Infiltrating micropapillary carcinoma chemotherapy benefit prediction and evaluation method and system

Abstract

The invention discloses a moistening micropapillary carcinoma chemotherapy benefit prediction evaluation method and system, and the method comprises the steps: building a data set based on the related data of the moistening micropapillary carcinoma chemotherapy, carrying out the screening of independent prognosis factors and interaction variables based on the data set, building a risk evaluation model, and carrying out the training, and finally, performing risk prediction and evaluation based on the trained risk evaluation model to complete prediction and evaluation of the benefit of the invasive micropapillary carcinoma chemotherapy. According to the method, baseline features are balanced based on a PSM method when multi-dimensional data are integrated, baseline features of patients in a chemotherapy group and a non-chemotherapy group are matched and corrected through tendency scores, hybrid bias is remarkably reduced, on one hand, interaction item analysis is introduced, and the method is provided for solving the problem that the interaction relation between clinical features and pathological features is not fully excavated in the prior art; according to the scheme, the joint effect between complex variables is analyzed, the dynamic relation having important influence on prognosis is captured, independent verification and performance verification are carried out on significant interaction variables, and the stability and clinical significance of the significant interaction variables are ensured.

Description

Technical Field

[0001] The present invention belongs to the field of medical information, and in particular relates to a method and system for predicting and evaluating the chemotherapy benefit of invasive micropapillary carcinoma. Background Art

[0002] Invasive micropapillary carcinoma (IMPC) is a special subtype of breast cancer. Its unique biological characteristics include high lymphovascular invasion, high lymph node metastasis rate, and high histological grade, which lead to significantly poor prognosis of patients. Although chemotherapy, as the main treatment for IMPC patients, has significantly improved the survival rate in some patients, many patients fail to benefit from chemotherapy due to individual differences, and chemotherapy may even bring unnecessary toxic side effects and economic burden. Therefore, accurately predicting the benefits of chemotherapy in IMPC patients has become an important issue that needs to be solved urgently.

[0003] At present, some breast cancer prognostic models have been applied to predict the chemotherapy benefits of different subtypes of breast cancer. These models usually integrate the patient's clinical characteristics (such as age, stage) and pathological characteristics (such as hormone receptor status, HER2 status), and are constructed based on Cox regression or nomogram methods. However, these models are mainly developed for the more common invasive ductal carcinoma (IDC) patients, and there is still a lack of special designs for the special subtype of IMPC. In addition, the existing models have limitations in the scientific selection of variables and the accuracy of risk stratification, resulting in low predictive ability and clinical applicability.

[0004] The problems with the existing technology are:

[0005] (1) Lack of a specific prediction model for IMPC: The biological characteristics of IMPC are different from other breast cancer subtypes, and existing breast cancer prediction models (such as models for IDC) cannot accurately reflect the prognosis and chemotherapy benefits of IMPC patients. This leads to a lack of effective tool support for predicting the benefits of chemotherapy for IMPC patients.

[0006] (2) Insufficient variable selection methods and model stability: Existing IMPC prognostic models mainly use univariate analysis and multivariate Cox regression to screen independent prognostic factors. However, these methods may have the following problems when dealing with high-dimensional data or small sample sizes: ① Uncorrected baseline characteristic bias: The baseline characteristics of patients in the chemotherapy group and the non-chemotherapy group (such as age, molecular subtype, etc.) may be significantly unbalanced, resulting in bias in the model prediction results. ② Neglect of dynamic interactive variables: The single regression screening method fails to capture the complex interactions between variables and may miss interactive variables that have a significant impact on prognosis. ③ Instability of variable screening: The traditional stepwise regression method relies on preset criteria, which may lead to the elimination of important variables due to sample fluctuations, affecting the stability of the model.

[0007] (3) Insufficient risk stratification tools: Existing models usually provide a single risk score and lack the ability to accurately stratify high and low risks, and cannot effectively distinguish between patients who benefit significantly from chemotherapy and those who benefit only slightly. This deficiency limits the practical application of the model in personalized treatment decision-making.

[0008] (4) The integration depth of multidimensional data is limited: Although existing literature has applied clinical characteristics (such as age, stage) and pathological characteristics (such as ER / PR status, HER2 status) to construct breast cancer prognosis models, the integration of these data remains at a basic level and has not yet explored their complex relationships. For example: Ignoring the interaction of variables: Existing models mostly treat clinical and pathological characteristics as independent variables and do not analyze the potential interaction effects between these variables. Lack of dynamic update mechanism: The current model has poor adaptability to new data and cannot dynamically adjust variable weights to improve prediction performance. Insufficient unified integration of multidimensional data: Existing models mostly only integrate some data dimensions (such as clinical or pathological characteristics) and do not combine dynamic scoring tools (such as X-tile) for risk classification, which lacks comprehensiveness. Summary of the invention

[0009] In view of the above problems, the object of the present invention is to provide a method and system for predicting and evaluating the chemotherapy benefit of invasive micropapillary carcinoma.

