Cesarean section scar pregnancy prognosis outcome classification prediction model establishment method and device

Abstract

The invention discloses a caesarean section scar pregnancy prognosis outcome classification prediction model establishment method and device. The caesarean section scar pregnancy prognosis outcome classification prediction model establishment method comprises the following steps: acquiring a patient database, wherein the patient database comprises at least one piece of patient data; classifying the patient data in the patient database according to a treatment mode and a prognosis condition so as to obtain a first type of patient data, and establishing a model for the first type of patient data by a second type of patient according to the first type of patient data; and establishing a model for the second type of patient data according to the first type of patient data, the second type of patient data and the third type of patient data. According to the method, a more accurate classification model is established by utilizing ultrasonic and clinical features and combining a machine learning model.

Description

Technical Field

[0001] The present application relates to the technical field of model establishment, and in particular to a method for establishing a classification prediction model for the prognosis outcome of cesarean scar pregnancy and a device for establishing a classification prediction model for the prognosis outcome of cesarean scar pregnancy. Background Art

[0002] Cesarean scar pregnancy (CSP) refers to a special type of pregnancy in which the gestational sac is implanted in the cesarean scar at the lower part of the anterior wall of the uterus. In recent years, with the increase in cesarean section rates and the advancement of imaging diagnostic technology, the incidence of cesarean scar pregnancy (CSP) has continued to increase. The pregnancy outcomes and prognosis of CSP cases vary, and some cases are accompanied by serious risks, such as life-threatening hemorrhage, uterine rupture, and even death. Transvaginal ultrasound is the preferred evaluation method. CSP can be divided into different types based on US results. For example, by observing the relationship between the gestational sac and the uterine cavity line and serosal line, the Delphi consensus divides it into three types. According to the thickness of the remaining myometrium and whether the gestational sac protrudes into the serosal layer, the Chinese consensus divides it into three types. At present, some studies have attempted to establish a clinical classification model for CSP to identify patients with large surgical bleeding in order to provide a reference for treatment decisions. For example, based on the data of patients with cesarean scar pregnancy, Shandong Qilu Hospital used univariate analysis and multivariate logistic regression analysis to explore the independent risk factors for bleeding (≥300 ml) during cesarean scar pregnancy surgery. The receiver operating characteristic curve method was used to determine the optimal threshold of the identified risk factors, and two variables, the thickness of the anterior scar wall myometrium and the mean diameter of the gestational sac, were used to establish a clinical classification of five cesarean scar pregnancies.

[0003] However, due to the changing understanding and treatment of CSP, the treatment of CSP in different periods, regions and different types varies, including surgical treatments such as curettage, lesion resection, and even hysterectomy, simple drug treatment or drug combined with surgical treatment, etc. In order to better avoid the risk of bleeding, auxiliary means such as uterine artery embolization and balloon compression hemostasis are also used. Therefore, it is not precise and accurate to use only the amount of intraoperative bleeding as the classification standard without considering the differences in treatment methods.

[0004] Therefore, it is hoped that a technical solution can be provided to solve or at least alleviate the above-mentioned deficiencies of the prior art. Summary of the invention

[0005] The object of the present invention is to provide a method for establishing a classification prediction model for the prognosis of cesarean scar pregnancy to at least solve one of the above-mentioned technical problems.

[0006] The present invention provides the following scheme:

[0007] According to one aspect of the present invention, a method for establishing a classification prediction model for the prognosis outcome of cesarean scar pregnancy is provided, and the method for establishing a classification prediction model for the prognosis outcome of cesarean scar pregnancy comprises:

[0008] Step 1: Acquire a patient database, wherein the patient database includes at least one patient data;

[0009] Step 2: Classifying the patient data in the patient database according to treatment methods and prognosis, thereby obtaining first-category patient data, second-category patient data, and third-category patient data;

[0010] Step 3: Establish a model for the first category of patient data and a model for the second category of patient data based on the first category of patient data, the second category of patient data, and the third category of patient data, respectively.

[0011] Optionally, the step 3: establishing a model for the first category of patient data according to the first category of patient data, the second category of patient data, and the third category of patient data respectively includes:

[0012] Step 31: obtaining the first round of multimodal data corresponding to the first category of patient data as positive samples, and obtaining the first round of multimodal data corresponding to the second category of patient data and the third category of patient data as negative samples;

[0013] Step 32: Get the trained large model;

[0014] Step 33: inputting the first round of multimodal data into the trained large model to obtain the first round of feature selection information;

[0015] Step 34: Establishing a first GDBT model according to the first round feature selection information;

[0016] Step 35: Optimize the first GDBT model to obtain a model for the first type of patient data.

[0017] Optionally, the step 35: optimizing the first GDBT model to obtain a model for the first type of patient data includes:

[0018] Step 351: Obtaining feature selection information for optimization based on the first GDBT model;

[0019] Step 352: Obtaining an optimization GDBT model according to the optimization feature selection information;

[0020] Step 353: Evaluate the model performance of the first GDBT model and the optimized GDBT model respectively, and retain the one with better performance as the model for the first category of patient data.

