091]

[0092] Among them, F t (x) is the target omics feature that includes both the target imaging omics feature and the target dose omics feature; F MRI (x i) is the target imaging features obtained from different sequence MRI image data; i is the weight of radiomics features of different sequences; F DS (x) is the target dose-omics feature obtained from the three-dimensional dose distribution data.

[0093] In the above-mentioned embodiment, the target imaging omics features are obtained by weighted averaging the imaging omics features of each sequence of the multi-sequence MRI image data, i.e., the imaging omics features of multiple sequences, taking into account the differences in imaging features in the lesion area between the multi-sequence MRI image data. Further, the target imaging omics features are spliced ​​and combined with the target dose omics features to obtain the target omics features. Since the target dose omics features can characterize the spatial heterogeneity of the dose distribution, the target imaging omics features characterizing the differences in imaging features are fused with the target dose omics features characterizing the spatial heterogeneity of the dose distribution, which reduces the computational complexity of medical image processing, reduces the complexity of modeling, simplifies the cumbersome operation process, and improves the accuracy of the prediction.

[0094] In some embodiments, please refer to the attached Figure 5 , the target imaging features and target dose features are combined to obtain the target omics features, including:

[0095] S510, combining the target imaging omics features and the target dose omics features to obtain fused omics features.

[0096] Among them, the fused omics features constitute the feature set to be screened. The feature set to be screened is a feature set that includes all fused omics features. Specifically, by splicing and fusing the target imaging omics features and the target dose omics features into a large feature matrix, the fused omics features that include both the target imaging omics features and the target dose omics features are obtained.

[0097] S520 , performing redundancy screening on the feature set to be screened according to the correlation and similarity between any two features to be screened in the feature set to be screened, to obtain an intermediate feature set.

[0098] Among them, screening can refer to the process of selecting the most informative and most helpful target omics features for prediction accuracy from a large number of fused omics feature sets. Specifically, redundant features can be identified based on correlation and similarity evaluation, and the correlation and similarity between any two features to be screened in the feature set to be screened are calculated, and features with low correlation and high similarity to other features to be screened in the feature set to be screened are identified and removed. Exemplarily, the maximum correlation minimum redundancy (mRMR) algorithm can be used to measure the correlation and similarity between any two features to be screened based on the Pearson correlation coefficient and mutual information, and features with low correlation and high similarity are eliminated, and features with high correlation and low similarity are retained to obtain an intermediate feature set.

[0099] S530 , performing shrinkage selection on the omics features in the intermediate feature set to obtain target omics features.

[0100] Among them, shrinkage selection can refer to selecting the most influential features for predicting and evaluating the adaptive radiotherapy regimen of the subject from the intermediate feature set. Specifically, the screened intermediate feature set is screened again. Exemplarily, the least absolute shrinkage and selection operator (LASSO) algorithm can be applied to take the intermediate feature set as input, process the intermediate feature set, and implement feature selection through the L1 norm penalty term to obtain the target feature set. Furthermore, the omics features in the target feature set that has undergone the screening operation can be regarded as the most representative and predictive features, that is, as the target omics features.

[0101] In the above embodiment, the target imaging omics features and the target dose omics features are spliced ​​and combined to obtain fused omics features. Further, the fused omics features are screened and processed. First, the maximum relevance minimum redundancy algorithm (mRMR) is used to reduce the redundancy between the features to be screened, and then the LASSO algorithm is applied to further streamline the features to retain the target omics features that are most informative and most helpful for prediction accuracy, reduce redundancy between features, and improve the accuracy of prediction evaluation.

[0102] This specification also provides a medical image processing method. Figure 6 , the method comprises the following steps:

[0103] S602: Obtain MRI image data and treatment plan data of the subject.

[0104] The MRI image data may be multi-sequence MRI image data, obtained by performing an MRI examination on the nasopharyngeal region of the subject. The treatment plan data may include three-dimensional dose distribution data, CT image data, and delineated structure data. Optionally, obtaining the MRI image data of the subject may include obtaining multi-sequence MRI image data.

[0105] S604 , registering the CT image data and the MRI image data to obtain registered MRI image data whose pixel positions are aligned with those of the CT image data.

[0106] Specifically, the CT image data and the MRI image data are registered using an image elastic registration method based on mutual information to obtain registered MRI image data whose pixel positions are aligned with those of the CT image data.