[0010] The specific technical solution for achieving the purpose of the present invention is:

[0011] A method for predicting and evaluating the benefit of chemotherapy for invasive micropapillary carcinoma comprises the following steps:

[0012] Step 1, obtaining chemotherapy-related data of invasive micropapillary carcinoma and preprocessing the data to construct a data set;

[0013] Step 2: Screening of independent prognostic factors and interaction variables based on the data set;

[0014] Step 3: Based on the screened independent prognostic factors and interaction variables, a risk assessment model is constructed, and the model is trained based on the data set;

[0015] Step 4: Perform risk prediction and assessment based on the trained risk assessment model to complete the prediction and assessment of chemotherapy benefits for invasive micropapillary carcinoma.

[0016] The present invention also provides a system for predicting and evaluating the chemotherapy benefit of invasive micropapillary carcinoma, comprising the following modules:

[0017] Data processing module: used to obtain chemotherapy-related data of invasive micropapillary carcinoma and pre-process the data, build a data set, and regularly update the data set;

[0018] The sources of the chemotherapy-related data for invasive micropapillary carcinoma include various public databases and clinical databases;

[0019] The data included clinical characteristics, case characteristics, treatment characteristics, and follow-up characteristics;

[0020] The clinical characteristics include age, race, marital status, and side of invasive micropapillary carcinoma;

[0021] The case characteristics included histological grade, tumor size, lymph node status, estrogen receptor status, progesterone receptor status, and human epidermal growth factor receptor 2 status;

[0022] The treatment characteristics include whether the patient has received radiotherapy, surgery, or chemotherapy;

[0023] The follow-up characteristics include survival status, survival time;

[0024] The data preprocessing includes:

[0025] Clean and standardize the data to remove missing values;

[0026] Data were coded for categorical variables and standardized for continuous variables;

[0027] The data preprocessing includes:

[0028] Clean and standardize the data to remove missing values;

[0029] Data were coded for categorical variables and standardized for continuous variables;

[0030] Screening module: used to screen independent prognostic factors and interaction variables based on the data set;

[0031] Risk assessment module: used to build a risk assessment model based on the screened independent prognostic factors and interaction variables, and train the model based on the data set;

[0032] Benefit prediction and evaluation module: used to predict and evaluate risks based on the trained risk assessment model, and complete the prediction and evaluation of the benefits of chemotherapy for invasive micropapillary carcinoma.

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] (1) The present invention has made innovations in integrating multi-dimensional data, surpassing the limitations of existing technologies. On the one hand, the baseline characteristics are balanced based on the PSM method, and the baseline characteristics of patients in the chemotherapy group and the non-chemotherapy group are corrected by propensity score matching (PSM), which significantly reduces confounding bias and makes the prediction results of the model more scientific and stable. This method has not been fully applied in existing models. On the one hand, interaction analysis is introduced. In view of the problem that the existing technology does not fully explore the interactive relationship between clinical characteristics and pathological characteristics, this scheme analyzes the joint effects between complex variables, captures the dynamic relationship that has an important impact on prognosis, and independently verifies and performs performance verification on significant interactive variables to ensure their stability and clinical significance.

[0035] (2) The solution of the present invention combines X-tile software to dynamically stratify the risk scores generated by the model, accurately distinguish high risk from low risk, dynamically calculate the optimal cutoff value, and accurately divide patients into a group with significant chemotherapy benefits (high-risk group) and a group with limited chemotherapy effects (low-risk group). Compared with traditional models, the X-tile method provides a more intuitive and flexible risk stratification tool, which significantly improves the clinical interpretability of the prediction;

[0036] (3) The solution of the present invention realizes the deep integration of multi-dimensional data. Through the dynamic update mechanism, new data is introduced regularly and variable weights are dynamically adjusted so that the model can continuously maintain the best prediction performance.

[0037] In summary, the present invention significantly improves the scientificity and practicality of IMPC chemotherapy benefit prediction through the comprehensive innovation of PSM method, interaction analysis and X-tile dynamic stratification, and provides solid technical support for personalized precision medicine.