[0021] Optionally, the step 35 of optimizing the first GDBT model to obtain a model for first-category patient data further comprises:

[0022] Step 354: Obtaining optimization feature selection information according to the optimization GDBT model obtained in step 352;

[0023] Step 355: Generate a new GDBT model for optimization according to the feature selection information for optimization;

[0024] Step 356: Evaluate the model performance of the new optimized GDBT model and compare its performance with the previously retained model, so as to retain a model with better performance as the model for the first category of patient data.

[0025] Optionally, the step 35 of optimizing the first GDBT model to obtain a model for first-category patient data further comprises:

[0026] Step 357: Repeatedly generate new GDBT models for optimization and compare their performance with the previously retained models until the model prediction performance converges or decreases to a preset value.

[0027] Optionally, the first round of multimodal data includes clinical knowledge and treatment data related to the patient's disease, and a set of clinical features available to the model.

[0028] Optionally, the step 351: obtaining feature selection information for optimization based on the first GDBT model includes:

[0029] Step 3511: Obtain the first round of multimodal data corresponding to the first category of patient data, the model data obtained by the established first GDBT model, the feature set selected by the current round of machine learning model, the performance information obtained by the current round of machine learning model when training the test set, and the model feature selection and performance iteration history information;

[0030] Step 3512: Obtain the trained large model;

[0031] Step 3513: Input the first round of multimodal data, the model data obtained by establishing the first GDBT model, the feature set selected by the current round of machine learning model, the performance information obtained by the current round of machine learning model when training the test set, the model feature selection and performance iteration history information into the trained large model, thereby obtaining feature selection information for optimization.

[0032] Optionally, establishing a model for the second category of patient data according to the first category of patient data, the second category of patient data, and the third category of patient data includes:

[0033] Obtain a set of univariate significant correlated variables of the first category of patient data and a set of univariate significant correlated variables of the third category of patient data;

[0034] Draw the ROC curve of the single factor significantly correlated variable set of the first category of patient data and the first category of patient data, and the ROC curve of the single factor significantly correlated variable set of the third category of patient data and the third category of patient data;

[0035] Find the threshold value with the maximum Youden index on each ROC curve, and convert the variable into a categorical variable ranging from 0 to 1 based on the threshold value;

[0036] For each categorical variable converted to 0 to 1, four weight scores of 0, 1, 2, and 3 were assigned, and the variable weight combination with the largest AUC value was found based on a computer algorithm.

[0037] The present application also provides a device for establishing a classification prediction model for the prognosis outcome of cesarean scar pregnancy, and the device for establishing a classification prediction model for the prognosis outcome of cesarean scar pregnancy comprises:

[0038] A patient database acquisition module, the patient database acquisition module is used to acquire a patient database, the patient database includes at least one patient data;

[0039] A classification module, wherein the classification module is used to classify the patient data in the patient database according to treatment methods and prognosis conditions, thereby obtaining first-category patient data, second-category patient data, and third-category patient data;

[0040] A first model building module, the first model building module is used to build a model for the first category of patient data according to the first category of patient data, the second category of patient data and the third category of patient data;

[0041] The second model building module is used to build a model for the second category of patient data based on the first category of patient data, the second category of patient data and the third category of patient data.

[0042] The method for establishing a classification prediction model for the prognosis outcome of cesarean scar pregnancy in this application utilizes ultrasound and clinical features, combined with a machine learning model, to establish a more accurate classification model to help clinicians select appropriate treatment methods, identify patients with a higher prognosis risk so that they can receive adequate attention and treatment, and identify patients with a lower prognosis risk to avoid unnecessary losses caused by overtreatment. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is a flowchart of a method for establishing a classification prediction model for the prognosis of cesarean scar pregnancy in one embodiment of the present application.

[0044] Figure 2 It is a schematic diagram of the structure of an electronic device in an embodiment of the present application.

[0045] Figure 3 It is a schematic diagram of the large model selection feature in one embodiment of the present application.

[0046] Figure 4 It is a schematic diagram of characteristic distribution of first-category patient data, second-category patient data, and third-category patient data in one embodiment of the present application.

[0047] Figure 5 Schematic diagram of importance scores of ultrasound and clinical features in one embodiment of the present application.

[0048] Figure 6 It is a schematic diagram of a ROC curve in one embodiment of the present application.

[0049] Figure 7 It is a schematic diagram of the ROC curve of the machine learning model of the first category of patient data in one embodiment of the present application. DETAILED DESCRIPTION

[0050] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. 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.

[0051] like Figure 1 The method for establishing a classification prediction model for the prognosis of cesarean scar pregnancy shown includes:

[0052] Step 1: Acquire a patient database, wherein the patient database includes at least one patient data;

[0053] Step 2: Classifying the patient data in the patient database according to treatment methods and prognosis, thereby obtaining first-category patient data, second-category patient data, and third-category patient data;

[0054] Step 3: Establish a model for the first category of patient data and a model for the second category of patient data based on the first category of patient data, the second category of patient data, and the third category of patient data, respectively.

[0055] The method for establishing a classification prediction model for the prognosis outcome of cesarean scar pregnancy in this application utilizes ultrasound and clinical features, combined with a machine learning model, to establish a more accurate classification model to help clinicians select appropriate treatment methods, identify patients with a higher prognosis risk so that they can receive adequate attention and treatment, and identify patients with a lower prognosis risk to avoid unnecessary losses caused by overtreatment.

