Construction method and application of CMS prediction model

Abstract

The invention relates to the technical field of medical data processing, and discloses a construction method and application of a CMS prediction model, and the method comprises the steps: obtaining a damage map of a CNS tumor patient group and a voxel value of each individual by using a preoperative MRI image and a tumor mask of the CNS tumor patient group with / without CMS after a child posterior fossal tumor resection operation; the method comprises the following steps of: selecting voxel value data and gender data, screening out significant variable data significantly related to CMS through a regression analysis method, verifying the correlation intensity of the voxel value data and the gender data to appearing / non-appearing CMS, and training to obtain a high-precision CMS prediction model by using the significant variable data of a target population related to the CMS and labels of appearing / non-appearing CMS on an MRI image. The method can assist in accurately evaluating whether a CMS risk occurs after a CNS tumor patient accepts a child posterior fossa tumor resection operation, and is beneficial for doctors and parents to prepare good emotional support and directional rehabilitation plans in advance.

Description

Technical Field

[0001] The present invention relates to the technical field of medical data processing, and in particular to a method for constructing a CMS (cerebellar aphasia syndrome) prediction model and its application. Background Art

[0002] Central nervous system (CNS) tumors are the most common solid tumors in children and adolescents and the leading cause of cancer death in children. Among childhood CNS tumors, approximately 45% to 60% occur in the posterior fossa. Cerebellar aphasia syndrome (CMS) is the most common complication after surgical resection of posterior fossa tumors in children. Typical features of CMS include language disorders, motor dysfunction, emotional instability, and cognitive impairment, and the long-term side effects of CMS have a significant impact on children's quality of life.

[0003] To date, the mechanism of CMS remains unclear. Previous studies have used midline location and medulloblastoma (MB) as risk factors for assessing CMS, while the role of gender, age, hydrocephalus, tumor size, and surgical approach remains controversial. Some studies have proposed a subtemporal approach to reduce the risk of CMS, but European studies have found conflicting evidence. As a result, some studies have begun to use risk factors to establish a CMS prediction model. Some studies have used preoperative MRI diagnosis of MB, midline tumor location, tumor invasion of the middle cerebral peduncle, and tumor invasion of the superior cerebral peduncle as risk factors for predicting CMS. Some studies have used preoperative age, gender, and Evans index as risk factors, but the CMS prediction models they obtained have low accuracy in predicting CMS and have failed to truly assist in CMS assessment.

[0004] Therefore, a new method for constructing CMS prediction model is urgently needed. Summary of the invention

[0005] The present invention provides a method for constructing a CMS prediction model and application thereof, so as to solve the defect that the existing CMS prediction model has low accuracy in predicting CMS.

[0006] The present invention provides a method for constructing a CMS prediction model, comprising: Obtaining MRI images of a target population, tumor masks obtained based on the MRI images, and clinical variable data, wherein the target population includes a CNS tumor patient group that develops CMS after undergoing posterior cranial fossa tumor resection surgery for children and a CNS tumor patient group that does not develop CMS after undergoing posterior cranial fossa tumor resection surgery for children, and the MRI images are MRI images of the target population before undergoing posterior cranial fossa tumor resection surgery for children, and the MRI images are marked with labels of developing CMS after undergoing posterior cranial fossa tumor resection surgery for children and not developing CMS after undergoing posterior cranial fossa tumor resection surgery for children; According to the labels of CMS after pediatric posterior fossa tumor resection surgery and CMS not after pediatric posterior fossa tumor resection surgery on the MRI images of the target population and the tumor mask obtained based on the MRI images, the damage map of the target population and the voxel value of each individual in the target population are obtained, wherein the damage map reflects the distribution of voxels related to CMS and the strength of the correlation between each voxel and CMS, and the voxel value indicates the degree of influence of the CNS tumor location on whether CMS occurs in the CNS tumor patient group after pediatric posterior fossa tumor resection surgery; According to the clinical variable data of the target population and the voxel value of each individual in the target population, the significant variable data related to CMS were obtained through regression analysis; A CMS prediction model was trained based on the significant variable data related to CMS in the target population and the labels of children who developed CMS after undergoing posterior fossa tumor resection surgery and those who did not develop CMS after undergoing posterior fossa tumor resection surgery on MRI images.

[0007] In one embodiment, the clinical variable data include data of any one of the following indicators or any combination thereof: gender, age at surgery, tumor size, tumor texture, whether there is hydrocephalus, whether there is paraventricular edema, whether there is preoperative VP shunt, extent of resection (EOR), surgical approach, and pathology.

