Glioma chemotherapy reaction prediction model construction method, equipment and medium

Abstract

The invention discloses a glioma chemotherapy reaction prediction model construction method, equipment and a medium, and relates to the technical field of medical instruments. According to the method, feature extraction and screening are performed on magnetic resonance images based on a radiomics technology, a sample data set is constructed, then different machine learning models are trained by using the sample data set, and then the optimal trained machine learning model is selected as a glioma chemotherapy reaction prediction model, so that the prediction accuracy of the glioma chemotherapy reaction is improved. And the prediction of the glioma chemotherapy reaction is realized.

Description

Technical Field

[0001] The present application relates to the field of medical device technology, and in particular to a method, device and medium for constructing a glioma chemotherapy response prediction model. Background Art

[0002] Since it is difficult to completely remove tumor cells in the brain through surgery, glioma patients need chemotherapy after receiving neurosurgery. The chemotherapy process will cause continuous and great pain to the patient's body and mind. Hair loss, vomiting, and extremely decreased immunity are common complications, and the patient's quality of life is seriously reduced; at the same time, the cost of chemotherapy is high, which brings a serious economic burden to the patient's family. More importantly, gliomas are highly heterogeneous, and the treatment effects after chemotherapy vary greatly. Some patients are not sensitive to chemotherapy, which means that patients endure great pain but cannot benefit from it. However, since it is not clear which part of the patients are sensitive to chemotherapy in clinical practice, most patients can only choose to receive unified standard chemotherapy, which is essentially a wide-ranging trial treatment. Therefore, how to predict the patient's response to chemotherapy before chemotherapy begins and avoid unnecessary physical and mental pain and economic burden is an important issue that needs to be solved in the treatment of brain gliomas. Summary of the invention

[0003] The purpose of this application is to provide a method, device and medium for constructing a glioma chemotherapy response prediction model, so as to establish a glioma chemotherapy response prediction model and realize the prediction of glioma chemotherapy response.

[0004] To achieve the above objectives, this application provides the following solutions.

[0005] In a first aspect, the present application provides a method for constructing a glioma chemotherapy response prediction model, comprising the following steps.

[0006] A longitudinal multi-time point magnetic resonance image set is acquired; the longitudinal multi-time point magnetic resonance image set includes a plurality of magnetic resonance image sequences, and the magnetic resonance image sequence includes a magnetic resonance image group of high-grade glioma of the same patient at least three time points; the three time points are respectively located before surgery, before chemotherapy after surgery, and after the end of chemotherapy; the magnetic resonance image group includes: one or more of a T1 weighted phase, a T2 weighted phase, and an enhanced T1 weighted phase.

[0007] The volume changes of the residual tumor in the magnetic resonance image group after surgery and before chemotherapy and the magnetic resonance image group after the chemotherapy in each magnetic resonance image sequence are calculated as the volume changes corresponding to each magnetic resonance image sequence.

[0008] The chemotherapy sensitivity corresponding to each magnetic resonance imaging sequence is determined according to the volume change corresponding to each magnetic resonance imaging sequence.

[0009] Feature extraction and screening are performed on the pre-operative magnetic resonance image group in each magnetic resonance image sequence to obtain a feature set corresponding to each magnetic resonance image sequence.

[0010] The feature set corresponding to each magnetic resonance imaging sequence is used as input, and the chemotherapy sensitivity corresponding to each magnetic resonance imaging sequence is used as a label to construct a sample data set.

[0011] Based on the sample data set, multiple machine learning models are trained respectively, and the optimal trained machine learning model is selected as a glioma chemotherapy response prediction model.

[0012] In a second aspect, the present application provides a computer device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-mentioned method for constructing a glioma chemotherapy response prediction model.

[0013] In a third aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned method for constructing a glioma chemotherapy response prediction model.

[0014] According to the specific embodiments provided in this application, this application has the following technical effects.

[0015] The present application provides a method, device and medium for constructing a glioma chemotherapy response prediction model. The present application extracts and screens features of magnetic resonance images based on imaging genomics technology, constructs a sample data set, and then uses the sample data set to train different machine learning models separately, and then selects the optimal trained machine learning model as the glioma chemotherapy response prediction model, thereby realizing the prediction of glioma chemotherapy response. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0017] Figure 1 A flowchart of a method for constructing a glioma chemotherapy response prediction model provided in one embodiment of the present application.