[0107] S606: Determine a first distribution of the lesion region on the MRI image data and a second distribution of the lesion region on the three-dimensional dose distribution data based on the delineated structure data and the CT image data.

[0108] Specifically, the tumor target area information on the CT image data is transferred to the MRI image data to obtain a first distribution of the lesion area on the MRI image data. Similarly, the tumor target area information on the CT image data is transferred to the three-dimensional dose distribution data to obtain a second distribution of the lesion area on the three-dimensional dose distribution data.

[0109] S608 , resample the three-dimensional dose distribution data to obtain intermediate dose distribution data having the same voxel dimension as the MRI image data.

[0110] Specifically, the three-dimensional dose distribution data are resampled to have the same pixel size as the MRI image data. If the dimension of the three-dimensional dose distribution data is smaller than that of the MRI image data, the MRI image data can be three-dimensionally cropped to obtain intermediate image data having the same pixel size and voxel dimension as the three-dimensional dose distribution data.

[0111] S610 , extracting features from the intermediate dose distribution data according to the second distribution to obtain intermediate dose omics features.

[0112] Specifically, the Pyradiomics toolkit was used to extract features based on the intermediate dose distribution data, including extracting first-order features, morphological features, texture features, and filtering features to obtain intermediate dose omics features.

[0113] S612. Obtain target dose omic characteristics by standardizing the intermediate dose omic characteristics.

[0114] Specifically, the intermediate dose-omics features were normalized using z-score normalization and ComBat compensation methods.

[0115] S614, performing bias field correction and image enhancement operations on the multiple sequences of MRI image data to obtain multiple sequences of enhanced image data.

[0116] Specifically, the N4 bias field correction is used to correct the inhomogeneity of the multi-sequence MRI image data, and an image enhancement operation is performed on the multi-sequence enhanced image data. Optionally, the processed image data is subjected to intensity standardization.

[0117] S616 . Perform feature extraction on the enhanced image data of the multiple sequences according to the first distribution to obtain intermediate image omics features of the multiple sequences.

[0118] Specifically, the Pyradiomics toolkit is used to extract features based on the enhanced image data, including extracting first-order features, morphological features, texture features, and filtering features to obtain intermediate imaging omics features.

[0119] S618. Standardize the intermediate imaging features of the multiple sequences to obtain the imaging features of each sequence of the multiple MRI image data.

[0120] Specifically, the intermediate radiomics features are standardized using z-score standardization and ComBat compensation methods. Optionally, the standardized intermediate radiomics features may include radiomics features of each sequence of multiple sequences of MRI image data.

[0121] S620 , performing weighted average processing on the radiomics features of multiple sequences to obtain target radiomics features.

[0122] Among them, a series of quantitative features extracted from specific MRI sequences that can describe and quantify image data are called sequence radiomics features. Furthermore, the sequence radiomics features are weighted and combined to obtain target radiomics features that integrate multiple sequence information data.

[0123] S622, combining the target imaging omics features and the target dose omics features to obtain fused omics features, wherein the fused omics features constitute a feature set to be screened.

[0124] The feature set to be screened is a feature set that includes all fused omics features. Specifically, by splicing and fusing the target imaging omics features and the target dose omics features into a large feature matrix, the fused omics features that include both the target imaging omics features and the target dose omics features are obtained.

[0125] S624 , performing redundancy screening on the feature set to be screened according to the correlation and similarity between any two features to be screened in the feature set to be screened, to obtain an intermediate feature set.

[0126] Specifically, the mRMR algorithm is used to measure the correlation and similarity between any two features to be screened based on the Pearson correlation coefficient and mutual information, and the features with low correlation and high similarity are eliminated, while the features with high correlation and low similarity are retained to obtain the intermediate feature set.

[0127] S626. Perform shrinkage selection on the omics features in the intermediate feature set to obtain target omics features.

[0128] Specifically, the least absolute shrinkage and selection operator (LASSO) algorithm is applied to take the intermediate feature set as input, process the intermediate feature set, and implement feature selection through the L1 norm penalty term to obtain the target feature set. Furthermore, the omics features in the target feature set that have been screened can be regarded as the most representative and predictive features, that is, as the target omics features.

[0129] S628. Predictively evaluate the subject's adaptive radiotherapy regimen based on the target omics features to obtain regimen evaluation data.