[0056] In this embodiment, the step 3: establishing a model for the first type of patient data according to the first type of patient data includes:

[0057] Step 31: Obtain the first round of multimodal data corresponding to the first category of patient data;

[0058] Step 32: Get the trained large model;

[0059] Step 33: inputting the first round of multimodal data into the trained large model to obtain the first round of feature selection information;

[0060] Step 34: Establishing a first GDBT model according to the first round feature selection information;

[0061] Step 35: Optimize the first GDBT model to obtain a model for the first type of patient data.

[0062] In this embodiment, the step 35: optimizing the first GDBT model to obtain a model for the first type of patient data includes:

[0063] Step 351: Obtaining feature selection information for optimization based on the first GDBT model;

[0064] Step 352: Obtaining an optimization GDBT model according to the optimization feature selection information;

[0065] Step 353: Evaluate the model performance of the first GDBT model and the optimized GDBT model respectively, and retain the one with better performance as the model for the first category of patient data.

[0066] In this embodiment, the step 35 of optimizing the first GDBT model to obtain a model for the first type of patient data further includes:

[0067] Step 354: Obtaining optimization feature selection information according to the optimization GDBT model obtained in step 352;

[0068] Step 355: Generate a new GDBT model for optimization according to the feature selection information for optimization;

[0069] Step 356: Evaluate the model performance of the new optimized GDBT model and compare its performance with the previously retained model, so as to retain a model with better performance as the model for the first category of patient data.

[0070] In this embodiment, the step 35 of optimizing the first GDBT model to obtain a model for the first type of patient data further includes:

[0071] Step 357: Repeatedly generate new GDBT models for optimization and compare their performance with the previously retained models until the model prediction performance converges or decreases to a preset value.

[0072] In this embodiment, the first round of multimodal data includes clinical knowledge and diagnosis and treatment data related to the patient's disease, and a set of clinical features available to the model.

[0073] For example, the GDBT model of the present application needs to obtain the optimal GDBT model (including the model parameters of the GDBT model and the selected features) through iterative training. For example, in the large model obtained in step 32, at this time, since it is the first round of training, there are only clinical knowledge and diagnosis and treatment data related to the patient's disease, and a set of clinical features available to the model. The features output by the large model can be considered as the first round of features, but the first round of features are not necessarily the best, because after the first round of features, the first GDBT model can be established to test the performance of the first GDBT model. After the test, it may not meet the requirements or is not the best. Therefore, a new round of feature selection will be performed again, that is, the first round of multimodal data corresponding to the first type of patient data, the model data obtained by the established first GDBT model, the feature set selected by the current round of machine learning model, the performance information obtained by the current round of machine learning model when training the test set, the model feature selection and performance iteration history information are re-input into the trained large model, thereby obtaining the second round of optimization feature selection information for establishing the GDBT model.

[0074] Then, a new GDBT model (temporarily referred to as the second GDBT model) is established based on the optimized feature selection information.

[0075] It is understandable that the second GDBT model may not be the best model. In this case, the model data obtained by the established second GDBT model, the feature set selected by the current round of machine learning model, the performance information obtained by the current round of machine learning model when training the test set, the model feature selection and performance iteration history information are re-input into the trained large model to obtain the third round of feature selection information for optimizing the GDBT model.

[0076] A new GDBT model (temporarily referred to as the third GDBT model) is established based on the feature selection information used in the third round of optimization for establishing the GDBT model.

[0077] It is understandable that the third GDBT model may not be the best model. At this time, the model data obtained by the established third GDBT model, the feature set selected by the current round of machine learning model, the performance information obtained by the current round of machine learning model when training the test set, the model feature selection and performance iteration history information are re-input into the trained large model to obtain the fourth round of feature selection information for optimizing the GDBT model.

[0078] Repeat the above iterations until the preset requirements are met.

[0079] In this embodiment, the step 351: obtaining feature selection information for optimization based on the first GDBT model includes:

[0080] Step 3511: Obtain the first round of multimodal data corresponding to the first category of patient data, the model data obtained by the established first GDBT model, the feature set selected by the current round of machine learning model, the performance information obtained by the current round of machine learning model when training the test set, and the model feature selection and performance iteration history information;

[0081] Step 3512: Obtain the trained large model;

[0082] Step 3513: Input the first round of multimodal data, the model data obtained by establishing the first GDBT model, the feature set selected by the current round of machine learning model, the performance information obtained by the current round of machine learning model when training the test set, the model feature selection and performance iteration history information into the trained large model, thereby obtaining feature selection information for optimization.

[0083] In this embodiment, establishing a model for the second type of patient data according to the first type of patient data, the second type of patient data, and the third type of patient data includes:

[0084] Obtain a set of univariate significantly correlated variables for the first category of patient data and a set of univariate significantly correlated variables for the second category of patient data; in this embodiment, the initial variables are derived by doctors from the patient's electronic medical records or test records, such as indicators recorded in the electronic medical record information system. We will manually extract and summarize the initial continuous variables for subsequent processing.

[0085] In this embodiment, single factor significant variables are selected from various initial variables through single factor analysis.