[0018] Figure 2 This is a receiver operating characteristic curve diagram of the glioma chemotherapy sensitivity prediction model provided in one embodiment of the present application.

[0019] Figure 3A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0020] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only 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 ordinary technicians in this field without creative work are within the scope of protection of this application.

[0021] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0022] Radiomics technology is the product of the development and progress of image pattern recognition technology. It can extract a large number of objective quantitative features that are difficult to identify with the naked eye from medical images, including signal intensity distribution, three-dimensional shape, texture characteristics, and wavelet transformed features of tumor images. This technology realizes the quantitative description of the characteristics of solid tumor images and provides a new direction for the predictive analysis of clinical diagnosis and treatment-related events of such diseases. In glioma research, radiomics has shown a wide range of application potential in many fields, especially in the prediction of treatment response. For example, the radiomics model developed by Grossmann P and Kickingereder P et al. can effectively screen out patients with recurrent glioblastoma who may benefit from bevacizumab (anti-angiogenic therapy). Similarly, George E et al. constructed another radiomics model to identify patients with recurrent glioblastoma who may benefit from durvalumab (immunotherapy). These models provide important support for personalized treatment and are expected to improve patient treatment outcomes.

[0023] Magnetic resonance imaging (MRI) plays a vital role in the diagnosis, treatment planning, and follow-up monitoring of gliomas. Different MRI sequences, including T1-weighted phase (T1 for short), T2-weighted phase (T2 for short), enhanced T1-weighted phase (CE for short), and fluid attenuated inversion recovery (FLAIR), provide complementary information for evaluating imaging characteristics such as tumor structure, edema, and necrosis. In clinical practice, the treatment response of gliomas is usually evaluated according to the Response Assessment in Neuro-Oncology (RANO) criteria, which is based on longitudinal MRI scans at multiple time points and evaluates the treatment effect by monitoring the dynamic changes of tumor volume. Given the core role of MRI in glioma management, imaging omics models based on MRI data have become an ideal tool for predicting the sensitivity of glioma chemotherapy, which helps to achieve personalized treatment plans.

[0024] Using MRI data to construct a glioma chemotherapy response prediction model to predict the treatment response of high-grade glioma patients to chemotherapy is expected to assist doctors in selecting the most appropriate individualized treatment plan, improve treatment outcomes, and avoid unnecessary physical and mental pain and economic burden.

[0025] In an exemplary embodiment, Figure 1 As shown, a method for constructing a glioma chemotherapy response prediction model is provided, including the following steps 101 to 106.

[0026] Step 101, obtaining a longitudinal multi-time point magnetic resonance image set; the longitudinal multi-time point magnetic resonance image set includes multiple magnetic resonance image sequences, and the magnetic resonance image sequence includes a magnetic resonance image group of high-grade glioma of the same patient at least three time points; the three time points are respectively located before surgery, before chemotherapy after surgery, and after the end of chemotherapy; the magnetic resonance image group includes: one or more of: T1 weighted phase, T2 weighted phase and enhanced T1 weighted phase.

[0027] Step 102 , calculating the volume change of the residual tumor in the magnetic resonance image group before chemotherapy and the magnetic resonance image group after chemotherapy in each magnetic resonance image sequence as the volume change corresponding to each magnetic resonance image sequence.

[0028] Step 103: determining the chemotherapy sensitivity corresponding to each magnetic resonance image sequence according to the volume change corresponding to each magnetic resonance image sequence.

[0029] Step 104 , extracting and screening features of the pre-operative magnetic resonance image groups in each magnetic resonance image sequence to obtain a feature set corresponding to each magnetic resonance image sequence.

[0030] Step 105 , taking the feature set corresponding to each magnetic resonance imaging sequence as input, taking the chemotherapy sensitivity corresponding to each magnetic resonance imaging sequence as a label, and constructing a sample data set.

[0031] Step 106: train multiple machine learning models based on the sample data set, and select the best trained machine learning model as the glioma chemotherapy response prediction model.