[0008] In one embodiment, the method of obtaining the lesion map of the target population and the voxel value of each individual in the target population based on the labels of the children who developed CMS after receiving posterior cranial fossa tumor resection surgery and the children who did not develop CMS after receiving posterior cranial fossa tumor resection surgery on the MRI images of the target population and the tumor mask obtained based on the MRI images includes: Aligning the MRI images of the target population and the tumor masks obtained based on the MRI images to the MNI standard space to obtain standardized MRI images and standardized tumor masks; Based on the labels of children with CMS and children without CMS after posterior fossa tumor resection on standardized MRI images and standardized tumor masks, a sparse canonical correlation analysis method was used to obtain a CMS-related lesion map, wherein the weight of each voxel in the CMS-related lesion map represents the strength of the correlation between the voxel and CMS; The standardized tumor mask of each individual in the target population is converted into a matrix M, and the lesion map associated with CMS is converted into a coefficient matrix S, wherein the coefficient matrix S reflects the voxel distribution associated with CMS and the strength of the correlation between each voxel and CMS; According to the matrix M and the coefficient matrix S, the voxel value of each individual in the target population is obtained.

[0009] In one embodiment, obtaining the voxel value of each individual in the target population according to the matrix M and the coefficient matrix S includes: The matrix M corresponding to the standardized tumor mask of each individual in the target population is multiplied by the coefficient matrix S to obtain the voxel value of each individual in the target population.

[0010] In one embodiment, the significant variable data related to CMS are obtained by regression analysis based on the clinical variable data of the target population and the voxel value of each individual in the target population, including: Based on the clinical variable data of the target population and the voxel values ​​of each individual in the target population, the H test was performed on the continuous variables and the chi-square test was performed on the categorical variables to obtain the preliminary significant variable data related to CMS; Using the Lasso regression analysis method, regression analysis and 10-fold cross-validation were performed on the preliminary significant variable data related to CMS to obtain the final significant variable data related to CMS, wherein the final significant variable data related to CMS included voxel value data and gender data.

[0011] In one embodiment, the CMS prediction model is trained based on the significant variable data related to CMS in the target population and the labels of CMS after pediatric posterior fossa tumor resection surgery and CMS not after pediatric posterior fossa tumor resection surgery on MRI images, including: Based on the final significant variable data related to CMS in the target population and the labels of children who developed CMS after undergoing posterior fossa tumor resection surgery and those who did not develop CMS after undergoing posterior fossa tumor resection surgery on MRI images, a generalized linear model was used to train an initial CMS prediction model; The AUC value of the ROC curve was used to detect the prediction performance of the initial CMS prediction model and optimize it to obtain the final CMS prediction model.

[0012] The present invention also provides a CMS prediction model construction system, comprising: A data acquisition module is used to: acquire MRI images of a target population, tumor masks obtained based on the MRI images, and clinical variable data, wherein the target population includes a CNS tumor patient group that develops CMS after undergoing posterior cranial fossa tumor resection surgery for children and a CNS tumor patient group that does not develop CMS after undergoing posterior cranial fossa tumor resection surgery for children, and the MRI images are MRI images of the target population before undergoing posterior cranial fossa tumor resection surgery for children, and the MRI images are marked with labels of CMS after undergoing posterior cranial fossa tumor resection surgery for children and CMS not developing after undergoing posterior cranial fossa tumor resection surgery for children; A voxel value acquisition module is used to: obtain a damage map of the target population and a voxel value of each individual in the target population based on the labels of CMS after pediatric posterior fossa tumor resection surgery and non-CMS after pediatric posterior fossa tumor resection surgery on the MRI images of the target population and a tumor mask obtained based on the MRI images, wherein the damage map reflects the distribution of voxels related to CMS and the strength of correlation between each voxel and CMS, and the voxel value indicates the degree of influence of the CNS tumor location on whether CMS occurs in the CNS tumor patient group after pediatric posterior fossa tumor resection surgery; The regression analysis module is used to obtain significant variable data related to CMS through regression analysis based on the clinical variable data of the target population and the voxel value of each individual in the target population; A module is constructed for training a CMS prediction model based on the significant variable data related to CMS in the target population and the labels of CMS and non-CMS after pediatric posterior fossa tumor resection surgery on MRI images.

[0013] The present invention also provides a CMS auxiliary evaluation system, comprising: The data receiving module is used to: receive an MRI image and gender data of a subject to be tested sent from at least one terminal, wherein the subject to be tested is a CNS tumor patient who is about to undergo a pediatric posterior cranial fossa tumor resection surgery; A prediction module is used to: obtain a CMS prediction result of the subject according to the MRI image and gender data of the subject by using a CMS prediction model obtained by any of the above-mentioned methods for constructing a CMS prediction model, wherein the CMS prediction result indicates whether the subject will or will not develop CMS after undergoing pediatric posterior cranial fossa tumor resection surgery; The data output module is used to send the CMS prediction result of the subject to be tested back to at least one terminal.