[0086] The univariate significant variable is evaluated by univariate logistic regression. If the P value is <0.05, the univariate is considered to be significantly correlated. This univariate significant variable will be obtained by statistical analysis of group A and group C respectively, that is, there will be a set of univariate significant correlated variables for group A, and another set of univariate significant correlated variables for group C. The variable sets of the two groups are not necessarily the same.

[0087] Draw the ROC curve of the single factor significantly correlated variable set of the first category of patient data and the first category of patient data, and the ROC curve of the single factor significantly correlated variable set of the third category of patient data and the third category of patient data;

[0088] Find the threshold with the largest Youden index on each ROC curve, and convert the variable into a categorical variable from 0 to 1 based on the threshold; in this embodiment, different points on the ROC curve correspond to the sensitivity and specificity of the model under different classification thresholds, Youden index = sensitivity + specificity - 1, we select a point with the largest Youden index, and the classification threshold corresponding to this point is used to determine whether the patient should belong to group A or group C. For example, the threshold corresponding to the point with the largest Youden index on the ROC curve is a probability value between 0-1. For example, if the threshold is 0.6, then based on the patient's characteristics, the probability predicted by the model is 0.8. If it is greater than 0.6, the patient's label is considered to be 1, and this 1 can represent group A or group C. This rule applies to both models constructed for group A and group C. If group A is modeled, then 1 means belonging to group A, and 0 means not group A (B+C).

[0089] For each categorical variable converted to 0 to 1, four weight scores of 0, 1, 2, and 3 are assigned, and the variable weight combination with the largest AUC value is found based on a computer algorithm. For example, the python hyperopt package can be used to achieve this.

[0090] The present application is further described in detail below by way of examples. It should be understood that the examples do not constitute any limitation to the present application.

[0091] The data used in this application is a dataset constructed based on hospitalized patients with uterine scar pregnancy (CSP) in a hospital in Beijing. Inclusion criteria are: diagnosed with uterine scar pregnancy (CSP) in a hospital in Beijing from January 2006 to October 2023; the final preoperative ultrasound examination is performed by a senior ultrasound physician with more than 15 years of experience. Exclusion criteria include patients in the second trimester of pregnancy, or patients diagnosed with inevitable miscarriage, lower uterine pregnancy, or cervical pregnancy. Patients with incomplete ultrasound or clinical data were also excluded.

[0092] The diagnosis of uterine scar pregnancy was determined by transvaginal ultrasonography according to the following criteria: (1) the uterine cavity and cervical canal were empty; (2) the gestational sac was implanted at the cesarean section scar on the lower anterior wall of the uterus; (3) the myometrium was thinned or disappeared at the cesarean section scar on the lower anterior wall of the uterus; and (4) color Doppler ultrasonography showed nutrient blood flow at the implantation site of the gestational sac. This study was approved by the Ethics Committee of Peking Union Medical College Hospital (Ethics Review Number: K3071).

[0093] Eligible patients were retrieved based on the diagnosis code for uterine scar pregnancy (ICD code: O00.807) in the hospital electronic medical record system. Patient identification information and risk factors related to the prognosis of uterine scar pregnancy were collected by three qualified physicians through the hospital electronic medical record system and reviewed by an obstetrical and gynecological ultrasound expert with more than 15 years of experience.

[0094] The variables included were the long diameter of the gestational sac or mass (GSCL), the short diameter of the gestational sac or mass (GSCH), the width of the gestational sac or mass (GSCW), the maximum diameter of the gestational sac or mass (GSDmax), the average outer diameter of the gestational sac or mass (GSDavg), the long diameter of the implantation part (IMPL), the short diameter of the implantation part (IMPW), the remaining myometrial thickness (RMT), the adjacent myometrial thickness (AMT), the fetus, the fetal bud length (FL), the fetal heart rate (FHB), the protrusion of the lower uterine segment serosal layer (Protrusion), the blood flow distribution around the gestational sac (blood sup), gestational sac blood flow grade (CDFI), peak blood flow velocity (PSV), resistance index (RI), caseous sign (Lacunae), height of gestational sac protruding from UCL (GSUCL), height of gestational sac protruding from SL (GSSL), anteroposterior diameter of gestational sac at the level of diverticulum (GSSH), Unamed29, GS-UA, days of amenorrhea (D), gravidity (P), parity (G), number of cesarean sections (CS), interval to last cesarean section (CST), blood β-HCG value (HCG), abdominal pain, and bleeding.

[0095] All patients were divided into three groups according to the treatment methods and prognoses (data of the first category of patients, data of the second category of patients and data of the third category of patients). Patients with better prognosis were recorded as Group A (data of the first category of patients): only receiving simple drug treatment and the condition improved quickly, or simple curettage treatment, no uterine artery embolization, smooth operation, and blood loss <200ml.