[0032] By implementing the above steps 101 to 106, a glioma chemotherapy response prediction model can be established to achieve the prediction of chemotherapy response.

[0033] In another exemplary embodiment of the present application, the above step 101 is implemented in the following manner.

[0034] A longitudinal multi-time point magnetic resonance image set of high-grade (Word Health Organization, WHO grade 3-4) gliomas was collected in the brain glioma database. The longitudinal multi-time point magnetic resonance image set includes multiple magnetic resonance image sequences, and the magnetic resonance image sequence includes a magnetic resonance image group of high-grade gliomas of the same patient at least three time points, and the magnetic resonance image group includes T1 weighted phase, T2 weighted phase and enhanced T1 weighted phase. The specific time points of collection are before surgery, within 1 week before postoperative chemotherapy, and 4 months after the end of postoperative chemotherapy.

[0035] Magnetic resonance imaging (MRI) technology can highlight different types of tissue or lesion characteristics through different scanning sequences and parameter settings. T1-weighted phase, T2-weighted phase and enhanced T1-weighted phase are several common imaging sequences, which are of great significance in different clinical applications.

[0036] T1-weighted phase is an image acquired by setting a shorter repetition time (TR) and echo time (TE). T1 stands for "spin-lattice relaxation time", which is the time required for water molecules to return to equilibrium due to the interaction between water molecules and surrounding tissues. T1-weighted phase is usually used to display anatomical structures.

[0037] T2-weighted phases are images acquired with longer repetition times and echo times. T2 stands for "spin-spin relaxation time," which is the time it takes for energy to disappear as a result of water molecules interacting with their surroundings. T2-weighted phases are suitable for showing brain edema, tumors, inflammation, and other pathological changes.

[0038] Enhanced T1-weighted phase is based on T1-weighted imaging, using contrast agents (usually contrast agents containing gadolinium) to enhance the signal of the image to help improve the visibility of certain lesions. The contrast agent enters the blood circulation through intravenous injection and accumulates at the specific lesion site, thereby enhancing the signal at that site. Enhanced T1-weighted phase is widely used for tumor detection, location and evaluation of tumor blood supply. It can also be used to identify some other types of lesions, such as infection, spinal cord lesions, etc.

[0039] In another exemplary embodiment of the present application, step 102 of the present application calculates the volume change of the residual tumor before and after chemotherapy on the enhanced T1-weighted phases before and after chemotherapy acquired in step 101, and specifically includes the following steps 201 to 204.

[0040] Step 201: Fully automatic tumor region annotation is performed on the enhanced T1-weighted phases before and after chemotherapy using a deep learning model.

[0041] Step 202: extract the number of voxels and volume information of each voxel in the residual tumor region.

[0042] Step 203, according to the formula and , calculate the residual tumor volume before and after chemotherapy; where, is the volume of the residual tumor in the magnetic resonance imaging group after surgery and before chemotherapy, that is, the first volume, is the number of voxels contained in the residual tumor area in the magnetic resonance imaging group after surgery and before chemotherapy, is the volume of a single voxel, The volume of the residual tumor in the magnetic resonance imaging group after chemotherapy, i.e., the second volume, is the number of voxels contained in the residual tumor area in the magnetic resonance imaging group after the end of chemotherapy.

[0043] Step 204, according to the formula , calculate the volume change; where, is the volume change.

[0044] In another exemplary embodiment of the present application, in the above step 103, based on the neuro-oncology response evaluation criteria, the change in tumor volume before and after chemotherapy is compared to determine chemotherapy sensitivity, and the specific evaluation criteria are as follows.

[0045] If the tumor disappears completely after treatment, it is considered a complete response; if the tumor volume decreases by ≥65% after treatment, it is considered a partial response; if the tumor volume decreases by <65% but increases by <40%, it is considered stable disease; if the tumor volume increases by ≥40%, it is considered progressive disease. Complete response and partial response are defined as chemotherapy sensitivity, and stable disease and progressive disease are defined as chemotherapy resistance.