[0130] Specifically, the constructed prediction and evaluation model can be used to predict and evaluate the adaptive radiotherapy regimen of the subject based on the target omics characteristics and obtain regimen evaluation data.

[0131] Exemplarily, a logistic regression algorithm and a Sigmoid function are used for predictive evaluation. Taking the target omics feature as the input variable, for each input target omics feature, this prediction model will output a probability value in the range of [0,1]. It can be understood that this probability value is the predictive evaluation data used to determine whether the subject needs to carry out an adaptive radiotherapy program, that is, the program evaluation data. Furthermore, if this probability value is closer to 1, it means that the patient who provides the fusion omics feature is more likely to need adaptive radiotherapy. Similarly, if this probability value is closer to 0, it means that the patient who provides the fusion omics feature is less likely to need adaptive radiotherapy.

[0132] Optionally, before applying the prediction and evaluation model, the receiver operating characteristic curve (ROC) can be used to verify the accuracy of the prediction results of the prediction and evaluation model.

[0133] Specifically, the test target omics features are input into the prediction and evaluation model to obtain the predicted probability value of whether the subject needs to carry out the adaptive radiotherapy scheme, that is, the test scheme evaluation data is obtained. Then the threshold is set. If the output predicted probability value is greater than the threshold, it is predicted that the case needs to perform ART. Otherwise, it is predicted that ART is not required. Further, the test target omics features and test scheme evaluation data used in constructing the prediction and evaluation model are analyzed with the MRI imaging data, treatment plan data and evaluation data of whether the subject needs to carry out the adaptive radiotherapy scheme in actual clinical treatment. The analysis results are divided into 4 different types: true positive (TP, cases correctly predicted to need ART), false positive (FP, cases wrongly predicted to need ART), true negative (TN, cases correctly predicted not to need ART) and false negative (FN, cases wrongly predicted not to need ART), thereby calculating the probability values ​​of true positive and false positive, and then a series of true positive and false positive probability values ​​are obtained by changing the threshold setting. Furthermore, the true positive probability value is used as the ordinate, and the false positive probability value is used as the abscissa, and points are drawn in the coordinate system and connected to obtain the ROC curve. The predictive ability of the prediction evaluation model is quantitatively verified by calculating the area under the ROC curve AUC value. For example, the AUC value is generally between 0.5 and 1, and an AUC value less than 0.5 rarely occurs in practical applications. The larger the AUC value, the better the predictive performance of the model. If the AUC value is 0.5, it means that it is a random prediction with no predictive value at all. The AUC value of 1 means that the prediction is completely correct.

[0134] It should be understood that, although the various steps in the above flowchart are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear description in this article, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least a part of the steps in the above flowchart may include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0135] The present specification also provides a medical image processing device 700, such as Figure 7 As shown, it includes: a data acquisition module 710, a feature determination module 720, a feature fusion module 730 and a prediction evaluation module 740, wherein:

[0136] The data acquisition module 710 is used to acquire the MRI image data and treatment plan data of the subject. The MRI image data is obtained by performing an MRI examination on the nasopharyngeal area of ​​the subject; the treatment plan data includes three-dimensional dose distribution data;

[0137] A feature determination module 720, for determining target radiomics features of MRI image data and target dosimics features of three-dimensional dose distribution data;

[0138] A feature fusion module 730 is used to fuse the target imaging omics features and the target dose omics features to obtain the target omics features;

[0139] The prediction and evaluation module 740 is used to predict and evaluate the adaptive radiotherapy regimen of the subject based on the target omics features to obtain regimen evaluation data.

[0140] In some embodiments, the MRI image data is multi-sequence MRI image data, which is obtained by performing an MRI examination on the nasopharyngeal region of the subject. The treatment plan data includes three-dimensional dose distribution data, CT image data, and delineation structure data.

[0141] In some embodiments, the medical image processing device 700 also includes a distribution determination module for determining a first distribution of the lesion area on the MRI image data and a second distribution of the lesion area on the three-dimensional dose distribution data based on the outline structure data and the CT image data.

[0142] The feature determination module 720 is also used to extract features from the MRI image data according to the first distribution to obtain target imaging features; and to extract features from the three-dimensional dose distribution data according to the second distribution to obtain target dose features.