[0038] The present invention is further described below in conjunction with specific implementation modes. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 The present invention is a schematic flow chart of the method for predicting and evaluating the chemotherapy benefit of invasive micropapillary carcinoma.

[0040] Figure 2 This is a schematic diagram of the framework of the system for predicting and evaluating the chemotherapy benefit of invasive micropapillary carcinoma of the present invention.

[0041] Figure 3 It is a schematic diagram of the Nomogram model of 1-year, 3-year, and 5-year OS of IMPC patients in an embodiment of the present invention.

[0042] Figure 4 Schematic diagram of ROC curves of 1-year, 3-year and 5-year OS of the training set (A) and validation set (B) in an embodiment of the present invention.

[0043] Figure 5Figure 3 is the calibration curve of OS at 1, 3 and 5 years for the training set (A) and validation set (B) in the embodiment of the present invention.

[0044] Figure 6 Decision curves for OS at 1, 3 and 5 years for the training set (A) and validation set (B) in an embodiment of the present invention.

[0045] Figure 7 This is the Kaplan-Meier analysis of OS of IMPC patients in different risk subgroups in the embodiments of the present invention, wherein (A) is the training set, (B) is the validation set, and (C) is the total population.

[0046] Figure 8 This is a Kaplan-Meier analysis of OS of low-risk IMPC patients with or without chemotherapy in the embodiments of the present invention. (A) Training set, (B) Validation set, (C) Total population.

[0047] Fig. 9 This is the Kaplan-Meier analysis of OS of high-risk IMPC patients with or without chemotherapy in the embodiments of the present invention. (A) Training set, (B) Validation set, (C) Total population. DETAILED DESCRIPTION

[0048] Example

[0049] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention. 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.

[0050] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an kind" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0051] Unless otherwise specifically stated, the relative arrangement, numerical expressions and numerical values ​​of the parts and steps set forth in these embodiments do not limit the scope of the present application. At the same time, it should be understood that, for ease of description, the sizes of the various parts shown in the accompanying drawings are not drawn according to the actual proportional relationship. The technology, method and equipment known to those of ordinary skill in the relevant field may not be discussed in detail, but in appropriate cases, the technology, method and equipment should be considered as a part of the authorization specification. In all examples shown and discussed here, any specific value should be interpreted as being merely exemplary, rather than as a limitation. Therefore, other examples of exemplary embodiments may have different values. It should be noted that similar reference numerals and letters represent similar items in the following drawings, so that once a certain item is defined in an accompanying drawing, it does not need to be further discussed in subsequent drawings.

[0052] Combination Figure 1 , a method for predicting and evaluating the benefit of chemotherapy for invasive micropapillary carcinoma, comprising the following steps:

[0053] Step 1: Obtain chemotherapy-related data for invasive micropapillary carcinoma, preprocess the data, construct a data set, and regularly update the data set;

[0054] Among them, the data include clinical characteristics, case characteristics, treatment characteristics and follow-up characteristics;

[0055] The clinical characteristics include age, race, marital status, side of invasive micropapillary carcinoma lesions, etc.;

[0056] The case characteristics include histological grade, tumor size, lymph node status, estrogen receptor status, progesterone receptor status, human epidermal growth factor receptor 2 status, etc.;

[0057] The treatment characteristics include whether the patient has received radiotherapy, surgery, chemotherapy, etc.;

[0058] The follow-up characteristics include survival status, survival time, etc.

[0059] The data preprocessing includes:

[0060] Clean and standardize the data to remove missing values;

[0061] The data were coded for categorical variables and standardized for continuous variables, such as age (<50 years = 1, ≥50 years = 2), race (white = 1, black = 2, other = 3), marital status (married = 1, single = 2), lesion side (left = 1, right = 2), histological grade (grade I-II = 1, grade III-IV = 2), tumor size (T1 = 1, T2 = 2, T3 = 3, T4 = 4), lymph node status (N0 = 0, N1 = 1, N2 = 3, N3 = 4), ER status (negative = 0, positive = 1), PR status (negative = 0, positive = 1), HER2 status (negative = 0, positive = 1), whether the patient had received radiotherapy (no = 0, yes = 1), whether the patient had received surgery (no = 0, yes = 1), whether the patient had received chemotherapy (no = 0, yes = 1), survival status (alive = 0, dead = 1), and survival time (months).