[0046] In another exemplary embodiment of the present application, the above step 104 extracts the imaging omics features on the pre-operative magnetic resonance imaging group, including T1 weighted phase, T2 weighted phase and enhanced T1 weighted phase, and then screens the obtained imaging omics features to screen out the differential imaging omics features of patients with different treatment responses.

[0047] Among them, the obtained radiomic features were screened to screen out the differential radiomic features of patients with different treatment responses, specifically including: for the obtained radiomic features, the maximum relevancy minimum redundancy (mRMR) algorithm was first used to identify the most discriminative feature subsets, and the top 50 features were retained. Subsequently, Spearman correlation analysis was performed to evaluate the correlation between these high-ranking features. For highly correlated features (r>0.9 and p<0.05, r is the correlation coefficient of the Spearman correlation test, and p is the statistical test value), only features with large variance and more difference information were retained. Finally, the least absolute shrinkage and selection operator (LASSO) algorithm was applied, combined with ten-fold cross validation, to further screen out the most predictive radiomic features from the remaining features to form a feature set. Exemplarily, the radiomic features selected in this step include the following 12 items.

[0048] 1. original_ngtdm_Strength_T2-T1: original neighborhood grayscale difference matrix strength_T2-T1.

[0049] 2. log-sigma-5-0-mm-3D_glszm_ZoneVariance_T1: log-sigma-5-0-mm-3D grayscale size zone matrix zone variance_T1.

[0050] 3. log-sigma-5-0-mm-3D_glszm_ZonePercentage_T1: log-sigma-5-0-mm-3D grayscale size zone matrix interval percentage_T1.

[0051] 4. wavelet-LLH_glcm_JointAverage_CE: wavelet-LLH gray-level co-occurrence matrix joint average_CE.

[0052] 5. log-sigma-5-0-mm-3D_ngtdm_Contrast_T2: log-sigma-5-0-mm-3D neighborhood grayscale difference matrix contrast_T2.

[0053] 6. wavelet-LLH_glszm_GrayLevelVariance_T2: wavelet-LLH grayscale size area matrix grayscale variance_T2.

[0054] 7. wavelet-LLH_glrlm_RunVariance_T2: wavelet-LLH grayscale run matrix run variance_T2.

[0055] 8. wavelet-LLH_glrlm_ShortRunEmphasis_CE: wavelet-LLH grayscale run-length matrix short run emphasis_CE.

[0056] 9. wavelet-LLH_glrlm_RunLengthNonUniformityNormalized_T2: wavelet-LLH grayscale run length matrix normalized run length non-uniformity_T2.

[0057] 10. wavelet-LLH_glrlm_RunLengthNonUniformity_T2: wavelet-LLH grayscale run length matrix run length non-uniformity_T2.

[0058] 11. wavelet-LLH_glrlm_RunEntropy_T2: wavelet-LLH grayscale run-length matrix run entropy_T2.

[0059] 12. wavelet-LLH_glrlm_LongRunLowGrayLevelEmphasis_T2: wavelet-LLH grayscale run matrix long run low grayscale emphasis_T2.

[0060] The suffixes T2, T1, and CE indicate that the feature is extracted from the T2-weighted phase, T1-weighted phase, and enhanced T1-weighted phase, respectively; and T2-T1 indicates that the feature is the difference between the feature values ​​extracted from the T2-weighted phase and the T1-weighted phase.

[0061] In an exemplary embodiment, in the above step 106, the machine learning model is trained using the screened imaging genomics features and their labels to form a glioma chemotherapy response prediction model, specifically as follows: steps 301 to 303.

[0062] Step 301, first divide the sample data set into a training set and a test set in a ratio of 8:2.

[0063] Step 302, using a machine learning algorithm to train a random forest, a support vector machine, an extreme gradient boosting model, and a linear regression model on the training set.

[0064] Step 303, evaluate the performance of each model on the test set, calculate its AUC (Area Under Curve), sensitivity, specificity and accuracy, and select the best performing model as the glioma chemotherapy response prediction model.

[0065] Exemplarily, after comparison, the linear regression model was the best, and the final glioma chemotherapy response prediction model was shown in the following formula.

[0066] .