[0096] Patients with severe conditions and poor prognosis were recorded as Group C (third category patient data): patients who needed hysterectomy due to their condition, or uterine curettage or lesion resection with surgical bleeding ≥ 200 ml, or had severe bleeding after surgery and needed surgical intervention, or the original surgery was not carried out smoothly and the surgical procedure was changed midway were recorded as Group C [the original surgery was not carried out smoothly and the surgical procedure needed to be changed midway, the intraoperative bleeding volume ≥ 200 ml, hysterectomy or severe complications such as severe bleeding after surgery];

[0097] Other patients with moderate prognosis were recorded as Group B (second category of patient data): including medication + curettage, such as MTX, KCL and other treatments followed by surgical treatment; UAE + curettage; or MTX supplementary treatment due to residual after curettage (unsatisfactory decrease in HCG or residual indicated by ultrasound).

[0098] The reason why this application analyzes three types of patient data and adopts two models is that, based on single-factor significant variables, this application visualizes the patient populations of groups A, B, and C through principal component analysis (PCA) to analyze the distribution patterns of clinical variables of the three groups of people, and thus finds that the people of groups ABC are displayed in three-dimensional space, and each person is a point on the graph, so that the similarity of the people of groups ABC in three-dimensional space can be evaluated, and the farther the distance, the easier it is to distinguish. Through PCA dimensionality reduction and visualization to three-dimensional space, it can be evaluated that group A is not easy to distinguish, and a more complex model is required, so GBDT modeling is adopted; group C is relatively far away from group AB and easy to distinguish, so a linear weighted model is adopted.

[0099] In this embodiment, considering the high correlation between the included variables, for variables with significant single-factor correlation, the importance of the variables will be scored and ranked using a machine learning model (Gradient Boosting Decision Tree (GDBT)). The greater the contribution of the variable to the machine learning model in predicting patients in group A or group C, the higher the importance score of the variable. Subsequent modeling will further screen the variables based on their importance to the prediction of the artificial intelligence model, and gradually remove the variables with lower weights to streamline the model.

[0100] In this embodiment, the construction of an artificial intelligence model for prognosis prediction using a gradient boosted decision tree (GDBT) includes: randomly extracting 25% of the data from the three groups of patient data, A, B, and C, as a model test set, and the remaining 75% as a model training set. First, all variables are included in the model construction, and then the variables with lower weights are deleted according to the importance of the variables, and the model is re-modeled. If the model prediction performance improves after deleting the variable, the deletion operation is retained, otherwise the variable is re-selected for deletion; repeat the above process multiple times until the model prediction performance converges or decreases. The model performance will be evaluated by calculating the model sensitivity, specificity, and area under the ROC curve (AUC).

[0101] In this embodiment, the second patient data model is established as follows:

[0102] For variables with significant univariate correlation, they will be further converted into categorical variables. The conversion method is to draw the ROC curve corresponding to the variable and the prognosis prediction of group A or group C, and find the threshold with the largest Youden index on the ROC curve, and convert the variable into a 0 / 1 categorical variable based on this threshold. Then for each categorical variable, four weight scores of 0, 1, 2, and 3 are assigned, and the variable weight combination with the largest AUC value is found based on the computer algorithm. After obtaining the optimal variable weight, the optimal scoring model division threshold will be selected by analyzing the ROC curve. When the score exceeds this threshold, the patient is considered to belong to group A or group C.

[0103] Continuous variables are presented as medians and quartiles, and categorical variables are presented as nonzero counts and proportions. Variables with a P value less than 0.05 were considered statistically significant. Statistical analyses were performed using IBM SPSS and Python 3.8.8. Machine learning models were constructed using the scikit-learn toolkit. The hyperparameters max_depth and n_estimators of the GBDT model were set to 2 and 50, respectively. The optimal combination of overall weights of the variables was selected using the “Hyperopt” Python package. The SHAP values ​​of the variables were calculated using the “shap” Python package.

[0104] Unsupervised cluster analysis of CSP patients was performed based on 20 univariate significant variables in group A or group C (GSCL, GSCH, GSCW, GSDmax, GSDavg, IMPL, RMT, AMT, FL, GSUCL, GSSH, Protrusion, GSTP, Blood sup, CDFI, PSV, RI, Lacunae, gestational age, abdominal pain) after removing variables with more than 50% missing values ​​(IMPW, GSSL, IMPA, and GSUA). The PCA method was used to reduce the dimension of each patient's feature vector and Figure 1The visualization results are shown in Figure 1. It can be observed that the patients in group C (marked with red dots) are mainly distributed in the upper right quadrant of the figure, showing obvious distribution differences compared with group A (blue dots) and group B (green dots). From the proximity of the green and blue dots in the figure, it can be seen that the patients in group B and group A show relatively similar characteristic distributions ( Figure 4 ).

[0105] Figure 4 . Visualization of the distribution of CSP patients. Each point in the figure represents a patient, and the three groups ABC are marked in blue, green, and red respectively. The distance between different points in the figure can represent the similarity of clinical variables between two patients.

[0106] Based on the results of univariate analysis, each feature was further evaluated for its importance in predicting the prognosis of patients in group C and group A by machine learning modeling and explanatory analysis. Before modeling, univariate significant variables were filtered according to the proportion of missing values ​​and clinical feasibility. Subsequently, GBDT models were established for group A and group C, respectively, and the SHAP value of each variable was calculated to obtain the importance scores of ultrasound and clinical features ( Figure 5 ). The average absolute value of the SHAP value of each feature reflects its importance in predicting prognosis. The variables in the figure are sorted from high to low according to their weight. It can be seen that for patients in group C, IMPL, GSUCL, and GSSH made the greatest contribution to prognosis prediction. For group A, these variables were IMPL, RMT, and PSV. The importance scores of these three variables, IMPL, GSUCL, and RMT, in groups A and C were always ranked in the top five.