[0014] The present invention also provides an electronic device, comprising a processor and a memory storing a computer program, wherein the processor implements any of the above-mentioned methods for constructing a CMS prediction model when executing the computer program.

[0015] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-described methods for constructing a CMS prediction model.

[0016] The present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute any of the above-mentioned methods for constructing a CMS prediction model.

[0017] The present invention provides a method for constructing a CMS prediction model and its application. Preoperative MRI images of a group of CNS tumor patients who developed / did not develop CMS after undergoing posterior cranial fossa tumor resection surgery in children and tumor masks obtained based on the MRI images are used to obtain a CMS-related injury map of the CNS tumor patient group and a voxel value of each individual in the CNS tumor patient group. Regression analysis is used to screen out significant variable data significantly correlated with CMS from clinical variable data and voxel value data, verify the correlation strength of voxel value data and gender data on the development / non-development of CMS after undergoing posterior cranial fossa tumor resection surgery in children, and use the significant variable data related to CMS in the target population and the labels of the development / non-development of CMS after undergoing posterior cranial fossa tumor resection surgery in children on MRI images to train a high-precision CMS prediction model. The model can assist in accurately assessing the risk of CMS in CNS tumor patients after undergoing posterior cranial fossa tumor resection surgery in children, is helpful for doctors and parents to prepare good emotional support and targeted rehabilitation plans in advance, and is of great benefit to children with CMS. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0019] Figure 1 A schematic diagram of a process for constructing a CMS prediction model provided by the present invention.

[0020] Figure 2 Figure 2 is an example of a damage map associated with CMS.

[0021] Figure 3 A histogram of non-zero voxel values ​​is shown, voxel value: voxel value, frequency: frequency, which can be understood as the weight of the voxel value in this embodiment.

[0022] Figure 4 The C index (AUC), sensitivity, and specificity of the training set, validation set, and prospective set under the ROC curve are shown.

[0023] Figure 5 The calibration curve for predicting CMS risk is shown.

[0024] Figure 6 A schematic diagram of the structure of a CMS-assisted evaluation system provided by the present invention.

[0025] Figure 7 This is a schematic structural diagram of an electronic device provided by the present invention. DETAILED DESCRIPTION

[0026] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings in the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments, and they should not be understood as limitations on the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. In the description of the present invention, it should be understood that the terms used are only for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0027] Combine the following Figure 1-Figure 7 The method for constructing a CMS prediction model provided by the present invention and its application are described. It should be noted that the execution subject of the method for constructing a CMS prediction model provided by the present invention can be any network side device / terminal side device that meets the technical requirements, such as a CMS prediction model construction device.

[0028] Figure 1 Schematic diagram of the process of constructing the CMS prediction model provided by the present invention. Figure 1 The present invention provides a method for constructing a CMS prediction model, which may include: Step S110, obtaining MRI images of the target population, tumor masks obtained based on the MRI images, and clinical variable data, wherein the target population includes a group of CNS tumor patients who developed CMS after undergoing posterior cranial fossa tumor resection surgery for children and a group of CNS tumor patients who did not develop CMS after undergoing posterior cranial fossa tumor resection surgery for children, and the MRI images are MRI images of the target population before undergoing posterior cranial fossa tumor resection surgery for children and the MRI images are marked with labels of CMS after undergoing posterior cranial fossa tumor resection surgery for children and CMS not occurring after undergoing posterior cranial fossa tumor resection surgery for children, and the clinical variable data include data of the following indicators: gender, age at surgery, tumor size, tumor texture, whether there is hydrocephalus (hydrocephalus is defined as an Evans index exceeding 0.3), whether there is paraventricular edema, whether there is preoperative VP shunt, extent of resection (EOR), surgical approach, and pathology.

[0029] Step S120, based on the labels of CMS occurring after pediatric posterior fossa tumor resection surgery and CMS not occurring after pediatric posterior fossa tumor resection surgery on the MRI images of the target population and the tumor mask obtained based on the MRI images, obtain a damage map of the target population and the voxel value of each individual in the target population, wherein the damage map reflects the voxel distribution related to CMS and the strength of the correlation between each voxel and CMS, and the voxel value indicates the degree of influence of the CNS tumor location on whether CMS occurs in the CNS tumor patient group after pediatric posterior fossa tumor resection surgery.