[0067] in, is the chemotherapy sensitivity score. If the prediction result is positive, it is determined to be chemotherapy sensitive; otherwise, it is determined to be radiotherapy resistant. , , , , , , , , , , , They are the intensity of the original neighborhood grayscale difference matrix, the interval variance of the log-σ-5-0-mm-3D grayscale size area matrix, the interval percentage of the log-σ-5-0-mm-3D grayscale size area matrix, the joint average of the wavelet-LLH grayscale co-occurrence matrix, the contrast of the log-σ-5-0-mm-3D neighborhood grayscale difference matrix, the grayscale variance of the wavelet-LLH grayscale size area matrix, the run variance of the wavelet-LLH grayscale run matrix, the short run emphasis of the wavelet-LLH grayscale run matrix, the standardized run length unevenness of the wavelet-LLH grayscale run matrix, the run length unevenness of the wavelet-LLH grayscale run matrix, the run entropy of the wavelet-LLH grayscale run matrix, and the long run low grayscale emphasis of the wavelet-LLH grayscale run matrix.

[0068] like Figure 2 As shown, Figure 2The receiver operating characteristic curve of the glioma chemotherapy sensitivity prediction model is shown in the figure. The X-axis is "1-specificity", which indicates the proportion of negative samples that the model misclassifies as positive samples; the Y-axis is "sensitivity", which indicates the proportion of positive samples correctly identified by the model. The area under the curve of the glioma chemotherapy sensitivity prediction model is 0.977, indicating that the overall performance of the glioma chemotherapy sensitivity prediction model is excellent and suitable for high-precision classification tasks.

[0069] According to the specific embodiments provided in this application, this application has the following technical effects.

[0070] The present application example first extracts high-throughput features from the preoperative magnetic resonance images of patients with brain glioma, and on this basis screens out the most critical features for predicting the response to postoperative chemotherapy. Subsequently, these key features are used to train different machine learning models to predict the patient's response to postoperative chemotherapy.

[0071] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 3 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. 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, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for constructing a glioma chemotherapy response prediction model is implemented.

[0072] Those skilled in the art will understand that Figure 3 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.

[0073] In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0074] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0075] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0076] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0077] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0078] The technical features of the above embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0079] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for constructing a glioma chemotherapy response prediction model, characterized in that: include: Acquire a longitudinal multi-time point magnetic resonance image set; the longitudinal multi-time point magnetic resonance image set includes a plurality of magnetic resonance image sequences, and the magnetic resonance image sequence includes a magnetic resonance image group of high-grade glioma at least three time points of the same patient; The three time points were before surgery, before chemotherapy after surgery, and after chemotherapy; the magnetic resonance imaging group included one or more of the T1-weighted phase, T2-weighted phase, and enhanced T1-weighted phase; Calculating the volume change of the residual tumor in the magnetic resonance imaging group after surgery and before chemotherapy and the magnetic resonance imaging group after the end of chemotherapy in each magnetic resonance imaging sequence as the volume change corresponding to each magnetic resonance imaging sequence; Determine the chemotherapy sensitivity corresponding to each magnetic resonance imaging sequence according to the volume change corresponding to each magnetic resonance imaging sequence; Extracting and screening features of the pre-operative magnetic resonance image group in each magnetic resonance image sequence to obtain a feature set corresponding to each magnetic resonance image sequence; The feature set corresponding to each magnetic resonance imaging sequence is used as input, and the chemotherapy sensitivity corresponding to each magnetic resonance imaging sequence is used as a label to construct a sample data set; Based on the sample data set, multiple machine learning models are trained respectively, and the optimal trained machine learning model is selected as a glioma chemotherapy response prediction model.

2. The method for constructing a glioma chemotherapy response prediction model according to claim 1, characterized in that: The volume changes of the residual tumor in the magnetic resonance imaging group after surgery and before chemotherapy and the magnetic resonance imaging group after the end of chemotherapy in each magnetic resonance imaging sequence are calculated, specifically including: The volume of residual tumor in the magnetic resonance imaging group after surgery and before chemotherapy was calculated as the first volume using the following formula; ； in, is the volume of the residual tumor in the magnetic resonance imaging group after surgery and before chemotherapy, that is, the first volume, is the number of voxels contained in the residual tumor area in the magnetic resonance imaging group after surgery and before chemotherapy, is the volume of a single voxel; The volume of the residual tumor in the magnetic resonance imaging group after the completion of concurrent chemotherapy was calculated as the second volume using the following formula; ； in, is the volume of the residual tumor in the magnetic resonance imaging group after the end of chemotherapy, that is, the second volume, is the number of voxels contained in the residual tumor area in the magnetic resonance imaging group after the end of chemotherapy; Calculating a volume change according to the first volume and the second volume using the following formula; ； in, is the volume change.