[0143] In some embodiments, the distribution determination module is also used to align the CT image data and the MRI image data to obtain the registered MRI image data aligned with the pixel position of the CT image data; determine the first distribution condition based on the outline structure data and the registered MRI image data, and determine the second distribution condition based on the outline structure data and the three-dimensional dose distribution data.

[0144] In some embodiments, the MRI image data is multiple sequences of MRI image data. The feature determination module 720 is also used to obtain the radiomics features of each sequence of the multiple sequences of MRI image data; and perform weighted average processing based on the radiomics features of multiple sequences to obtain the target radiomics features.

[0145] In some embodiments, the feature determination module 720 is also used to resample the three-dimensional dose distribution data to obtain intermediate dose distribution data with the same voxel dimension as the MRI image data; perform bias field correction and image enhancement operations on the MRI image data to obtain enhanced image data; perform feature extraction based on the enhanced image data to obtain target imaging omics features; perform feature extraction based on the intermediate dose distribution data to obtain target dose omics features.

[0146] In some embodiments, the feature determination module 720 is also used to extract first-order features, morphological features, texture features, and filtering features based on MRI image data to obtain intermediate imaging omics features; extract first-order features, morphological features, texture features, and filtering features based on three-dimensional dose distribution data to obtain intermediate dose omics features; obtain target imaging omics features by standardizing the intermediate imaging omics features; obtain target dose omics features by standardizing the intermediate dose omics features.

[0147] In some embodiments, the feature fusion module 730 is further used to perform weighted average processing on the radiomic features of the multiple sequences to obtain the target radiomic features; and to splice and combine the target radiomic features with the target dosimic features to obtain the target radiomic features. Specifically, by splicing and merging the selected target radiomic features and target dosimic features into a large feature matrix, a target radiomic feature that includes both the target radiomic features and the target dosimic features is obtained.

[0148] In some embodiments, the medical image processing device 700 further includes a feature screening module, wherein the target imaging omics features and the target dose omics features are spliced ​​and combined to obtain fused omics features; the fused omics features constitute a feature set to be screened. The feature screening module is used to perform redundant screening of the feature set to be screened according to the correlation and similarity between any two features to be screened in the feature set to be screened, to obtain an intermediate feature set; and to perform shrinkage selection on the omics features in the intermediate feature set to obtain the target omics features.

[0149] In some embodiments, the prediction and evaluation module 740 is further used to utilize the constructed prediction and evaluation model to perform prediction and evaluation on the adaptive radiotherapy regimen of the subject based on the target omics features and obtain regimen evaluation data.

[0150] For the specific definition of a medical image processing device, please refer to the definition of a medical image processing method above, which will not be repeated here. Each module in the above medical image processing device can be implemented in whole or in part by software, hardware and a combination thereof. Each of the above modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.

[0151] A medical image processing device in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.

[0152] The embodiment of the present application also provides a computer device, which may be a terminal, and its internal structure diagram may be as follows: Figure 8 As shown. The computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a medical image processing method is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a key, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse, etc.

[0153] Those skilled in the art will understand that Figure 8 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0154] The embodiment of the present application also provides a computer-readable storage medium, and the above-mentioned method according to the embodiment of the present application can be implemented in hardware and firmware, or can be implemented as a computer code that can be recorded in a storage medium, or can be implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and will be stored in a local storage medium through a network download, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state hard disk, etc.; further, the storage medium can also include a combination of the above-mentioned types of memory. It can be understood that a computer, a processor, a microprocessor controller or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor or hardware, the method shown in the above embodiment is implemented.

[0155] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0156] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0157] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of this application, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0158] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0159] Each embodiment in this specification is described in a progressive 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 computer devices and computer-readable storage media, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiments.

[0160] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent substitution and improvement made within the spirit and principle of the present application, etc., should all be included within the scope of the claims of the present application.

[0161] Although the embodiments of the present application are described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present application, and such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A medical image processing method, characterized in that: The method comprises: Acquire MRI image data and treatment plan data of the subject; wherein the MRI image data is obtained by performing an MRI examination on the nasopharyngeal region of the subject; and the treatment plan data includes three-dimensional dose distribution data; Determining target radiomics features of the MRI image data and target dosimics features of the three-dimensional dose distribution data; fusing the target imaging omics features and the target dose omics features to obtain target omics features; Based on the target omics features, a predictive evaluation is performed on the adaptive radiotherapy regimen of the subject to obtain regimen evaluation data.