[0030] In one embodiment, step S120 may include: Aligning the MRI images of the target population and the tumor masks obtained based on the MRI images to the MNI standard space to obtain standardized MRI images and standardized tumor masks; Based on the labels of CMS and non-CMS after posterior fossa tumor resection in children on standardized MRI images and standardized tumor masks, sparse canonical correlation analysis was used to obtain a lesion map associated with CMS (see the example figure for details). Figure 2 ), where the weight of each voxel in the CMS-related damage map represents the strength of the correlation between the voxel and CMS; The standardized tumor mask of each individual in the target population is converted into a matrix M, and the lesion map associated with CMS is converted into a coefficient matrix S, where the coefficient matrix S reflects the voxel distribution associated with CMS and the strength of the correlation between each voxel and CMS. Figure 3 The weight values ​​(Frequency) of non-zero voxel values ​​are shown; The matrix M corresponding to the standardized tumor mask of each individual in the target population is multiplied by the coefficient matrix S to obtain the voxel value of each individual in the target population.

[0031] Step S130: Obtain significant variable data related to CMS through regression analysis based on the clinical variable data of the target population and the voxel value of each individual in the target population.

[0032] In one embodiment, step S130 may include: Based on the clinical variable data of the target population and the voxel values ​​of each individual in the target population, the H test was performed on the continuous variables and the chi-square test was performed on the categorical variables to obtain the preliminary significant variable data related to CMS; Using the Lasso regression analysis method, regression analysis and 10-fold cross-validation were performed on the preliminary significant variable data related to CMS to obtain the final significant variable data related to CMS, wherein the final significant variable data related to CMS included voxel value data and gender data.

[0033] Step S140, training a CMS prediction model based on the significant variable data related to CMS in the target population and the labels of CMS and non-CMS after pediatric posterior fossa tumor resection surgery on MRI images.

[0034] In one embodiment, step S140 can use a generalized linear model to train an initial CMS prediction model based on the final significant variable data related to CMS in the target population and the labels of CMS and non-CMS in children who underwent posterior cranial fossa tumor resection surgery on MRI images; then use the AUC value of the ROC curve to detect the prediction performance of the initial CMS prediction model and optimize it to obtain the final CMS prediction model.

[0035] The target population of this example study was patients who were diagnosed with posterior fossa tumors and received surgical treatment at Beijing Children's Hospital from July 2013 to December 2023. The patients were divided into a retrospective cohort (July 2013 to August 2020) and a prospective cohort (September 2020 to December 2023). The inclusion criteria of the study included: (1) age between 0 and 18 years; (2) diagnosis of posterior fossa tumors; (3) posterior fossa tumor resection surgery; (4) confirmed information of CMS or non-CMS after posterior fossa tumor resection surgery. Exclusion criteria included: (1) incomplete medical data; (2) lack of preoperative MRI images or records; (3) unsatisfactory standardization during visual inspection. According to the CMS status of the target population (CMS or non-CMS after posterior fossa tumor resection surgery), the patients in each cohort were divided into a CMS group or a non-CMS group.

[0036] Regarding the acquisition of preoperative MRI images of the target groups, the patients in the retrospective cohort were scanned on a 3.0T Discovery 750 scanner (GE Healthcare, Milwaukee, WI) using an 8-channel head coil (voxel size, 0.6 × 0.6 × 5 mm3), while the patients in the prospective cohort were scanned on a 3.0T Ingenia CX scanner (Philips Healthcare, Best, The Netherlands) using a 32-channel head coil (voxel size, 1 × 1 × 1 mm3).

[0037] The tumor mask based on MRI images can be manually drawn by a senior neurosurgeon on ITK-SNAP (https: / / www.itksnap.org) (T2W images and FLAIR images can be referred to during the drawing of the tumor mask). The tumor mask is a normal T1W image that does not include the surrounding edema area. Then, all tumor masks can be checked by a senior radiologist. Marking the CMS status label on the MRI image can also be understood as marking the CMS status label on the tumor mask. The marking of the CMS status label can be double-blinded by neurosurgeons and radiologists to reduce human bias.

[0038] MRI images and tumor masks were then aligned to the Montreal Neurological Institute (MNI) space in Statistical Parametric Mapping (SPM12, Wellcome Department of Neuroscience, London, UK; https: / / www.fl.ion.ucl.ac.uk / spm / software / spm12 / ) using the Clinical Toolbox (https: / / www.nitrc.org / projects / clinicaltbx) to standardize the MRI images and tumor masks. Lesion maps for each individual in the target population were generated using sparse canonical correlation analysis in (LESYMAP) in R, where the standardized tumor masks and MRI images (including CMS status labels) were the inputs to LESYMAP. Example images of lesion maps can be found in Figure 2 ,Detailed information of the damage map can be found in Table 1.