[0107] Figure 5 .Importance of variables used to predict the prognosis of group C and group A. Variables are sorted in descending order according to their average contribution to the model prediction results. Variables with larger values ​​mean they are more important to the model.

[0108] For patients in group C, a machine learning model and a traditional linear scoring model were constructed, and the ROC curves of the models are shown in Figure 6 After variable selection, the final GBDT model contains three features, and the corresponding feature importance values ​​are as follows: IMPL is 0.762, GSUCL is 0.550, and GSSH is 0.376. The GBDT model achieved a sensitivity of 0.875 (0.473, 0.997), a specificity of 0.857 (0.728, 0.941), and an AUC value of 0.927 (0.856, 0.999) on the test dataset (classification threshold is 0.244) (Table 7).

[0109] Figure 6.ROC curves of the two prognostic models in group C. The left side is the ROC curve of the GBDT model, and the right side is the ROC curve of the traditional linear scoring model. The blue line is the ROC curve on the test data set, and the orange line is the ROC curve on the training data set.

[0110] Based on the importance and clinical feasibility of the features, a simplified linear scoring model was derived, as listed in Table 3. In this model, IMPL, GSUCL, and RMT were converted into categorical variables and assigned integer scores (3 points for IMPL ≥ 2.43 cm, 2 points for GSUCL ≥ 1.4 cm, and 1 point for missing RMT). The model achieved a sensitivity of 0.857 (0.421, 0.996), a specificity of 0.840 (0.709, 0.928), and an AUC value of 0.939 (0.872, 1.000) on the test set (Table 4). Patients were classified as high risk if the weighted score was ≥ 3 points, i.e., a value of IMPL ≥ 2.43 cm, or a GSUCL ≥ 1.5 cm and missing RMT.

[0111] Table 3. Scoring model for group C.

[0112]

[0113] Note: A weighted score ≥3 points is considered high risk and stratified into group C, while a score below 3 points is considered low risk and stratified into group A or group B.

[0114] For patients in group A, a GBDT machine learning prognostic model containing 13 significant variables was established by combining the large model with traditional machine learning for automated feature selection and model optimization, and the weight of each feature variable is shown in Figure 5 The ROC curve is drawn in Figure 7 The sensitivity and specificity of the model were 0.867 (0.595, 0.983) and 0.881 (0.744, 0.960), respectively (Table 4). The AUC value of the model on the test data set was 0.917 (0.842, 0.993) (classification threshold was 0.233).

[0115] Figure 7 ROC curve of the machine learning model of Group A. The blue line is the ROC curve on the test dataset, and the orange line is the ROC curve on the training dataset.

[0116] Table 4. Predictive validity of two sets of machine learning and linear scoring models.

[0117]

[0118] AUC, area under the curve; PPV, positive predictive value; NPV, negative predictive value; PLHR, positive likelihood ratio. NLHR, negative likelihood ratio.

[0119] The method of this application has the following advantages:

[0120] 1. Based on the clustering results of patients with scar pregnancy, the difficulty of classification prediction of the three groups A / B / C was evaluated. Since the distribution distance between group C and group A / B was far, the traditional linear scoring model was selected for group C to distinguish them; a machine learning model was constructed for group A to identify them. In this application, model selection based on the results of population clustering can more accurately select more suitable models and methods based on population distribution and modeling complexity; for the convenience of clinical application, when group C patients are relatively easy to distinguish, there is no need to use more complex models for modeling prediction, but a more concise model with good interpretability should be selected for construction.

[0121] 2. Based on traditional logistic regression, single-factor and multi-factor significant variables were identified, and the optimal variable combination was found in combination with computer algorithms. The scoring model for group C includes three 0 / 1 variables: IMPL ≥ 2.43 cm, GSUCL ≥ 1.5 cm, and RMT missing, and the scores of each variable are 3 points, 2 points, and 1 point. For any patient, a weighted sum is performed according to the three variables. A weighted score ≥ 3 points is considered high risk and stratified into group C, while a score below 3 points is considered low risk and stratified into group A or group B. In this application, continuous variables are converted into categorical variables and scored, which can have better interpretability and are easy for doctors to remember and apply clinically.

[0122] 3. For patients not in group C, the large model is combined with the traditional machine learning algorithm to automatically select feature variables and build models, and the optimal variable combination and model prediction results are selected. In this embodiment, the combination of the large model and the machine learning algorithm can automatically perform feature selection and model optimization, and based on the powerful text understanding ability of the large model, it can better combine clinical expertise and select features from a clinical perspective like a doctor. On the one hand, it avoids the time-consuming and laborious problem of traditional manual feature selection, and on the other hand, it has better interpretability than the strategy of selecting features based purely on statistical methods.