2. The method according to claim 1, characterized in that The treatment plan data also includes delineation structure data and CT image data; before determining the target imaging features of the MRI image data and the target dose features of the three-dimensional dose distribution data, the method also includes: Determine a first distribution of the lesion region on the MRI image data and a second distribution of the lesion region on the three-dimensional dose distribution data based on the delineated structure data and the CT image data; The determining of the target radiomics features of the MRI image data and the target dosemic features of the three-dimensional dose distribution data comprises: Extracting features from the MRI image data according to the first distribution to obtain the target imaging features; Feature extraction is performed on the three-dimensional dose distribution data according to the second distribution to obtain the target dose omic feature.

3. The method according to claim 2, characterized in that The determining, based on the delineated structure data and the CT image data, a first distribution of the lesion area on the MRI image data, and a second distribution of the lesion area on the three-dimensional dose distribution data, comprises: Registering the CT image data and the MRI image data to obtain registered MRI image data aligned with pixel positions of the CT image data; The first distribution condition is determined based on the delineated structure data and the registered MRI image data, and the second distribution condition is determined based on the delineated structure data and the three-dimensional dose distribution data.

4. The method according to claim 1, characterized in that: The determining of the target radiomics features of the MRI image data and the target dosemic features of the three-dimensional dose distribution data comprises: Resampling the three-dimensional dose distribution data to obtain intermediate dose distribution data having the same voxel dimension as the MRI image data; Performing bias field correction and image enhancement operations on the MRI image data to obtain enhanced image data; Feature extraction is performed based on the enhanced image data to obtain the target imaging omics features; feature extraction is performed based on the intermediate dose distribution data to obtain the target dose omics features.

5. The method according to claim 1, characterized in that The determining of the target radiomics features of the MRI image data and the target dosemic features of the three-dimensional dose distribution data comprises: Extracting first-order features, morphological features, texture features, and filtering features based on the MRI image data to obtain intermediate imaging features; Extracting first-order features, morphological features, texture features, and filtering features based on the three-dimensional dose distribution data to obtain intermediate dose-omics features; By performing standardization processing on the intermediate radiomics features, the target radiomics features are obtained; The target dose omic characteristics are obtained by standardizing the intermediate dose omic characteristics.

6. The method according to claim 1, characterized in that The MRI image data is multi-sequence MRI image data; and determining the target imaging features of the MRI image data includes: Acquiring radiomics features of each sequence of the multiple sequences of MRI image data; Performing weighted average processing on the radiomics features of multiple sequences to obtain the target radiomics features; The fusion of the target imaging omics feature and the target dose omics feature to obtain the target omics feature includes: The target imaging omics features and the target dose omics features are spliced ​​and combined to obtain the target omics features.

7. The method according to claim 6, characterized in that The step of combining the target imaging omics feature with the target dose omics feature to obtain the target omics feature includes: The target imaging omics features and the target dose omics features are combined to obtain fused omics features; wherein the fused omics features constitute a feature set to be screened; Performing redundancy screening on the feature set to be screened according to the correlation and similarity between any two features to be screened in the feature set to be screened, to obtain an intermediate feature set; The omics features in the intermediate feature set are subjected to shrinkage selection to obtain target omics features.

8. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the steps of the method according to any one of claims 1 to 7 by executing the computer instructions.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the steps of any one of the methods of claims 1 to 7.

10. A medical image processing device, characterized in that: The device comprises: A data acquisition module, used to acquire MRI image data and treatment plan data of the subject; wherein the MRI image data is obtained by performing an MRI examination on the nasopharyngeal region of the subject; and the treatment plan data includes three-dimensional dose distribution data; A feature determination module, used to determine the target radiomics feature of the MRI image data and the target dosemic feature of the three-dimensional dose distribution data; A feature fusion module, used for fusing the target imaging omics feature and the target dose omics feature to obtain the target omics feature; A prediction and evaluation module is used to predict and evaluate the adaptive radiotherapy regimen of the subject based on the target omics features to obtain regimen evaluation data.

Citation Information

Patent Citations

  • Method for performing multi-omics processing through data of patient with head and neck cancer, related computer readable medium and computing equipment

    CN116563192A

Cited By

  • Medical image enhancement processing method and system based on deep learning

    CN119671884A

  • A medical image enhancement processing method and system based on deep learning

    CN119671884B