[0039] Table 1 Voxel coordinate information of the lesion atlas in the MNI standard system

[0040] The tumor mask of each patient was then converted into a matrix M represented by Matlab, and the lesion map was converted into a coefficient matrix S. The coefficient matrix S reflects the strength of the correlation between the corresponding voxels and CMS, as well as the distribution of voxels associated with CMS. The voxel value represents the degree of influence of the CNS tumor location on whether CMS occurs after pediatric posterior fossa tumor resection in the CNS tumor patient population. The matrix M of each patient was then multiplied by the coefficient matrix S of the patient population to obtain the voxel value of each patient.

[0041] This example defines CMS as aphasia or severely reduced speech after surgery. Other symptoms may occur simultaneously, including hypotonia, dysphagia, irritability, and involuntary movements. This example follows up all patients at least once and records the duration of aphasia to further confirm the diagnosis of CMS. The clinical variables of this example include: gender, age at surgery, tumor size, tumor texture, whether there is hydrocephalus, whether there is paraventricular edema, whether there is preoperative VP shunt, extent of resection (EOR), surgical approach, and pathology.

[0042] The Kruskal-Willis test (H test) was used to compare the continuous variables in the clinical variable data and voxel value data of the target group, and the chi-square test was used for the categorical variables to obtain preliminary significant variable data. P < 0.05 was considered to be the standard of statistical significance.

[0043] Then, Lasso regression analysis was used to perform regression analysis and 10-fold cross validation on the included variables (preliminary significant variable data) to achieve centering and standardization, and the final significant variable data, including body value data and gender data, were obtained. The CMS prediction model was then fitted using a generalized linear model, which associated the CMS status label with the final significant variable data (i.e., body value data and gender data). Finally, the AUC value of the ROC curve can be used to evaluate the prediction performance of the CMS prediction model and optimize its performance to obtain the final CMS prediction model.

[0044] During the data statistics process, a total of 197 patients were included in the study of this example. Among them, 102 were in the training set, 43 were in the validation set, and 52 were in the prospective set. There was no significant difference in CMS rate (training group: 32.4%, validation group: 32.6%, prospective group: 38.5%, p=0.731), median age (training group: 5.0 [3.2, 7.8] years, validation group: 5.2 [3.0, 8.1] years, prospective group: 4.6 There were no significant differences in the age (2.4, 7.0) years, p = 0.297), male sex (training group: 57.8%, validation group: 62.8%, prospective group: 63.5%, p = 0.746), tumor size > 5 mm (training group: 47.1%, validation group: 44.2%, prospective group: 34.6%, p = 0.333), solid tumors (training group: 76.5%, validation group: 81.4%, prospective group: 82.7%, p = 0.615), hydrocephalus (training group: 46.1%, validation group: 55.8%, prospective group: 48.1%, p = 0.560), paraventricular edema (training group: 61.8%, validation group: 67.4%, prospective group: 59.6%, p = 0.720), and preoperative VP shunt (training group: 9.8%, validation group: 14.0%, prospective group: 9.6%, p = 0.730). However, significant differences were observed in terms of surgical route (p < 0.001).

[0045] There were no significant differences in tumor size >5 mm (non-CMS: 42.3% vs CMS: 44.8%, p = 0.740), preoperative VP shunt rate (non-CMS: 9.2% vs CMS: 13.4%, p = 0.365), and gross total resection rate (non-CMS: 89.2% vs CMS: 85.1%, p = 0.398) between the CMS and non-CMS groups. Male gender (non-CMS: 52.3% vs CMS: 76.1%, p = 0.001), solid tumors (non-CMS: 70.8% vs CMS: 95.5%, p < 0.001), presence of hydrocephalus (non-CMS: 49.8% vs CMS: 64.2%, p = 0.001), and periventricular edema (non-CMS: 55.4% vs CMS: 76.1%, p = 0.004) showed significant differences between the two groups. Surgical route (p < 0.001) and pathology (p < 0.001) also showed significant differences between the groups.

[0046] Feature selection was performed in the training set using the Lasso regression method. All 11 variables were initially included (sex, age at surgery, tumor size, tumor texture, presence of hydrocephalus, presence of paraventricular edema, presence of preoperative VP shunt, extent of resection (EOR), surgical approach, pathology, and voxel value), and finally 2 potential predictors with non-zero coefficients were selected in the Lasso regression analysis, namely sex and voxel value.

[0047] In this example, a CMS prediction model was constructed using a logistic regression model using gender and voxel values. The C indexes of the training group, validation group, and prospective group were 0.956, 0.933, and 0.892, respectively, indicating that the model had quite satisfactory performance ( Figure 3 The calibration curves used to predict CMS risk also showed good agreement ( Figure 4 ).

[0048] Compared with previous studies that divided tumors into midline and lateral regions, the CMS prediction model obtained in this example can more accurately describe the location and extent of the tumor using tumor voxel information. With the inclusion of clinical variables, the CMS prediction model significantly improved in predictive performance, indicating that tumor location and other clinical factors contribute independently to CMS.