[0123] Finally, 13 variables were selected from the 20 variables: IMPL, RMT, PSV, GSUCL, GSSH, GSCL, FL, GSDmax, RI, GSSL, GSCH, GSDavg, and CDFI. The feature weights of the 13 variables in GBDT are:

[0124]

[0125]

[0126] In this embodiment, the big model receives five inputs: a. Medical knowledge and clinical diagnosis and treatment data related to scar pregnancy patients; b. Features selected by the current machine learning model; c. The currently available clinical feature set; d. Performance indicators of the current machine learning model on the training test set; e. Historical information on model features and performance iterations, and then the big model outputs a new feature set for the machine learning model to train. After multiple rounds of iterations, until the performance of the machine learning model is not significantly improved, the training stops and the currently selected clinical feature weights and training model are output.

[0127] The present application also provides a device for establishing a classification prediction model for the prognosis outcome of cesarean scar pregnancy, the device comprising a patient database acquisition module, a classification module, a first model establishment module and a second model establishment module, wherein:

[0128] The patient database acquisition module is used to acquire a patient database, wherein the patient database includes at least one patient data;

[0129] The classification module is used to classify the patient data in the patient database according to the treatment method and the prognosis, so as to obtain the first category of patient data, the second category of patient data and the third category of patient data;

[0130] The first model building module is used to build a model for the first category of patient data according to the first category of patient data, the second category of patient data and the third category of patient data;

[0131] The second model building module is used to build a model for the second category of patient data based on the first category of patient data, the second category of patient data and the third category of patient data.

[0132] Figure 2 It is a structural block diagram of an electronic device provided by one or more embodiments of the present invention.

[0133] like Figure 2As shown, the present application also discloses an electronic device, including: a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other through the communication bus; a computer program is stored in the memory, and when the computer program is executed by the processor, the processor executes the steps of a method for establishing a classification prediction model for the prognosis outcome of cesarean scar pregnancy.

[0134] The present application also provides a computer-readable storage medium storing a computer program executable by an electronic device. When the computer program is run on the electronic device, the steps of the method for establishing a classification prediction model for the prognosis outcome of cesarean scar pregnancy can be implemented.

[0135] The communication bus mentioned in the above electronic device can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The communication bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus.

[0136] The electronic device includes a hardware layer, an operating system layer running on the hardware layer, and an application layer running on the operating system. The hardware layer includes hardware such as a central processing unit (CPU), a memory management unit (MMU), and a memory. The operating system can be any one or more computer operating systems that implement electronic device control through a process, such as a Linux operating system, a Unix operating system, an Android operating system, an iOS operating system, or a Windows operating system. In addition, in an embodiment of the present invention, the electronic device can be a handheld device such as a smart phone or a tablet computer, or can be an electronic device such as a desktop computer or a portable computer, which is not particularly limited in the embodiment of the present invention.

[0137] The execution subject of the electronic device control in the embodiment of the present invention may be an electronic device, or a functional module in the electronic device that can call and execute a program. The electronic device may obtain the firmware corresponding to the storage medium. The firmware corresponding to the storage medium is provided by the supplier. The firmware corresponding to different storage media may be the same or different, which is not limited here. After the electronic device obtains the firmware corresponding to the storage medium, the firmware corresponding to the storage medium may be written into the storage medium, specifically, the firmware corresponding to the storage medium may be burned into the storage medium. The process of burning the firmware into the storage medium may be implemented using existing technology, which will not be described in detail in the embodiment of the present invention.

[0138] The electronic device may also obtain a reset command corresponding to the storage medium. The reset command corresponding to the storage medium is provided by the supplier. The reset commands corresponding to different storage media may be the same or different, and are not limited here.

[0139] At this time, the storage medium of the electronic device is a storage medium in which the corresponding firmware is written, and the electronic device can respond to the reset command corresponding to the storage medium in the storage medium in which the corresponding firmware is written, so that the electronic device resets the storage medium in which the corresponding firmware is written according to the reset command corresponding to the storage medium. The process of resetting the storage medium according to the reset command can be implemented by the existing technology and will not be described in detail in the embodiments of the present invention.

[0140] For the convenience of description, the above devices are described in terms of functions and are divided into various units and modules. Of course, when implementing the present application, the functions of each unit and module can be implemented in the same or multiple software and / or hardware.

[0141] Those skilled in the art will appreciate that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as those generally understood by those skilled in the art in the art to which the present invention belongs. It should also be understood that terms such as those defined in general dictionaries should be understood to have meanings consistent with the meanings in the context of the prior art, and will not be interpreted with idealized or overly formal meanings unless specifically defined.

[0142] For the method embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should know that the embodiments of the present invention are not limited by the order of the actions described, because according to the embodiments of the present invention, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions involved are not necessarily required by the embodiments of the present invention.

[0143] It can be known from the description of the above implementation modes that those skilled in the art can clearly understand that the present application can be implemented by means of software plus a necessary general hardware platform. Based on such an understanding, the technical solution of the present application can be essentially or partly contributed to the prior art in the form of a software product, which can be stored in a storage medium such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes several instructions for enabling a computer device (which can be a personal computer, a server or a network device, etc.) to execute the methods described in the various implementation modes of the present application or certain parts of the implementation modes.