3. The method for constructing a glioma chemotherapy response prediction model according to claim 2, characterized in that: The chemotherapy sensitivity corresponding to each magnetic resonance imaging sequence is determined according to the volume change corresponding to each magnetic resonance imaging sequence, specifically including: like , the chemotherapy sensitivity is determined to be chemotherapy sensitive, otherwise, the chemotherapy sensitivity is determined to be chemotherapy resistant.

4. The method for constructing a glioma chemotherapy response prediction model according to claim 2, characterized in that: Feature extraction and screening are performed on the preoperative magnetic resonance image group in each magnetic resonance image sequence to obtain a feature set corresponding to each magnetic resonance image sequence, specifically including: Extracting features of the magnetic resonance image group before surgery in the magnetic resonance image sequence to obtain imaging features of the magnetic resonance image group before surgery in the magnetic resonance image sequence; The mRMR algorithm was used to calculate the discriminability of each radiomics feature, and the radiomics features were sorted in descending order of discriminability to obtain the radiomics feature sequence; Performing correlation analysis on a preset number of radiomics features in the radiomics feature sequence, selecting radiomics features that meet preset conditions from the preset number of radiomics features in the radiomics feature sequence to form a radiomics feature candidate set; the preset conditions are: r>0.9 and p<0.05, where r is the correlation coefficient of the Spearman correlation test, and p is the statistical test value; The LASSO algorithm is used to select a second preset number of radiomics features from the radiomics feature candidate set to form a feature set.

5. The method for constructing a glioma chemotherapy response prediction model according to claim 1, characterized in that: Based on the sample data set, multiple machine learning models are trained respectively, and the best trained machine learning model is selected as a glioma chemotherapy response prediction model, specifically including: Dividing the sample data set into a training set and a test set; Using the training set to train the plurality of machine learning models respectively, to obtain a plurality of trained machine learning models; Using the test set to test multiple trained machine learning models, to obtain a test result of each trained machine learning model; the test result includes: one or more of AUC, sensitivity, specificity and accuracy; The trained machine learning model with the best test results is selected as the glioma chemotherapy response prediction model.

6. The method for constructing a glioma chemotherapy response prediction model according to claim 1 or 5, characterized in that: Multiple machine learning models include random forest model, support vector machine model, extreme gradient boosting model, and linear regression model.

7. The method for constructing a glioma chemotherapy response prediction model according to claim 1, characterized in that: The glioma chemotherapy response prediction model is: ； in, is the chemotherapy sensitivity score. If the prediction result is positive, it is determined to be chemotherapy sensitive; otherwise, it is determined to be chemotherapy resistant. , , , , , , , , , , , They are the intensity of the original neighborhood grayscale difference matrix, the interval variance of the log-σ-5-0-mm-3D grayscale size area matrix, the interval percentage of the log-σ-5-0-mm-3D grayscale size area matrix, the joint average of the wavelet-LLH grayscale co-occurrence matrix, the contrast of the log-σ-5-0-mm-3D neighborhood grayscale difference matrix, the grayscale variance of the wavelet-LLH grayscale size area matrix, the run variance of the wavelet-LLH grayscale run matrix, the short run emphasis of the wavelet-LLH grayscale run matrix, the standardized run length unevenness of the wavelet-LLH grayscale run matrix, the run length unevenness of the wavelet-LLH grayscale run matrix, the run entropy of the wavelet-LLH grayscale run matrix, and the long run low grayscale emphasis of the wavelet-LLH grayscale run matrix.

8. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method for constructing a glioma chemotherapy response prediction model as described in any one of claims 1-7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for constructing a glioma chemotherapy response prediction model described in any one of claims 1 to 7 is implemented.

Images