[0049] In order to combine the injury map and clinical factors, the present embodiment converts the injury map into a coefficient matrix S. Then, regression analysis is used to select significantly correlated variables from other clinical variables and voxel values. Afterwards, a CMS prediction model is established based on the selected significantly correlated variables. This method can ensure that the included variables are independently correlated with CMS, and ultimately generate a simple but efficient CMS prediction model.

[0050] Postoperative CMS remains the most notable challenge for pediatric patients with CNS tumors. Children with CMS often suffer significant impairment in social skills and education due to neurological deficits, emotional and cognitive dysfunction. The long-term quality of life of these children remains a concern. Early intervention and speech therapy have been reported to be effective in minimizing long-term speech and language deficits in patients with postoperative CMS. Therefore, assessing the risk of CMS in patients with posterior fossa tumors is crucial in clinical practice.

[0051] The present invention provides a method for constructing a CMS prediction model and its application. The method uses preoperative MRI images of a group of CNS tumor patients who develop / do not develop CMS after undergoing posterior cranial fossa tumor resection surgery in children and tumor masks obtained based on the MRI images to obtain the injury map and voxel value of each individual in the CNS tumor patient group, and uses regression analysis to screen out significant variable data significantly correlated with CMS from clinical variable data and voxel value data, verify the correlation strength of voxel value data and gender data on the development / non-development of CMS after undergoing posterior cranial fossa tumor resection surgery in children, and use the significant variable data related to CMS in the target population and the labels of the development / non-development of CMS after undergoing posterior cranial fossa tumor resection surgery in children on MRI images to train a high-precision CMS prediction model. The model can assist in accurately assessing the risk of CMS in CNS tumor patients after undergoing posterior cranial fossa tumor resection surgery in children, is helpful for doctors and parents to prepare good emotional support and targeted rehabilitation plans in advance, and is of great benefit to children with CMS.

[0052] The CMS assisted evaluation system provided by the present invention is described below.

[0053] Reference Figure 5 The present invention provides a CMS auxiliary evaluation system, which may include: The data receiving module is used to: receive an MRI image and gender data of a subject to be tested sent from at least one terminal, wherein the subject to be tested is a CNS tumor patient who is about to undergo a pediatric posterior cranial fossa tumor resection surgery; A prediction module is used to: obtain a CMS prediction result of the subject according to the MRI image and gender data of the subject by using a CMS prediction model obtained by any of the above-mentioned methods for constructing a CMS prediction model, wherein the CMS prediction result indicates whether the subject will or will not develop CMS after undergoing pediatric posterior cranial fossa tumor resection surgery; The data output module is used to send the CMS prediction result of the subject to be tested back to at least one terminal.

[0054] Specifically, after obtaining the preoperative MRI image of a CNS tumor patient who is about to undergo posterior cranial fossa tumor resection surgery, the doctor can use the terminal to input the preoperative MRI image and gender data of the patient into the CMS auxiliary evaluation system. The data receiving module receives the MRI image and gender data of the patient, and the prediction module uses the CMS prediction model to obtain the CMS prediction result of the patient. The data output module then sends the CMS prediction result of the patient back to the terminal. The doctor can obtain the prediction result of whether the patient will or will not develop CMS after undergoing posterior cranial fossa tumor resection surgery on the terminal, which helps to formulate a better treatment plan for the patient.

[0055] Figure 6An example of a physical structure diagram of an electronic device is shown in FIG. Figure 4 As shown, the electronic device may include: a processor 810, a communication interface 820, a memory 830 and a communication bus 840, wherein the processor 810, the communication interface 820 and the memory 830 communicate with each other through the communication bus 840. The processor 810 may call the logic instructions in the memory 830 to perform the following steps: Receiving an MRI image and gender data of a subject sent from at least one terminal, wherein the subject is a CNS tumor patient who is about to undergo a pediatric posterior cranial fossa tumor resection surgery; According to the MRI image and gender data of the subject, a CMS prediction model is obtained by any of the above-mentioned methods for constructing a CMS prediction model to obtain a CMS prediction result of the subject, wherein the CMS prediction result indicates whether the subject will or will not develop CMS after undergoing pediatric posterior cranial fossa tumor resection surgery; The CMS prediction result of the subject to be tested is sent back to at least one terminal.