[0062] Step 2: Screening of independent prognostic factors and interaction variables based on the data set:

[0063] Step 2-1, perform propensity score matching and baseline correction on the data to balance the baseline characteristics of the data that have undergone chemotherapy and the data that have not undergone chemotherapy, and eliminate confounding bias. Compared with directly screening the prognostic factors of the included population, the PSM-matched data set significantly reduces the prediction bias caused by differences in baseline characteristics, providing scientific support for subsequent analysis. Specifically, it is based on the 1:1 proximity matching method, in which each data point that has undergone chemotherapy is matched with a data point that has not undergone chemotherapy and has a set caliper value (for example, 0.2 times the standard deviation), and the Chi squared test is used to verify the balance of the matched characteristics:

[0064] Step 2-2: Screening of independent prognostic factors and interaction variables. Compared with the traditional stepwise regression model, this step systematically introduces interaction terms to improve the dynamic adaptability of the model and the comprehensiveness of variable screening, avoiding the omission of complex relationships that are important for prognosis. Specifically, it includes the following steps:

[0065] Step 2-2-1, determine the univariate based on Cox regression analysis, and screen the variables significantly correlated with the overall survival rate (OS), that is, set a threshold, and the univariate with a value less than the threshold (for example, P ≤ 0.05) is considered to be a relevant univariate. In this embodiment, the univariate determined based on Cox regression analysis is: marital status, tumor size, ER status, PR status, and whether or not radiotherapy has been received;

[0066] Step 2-2-2, in combination with the characteristics of IMPC with a high lymph node metastasis rate and high-level histological grade, as well as clinical practice experience, in order to capture the dynamic relationship between features, an interaction term detection method is used to gradually introduce interaction terms between variables. The interaction terms determined in this embodiment include: N stage × histological grade, N stage × radiotherapy, N stage × T stage, T stage × radiotherapy, HER2 × surgery, age × histological grade. The specific meanings of different variables are as follows: T stage: refers to the local extension of the tumor, especially the size of the tumor, the scope of invasion, and whether it invades surrounding tissues (T1: the maximum diameter of the primary tumor is ≤2 cm; T2: the maximum diameter of the primary tumor is >2 cm but ≤5 cm; T3: the maximum diameter of the primary tumor is >5 cm, or the tumor has invaded the chest wall or skin; T4: the primary tumor invades the chest wall or skin). N stage: refers to the lymph node stage of the tumor, often refers to the lymph node involvement of the tumor (N0: no regional lymph node metastasis, indicating that the tumor has not spread to the lymph nodes; N1: mild regional lymph node metastasis, usually 1-3 axillary or clavicular lymph nodes are involved; N2: moderate regional lymph node metastasis, usually 4-9 axillary lymph nodes or deeper lymph nodes are involved; N3: extensive regional lymph node metastasis, usually more than 10 axillary lymph nodes or other deep lymph nodes are metastatic). Histological grade: Tumors are graded according to the degree of differentiation of tumor cells (grade I: well differentiated; grade II: moderately differentiated; grade III: poorly differentiated; grade IV: undifferentiated). HER2 status: refers to the expression level or amplification status of HER2 in breast cancer. It is divided into HER2-positive and HER2-negative according to the results of immunohistochemistry and fluorescence in situ hybridization. Age: refers to the age of the patient, which can be divided into two categories: <50 years old and ≥50 years old. Radiotherapy: refers to whether the patient has received radiotherapy for breast tumors (yes, no). Surgery: refers to whether the patient has received surgical treatment for breast tumors (yes, no);

[0067] The Wald test was used to screen significant interaction terms. If the empirical value was less than the set threshold (such as P < 0.05), it indicated that the interaction term had a significant effect on prognosis.

[0068] The significant interaction terms screened in this example are: age × histological grade (P = 0.003); in addition, this example also uses the survival package and survcomp package of the R language to compare the model performance (such as AIC, BIC, C-index) before and after the introduction of the interaction term to confirm whether the interaction term improves the predictive ability of the model. The results of AIC (858.68 vs 851.05), BIC (870.20 vs 864.87), and C-index (0.76 vs 0.77) before and after the introduction of the interaction term suggest that the introduction of the interaction term (age × histological grade) can improve the predictive ability of the model;

[0069] Step 2-2-3: Incorporate the selected univariate variables and interaction terms into the multivariate Cox regression model to determine the final independent prognostic factors and interaction terms:

[0070] First, the regression coefficients of the selected single variables and interaction terms are determined based on the multivariate Cox regression model:

[0071] h(t)=h0(t)exp(β1X1+β2X2+…+β i X i )

[0072] Among them, β i represents the selected single variable or interaction term, β i represents the regression coefficient of the single variable or interaction term to be determined, indicating the strength and direction of the interaction effect, h(t) is the risk function, which refers to the risk of an individual experiencing an event at a specific time t, and h0(t) is the baseline risk function, which represents the basic risk of an individual experiencing an event without the influence of any predictor variable;

[0073] The Wald test is then used to determine whether the selected single variables and interaction terms are significant based on the determined regression coefficient, and the P value is obtained based on the regression coefficient (β) and standard error (SE). If the P value is less than the set significance level threshold (usually 0.05), the regression coefficient is significant, indicating that the variable has a significant impact on the risk, thereby determining the final selected single variables and interaction terms.