[0144] 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 replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for establishing a classification prediction model for the prognosis of cesarean scar pregnancy, characterized in that: The method for establishing a classification prediction model for the prognosis outcome of cesarean scar pregnancy comprises: Step 1: Acquire a patient database, wherein the patient database includes at least one patient data; Step 2: Classifying the patient data in the patient database according to treatment methods and prognosis, thereby obtaining first-category patient data, second-category patient data, and third-category patient data; Step 3: Establish a model for the first category of patient data and a model for the second category of patient data based on the first category of patient data, the second category of patient data, and the third category of patient data, respectively.

2. The method for establishing a classification prediction model for the prognosis of cesarean scar pregnancy according to claim 1, characterized in that: The step 3: establishing a model for the first category of patient data according to the first category of patient data, the second category of patient data and the third category of patient data respectively comprises: Step 31: obtaining the first round of multimodal data corresponding to the first category of patient data as positive samples, and obtaining the first round of multimodal data corresponding to the second category of patient data and the third category of patient data as negative samples; Step 32: Get the trained large model; Step 33: inputting the first round of multimodal data into the trained large model to obtain the first round of feature selection information; Step 34: Establishing a first GDBT model according to the first round feature selection information; Step 35: Optimize the first GDBT model to obtain a model for the first type of patient data.

3. The method for establishing a classification prediction model for the prognosis of cesarean scar pregnancy according to claim 2, characterized in that: The step 35: optimizing the first GDBT model to obtain a model for the first type of patient data includes: Step 351: Obtaining feature selection information for optimization based on the first GDBT model; Step 352: Obtaining an optimization GDBT model according to the optimization feature selection information; Step 353: Evaluate the model performance of the first GDBT model and the optimized GDBT model respectively, and retain the one with better performance as the model for the first category of patient data.

4. The method for establishing a classification prediction model for the prognosis of cesarean scar pregnancy according to claim 3, characterized in that: The step 35: optimizing the first GDBT model to obtain a model for first-category patient data further includes: Step 354: Obtaining optimization feature selection information according to the optimization GDBT model obtained in step 352; Step 355: Generate a new GDBT model for optimization according to the feature selection information for optimization; Step 356: Evaluate the model performance of the new optimized GDBT model and compare its performance with the previously retained model, so as to retain a model with better performance as the model for the first category of patient data.

5. The method for establishing a classification prediction model for the prognosis of cesarean scar pregnancy according to claim 4, characterized in that: The step 35: optimizing the first GDBT model to obtain a model for first-category patient data further includes: Step 357: Repeatedly generate new GDBT models for optimization and compare their performance with the previously retained models until the model prediction performance converges or decreases to a preset value.

6. The method for establishing a classification prediction model for the prognosis of cesarean scar pregnancy according to claim 5, characterized in that: The first round of multimodal data includes clinical knowledge and diagnosis and treatment data related to the patient's disease, and a set of clinical features available to the model.

7. The method for establishing a classification prediction model for the prognosis of cesarean scar pregnancy according to claim 6, characterized in that: The step 351: obtaining feature selection information for optimization based on the first GDBT model includes: Step 3511: Obtain the first round of multimodal data corresponding to the first category of patient data, the model data obtained by the established first GDBT model, the feature set selected by the current round of machine learning model, the performance information obtained by the current round of machine learning model when training the test set, and the model feature selection and performance iteration history information; Step 3512: Obtain the trained large model; Step 3513: Input the first round of multimodal data, the model data obtained by establishing the first GDBT model, the feature set selected by the current round of machine learning model, the performance information obtained by the current round of machine learning model when training the test set, the model feature selection and performance iteration history information into the trained large model, thereby obtaining feature selection information for optimization.

8. The method for establishing a classification prediction model for the prognosis of cesarean scar pregnancy according to claim 7, characterized in that: The establishing of a model for the second-category patient data according to the first-category patient data, the second-category patient data, and the third-category patient data includes: Obtain a set of univariate significant correlated variables of the first category of patient data and a set of univariate significant correlated variables of the third category of patient data; Draw the ROC curve of the single factor significantly correlated variable set of the first category of patient data and the first category of patient data, and the ROC curve of the single factor significantly correlated variable set of the third category of patient data and the third category of patient data; Find the maximum threshold of Youden index on each ROC curve, and convert the variable into a 0 / 1 categorical variable based on the threshold; For each variable converted to 0 / 1 categorical variables, four weight scores of 0, 1, 2, and 3 were assigned, and the variable weight combination with the largest AUC value was found based on a computer algorithm.

9. A device for establishing a classification prediction model for the prognosis of cesarean scar pregnancy, characterized in that: The cesarean scar pregnancy prognosis outcome classification prediction model establishment device comprises: A patient database acquisition module, the patient database acquisition module is used to acquire a patient database, the patient database includes at least one patient data; A classification module, wherein the classification module is used to classify the patient data in the patient database according to treatment methods and prognosis conditions, thereby obtaining first-category patient data, second-category patient data, and third-category patient data; A first model building module, the first model building module is used to build a model for the first category of patient data according to the first category of patient data, the second category of patient data and the third category of patient data; The second model building module is used to build a model for the second category of patient data based on the first category of patient data, the second category of patient data and the third category of patient data.

Images