[0056] In addition, the logic instructions in the above-mentioned memory 830 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0057] On the other hand, the present invention further provides a computer program product, the computer program product comprising a computer program, the computer program can be stored on a non-transitory computer readable storage medium, and when the computer program is executed by a processor, the computer can perform the following steps: Receiving an MRI image and gender data of a subject sent from at least one terminal, wherein the subject is a CNS tumor patient who is about to undergo a pediatric posterior cranial fossa tumor resection surgery; According to the MRI image and gender data of the subject, a CMS prediction model is obtained by any of the above-mentioned methods for constructing a CMS prediction model to obtain a CMS prediction result of the subject, wherein the CMS prediction result indicates whether the subject will or will not develop CMS after undergoing pediatric posterior cranial fossa tumor resection surgery; The CMS prediction result of the subject to be tested is sent back to at least one terminal.

[0058] In another aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which is implemented to perform the following steps when executed by a processor: Receiving an MRI image and gender data of a subject sent from at least one terminal, wherein the subject is a CNS tumor patient who is about to undergo a pediatric posterior cranial fossa tumor resection surgery; According to the MRI image and gender data of the subject, a CMS prediction model is obtained by any of the above-mentioned methods for constructing a CMS prediction model to obtain a CMS prediction result of the subject, wherein the CMS prediction result indicates whether the subject will or will not develop CMS after undergoing pediatric posterior cranial fossa tumor resection surgery; The CMS prediction result of the subject to be tested is sent back to at least one terminal.

[0059] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0060] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0061] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for constructing a CMS prediction model, characterized in that: include: Obtaining MRI images of a target population, tumor masks obtained based on the MRI images, and clinical variable data, wherein the target population includes a CNS tumor patient group that develops CMS after undergoing posterior cranial fossa tumor resection surgery for children and a CNS tumor patient group that does not develop CMS after undergoing posterior cranial fossa tumor resection surgery for children, and the MRI images are MRI images of the target population before undergoing posterior cranial fossa tumor resection surgery for children, and the MRI images are marked with labels of developing CMS after undergoing posterior cranial fossa tumor resection surgery for children and not developing CMS after undergoing posterior cranial fossa tumor resection surgery for children; According to the labels of CMS after pediatric posterior fossa tumor resection surgery and CMS not after pediatric posterior fossa tumor resection surgery on the MRI images of the target population and the tumor mask obtained based on the MRI images, the damage map of the target population and the voxel value of each individual in the target population are obtained, wherein the damage map reflects the distribution of voxels related to CMS and the strength of the correlation between each voxel and CMS, and the voxel value indicates the degree of influence of the CNS tumor location on whether CMS occurs in the CNS tumor patient group after pediatric posterior fossa tumor resection surgery; According to the clinical variable data of the target population and the voxel value of each individual in the target population, the significant variable data related to CMS were obtained through regression analysis; A CMS prediction model was trained based on the significant variable data related to CMS in the target population and the labels of children who developed CMS after undergoing posterior fossa tumor resection surgery and those who did not develop CMS after undergoing posterior fossa tumor resection surgery on MRI images.

2. The method for constructing a CMS prediction model according to claim 1, characterized in that: Clinical variable data included any one of the following indicators or any combination thereof: gender, age at surgery, tumor size, tumor texture, presence of hydrocephalus, presence of paraventricular edema, presence of preoperative VP shunt, extent of resection, surgical approach, and pathology.

3. The method for constructing a CMS prediction model according to claim 2, characterized in that: The method of obtaining the lesion map of the target population and the voxel value of each individual in the target population based on the labels of the children who developed CMS after receiving posterior cranial fossa tumor resection surgery and the children who did not develop CMS after receiving posterior cranial fossa tumor resection surgery on the MRI images of the target population and the tumor mask obtained based on the MRI images includes: Aligning the MRI images of the target population and the tumor masks obtained based on the MRI images to the MNI standard space to obtain standardized MRI images and standardized tumor masks; Based on the labels of children with CMS and children without CMS after posterior fossa tumor resection on standardized MRI images and standardized tumor masks, a sparse canonical correlation analysis method was used to obtain a CMS-related lesion map, wherein the weight of each voxel in the CMS-related lesion map represents the strength of the correlation between the voxel and CMS; The standardized tumor mask of each individual in the target population is converted into a matrix M, and the lesion map associated with CMS is converted into a coefficient matrix S, wherein the coefficient matrix S reflects the voxel distribution associated with CMS and the strength of the correlation between each voxel and CMS; According to the matrix M and the coefficient matrix S, the voxel value of each individual in the target population is obtained.

4. The method for constructing a CMS prediction model according to claim 3, characterized in that: The voxel value of each individual in the target population is obtained according to the matrix M and the coefficient matrix S, including: The matrix M corresponding to the standardized tumor mask of each individual in the target population is multiplied by the coefficient matrix S to obtain the voxel value of each individual in the target population.

5. The method for constructing a CMS prediction model according to claim 4, characterized in that: The significant variable data related to CMS are obtained by regression analysis based on the clinical variable data of the target population and the voxel value of each individual in the target population, including: Based on the clinical variable data of the target population and the voxel values ​​of each individual in the target population, the H test was performed on the continuous variables and the chi-square test was performed on the categorical variables to obtain the preliminary significant variable data related to CMS; Using the Lasso regression analysis method, regression analysis and 10-fold cross-validation were performed on the preliminary significant variable data related to CMS to obtain the final significant variable data related to CMS, wherein the final significant variable data related to CMS included voxel value data and gender data.

6. The method for constructing a CMS prediction model according to claim 5, characterized in that: The CMS prediction model is trained based on the significant variable data related to CMS in the target population and the labels of CMS and non-CMS after posterior cranial fossa tumor resection surgery in children on MRI images, including: Based on the final significant variable data related to CMS in the target population and the labels of children who developed CMS after undergoing posterior fossa tumor resection surgery and those who did not develop CMS after undergoing posterior fossa tumor resection surgery on MRI images, a generalized linear model was used to train an initial CMS prediction model; The AUC value of the ROC curve was used to detect the prediction performance of the initial CMS prediction model and optimize it to obtain the final CMS prediction model.

7. A system for constructing a CMS prediction model, characterized in that: include: A data acquisition module is used to: acquire MRI images of a target population, tumor masks obtained based on the MRI images, and clinical variable data, wherein the target population includes a CNS tumor patient group that develops CMS after undergoing posterior cranial fossa tumor resection surgery for children and a CNS tumor patient group that does not develop CMS after undergoing posterior cranial fossa tumor resection surgery for children, and the MRI images are MRI images of the target population before undergoing posterior cranial fossa tumor resection surgery for children, and the MRI images are marked with labels of CMS after undergoing posterior cranial fossa tumor resection surgery for children and CMS not developing after undergoing posterior cranial fossa tumor resection surgery for children; A voxel value acquisition module is used to: obtain a damage map of the target population and a voxel value of each individual in the target population based on the labels of CMS after pediatric posterior fossa tumor resection surgery and non-CMS after pediatric posterior fossa tumor resection surgery on the MRI images of the target population and a tumor mask obtained based on the MRI images, wherein the damage map reflects the distribution of voxels related to CMS and the strength of correlation between each voxel and CMS, and the voxel value indicates the degree of influence of the CNS tumor location on whether CMS occurs in the CNS tumor patient group after pediatric posterior fossa tumor resection surgery; The regression analysis module is used to obtain significant variable data related to CMS through regression analysis based on the clinical variable data of the target population and the voxel value of each individual in the target population; A module is constructed for training a CMS prediction model based on the significant variable data related to CMS in the target population and the labels of CMS and non-CMS after pediatric posterior fossa tumor resection surgery on MRI images.

8. A CMS-assisted evaluation system, characterized in that: include: The data receiving module is used to: receive an MRI image and gender data of a subject to be tested sent from at least one terminal, wherein the subject to be tested is a CNS tumor patient who is about to undergo a pediatric posterior cranial fossa tumor resection surgery; A prediction module, used to: obtain a CMS prediction result of the subject according to the MRI image and gender data of the subject by using a CMS prediction model obtained by the method for constructing a CMS prediction model according to any one of claims 1 to 6, wherein the CMS prediction result indicates whether the subject will or will not develop CMS after undergoing pediatric posterior cranial fossa tumor resection surgery; The data output module is used to send the CMS prediction result of the subject to be tested back to at least one terminal.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: The processor performs the following steps: Receiving an MRI image and gender data of a subject sent from at least one terminal, wherein the subject is a CNS tumor patient who is about to undergo a pediatric posterior cranial fossa tumor resection surgery; According to the MRI image and gender data of the subject, a CMS prediction model is obtained by the method for constructing a CMS prediction model according to any one of claims 1 to 6, to obtain a CMS prediction result of the subject, wherein the CMS prediction result indicates whether the subject will or will not develop CMS after undergoing pediatric posterior cranial fossa tumor resection surgery; The CMS prediction result of the subject to be tested is sent back to at least one terminal.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: The computer program is executed by the processor to perform the following steps: Receiving an MRI image and gender data of a subject sent from at least one terminal, wherein the subject is a CNS tumor patient who is about to undergo a pediatric posterior cranial fossa tumor resection surgery; According to the MRI image and gender data of the subject, a CMS prediction model is obtained by the method for constructing a CMS prediction model according to any one of claims 1 to 6, to obtain a CMS prediction result of the subject, wherein the CMS prediction result indicates whether the subject will or will not develop CMS after undergoing pediatric posterior cranial fossa tumor resection surgery; The CMS prediction result of the subject to be tested is sent back to at least one terminal.

Images