[0074] The final independent prognostic factors and interaction terms determined in this example are: marital status, tumor size, ER status, PR status, whether or not radiotherapy was received, and age × histological grade;

[0075] Step 3: Based on the screened independent prognostic factors and interaction variables, a risk assessment model is constructed and trained based on the data set:

[0076] Divide the data set into a training set and a validation set in a certain ratio, for example, the ratio is 7:3;

[0077] The Nomogram model was constructed using the rms package of the R language, and the risk assessment model was determined as follows:

[0078] Risk Score=α1Y1+α2Y2+…+α i Y i

[0079] Among them, α i is the regression coefficient of the i-th variable, Y i is the value of the variable;

[0080] In this embodiment, based on the single variables and interaction terms determined in step 2, the risk assessment model is obtained as follows:

[0081] Risk score=P martial status ×a+P T stage ×b+P ER ×c+P PR ×d+P radiotherapy ×e

[0082] +P age_grade ×f;

[0083] The value of P is determined according to the classification of different factors, among which married P martial status =0, unmarried martial status =1; T1 P T stage =0, T2-4 P T stage =1; ER positive P ER =0, ER negative P ER =1; PR positive P PR =0, PR negative PR =1; not received radiotherapy P radiotherapy =0, received radiotherapy P radiotherapy =1; <50 years old and histological grade I-II P age_grade =0, otherwise P age_grade =1;

[0084] a, b, c, d, e, and f in the risk score are preset constant values, i.e., the corresponding goose coefficients, a=46, c=35, d=15, e=20. The preset constant value of b varies according to different T stages, b=34 at T2, b=75 at T3, and b=100 at T4. The preset constant value of f varies according to different ages and histological grades, f=62 when <50 years old and histological grade III-IV, f=92 when ≥50 years old and histological grade I-II, and f=100 when ≥50 years old and histological grade III-IV.

[0085] The coefficients in the risk assessment model were trained and adjusted based on the data set, and the risk assessment model was verified using ROC curve analysis (using the AUC value to evaluate the predictive ability of the model), calibration curve (testing the consistency between the predicted results of the Nomogram model and the actual results) and decision curve analysis (evaluating the net benefit of the model under different thresholds).

[0086] Step 4: Perform risk prediction and evaluation based on the trained risk assessment model to complete the prediction and evaluation of the chemotherapy benefit of invasive micropapillary carcinoma, that is, determine the risk assessment score based on the trained risk assessment model. If the risk assessment score is higher than the determined cutoff value, it is considered that there is a significant benefit from chemotherapy, otherwise it is considered that there is no significant benefit from chemotherapy;

[0087] The cutoff value is obtained by analyzing the continuous variables in the risk assessment model according to X-tile, and the specific steps include:

[0088] ① Input data: Input clinical data and prognosis results (survival time and survival status).

[0089] ② Split variables: Select a continuous variable (risk score), and then the X-tile software will automatically calculate the different split points of the variable for survival analysis.

[0090] ③ Selection of the optimal cutoff value: X-tile selects the optimal cutoff point that can significantly distinguish the survival or prognosis difference between the two groups (high-risk group and low-risk group) by comparing the Log-rank test results of different cutoff points. These optimal cutoff values ​​are usually determined by maximizing statistical significance or minimizing the difference between groups. The optimal cutoff value in this example is 154.

[0091] In addition, this scheme regularly updates the database and the data and follow-up results of IMPC patients in clinical practice, repeats the above modeling and verification steps 1-4 to retrain the model parameters, so that the model can adapt to the changing characteristics of the patient population. This dynamic update mechanism avoids the problem of model aging and ensures the real-time and reliability of the prediction results. Compared with the traditional static model, this real-time update method significantly improves the flexibility and adaptability of the model, and can maintain the real-time and accuracy of the prediction results, thus being closer to the needs of individualized treatment.

[0092] The present invention also provides a system for predicting and evaluating the chemotherapy benefit of invasive micropapillary carcinoma, comprising the following modules:

[0093] Data processing module: used to obtain chemotherapy-related data of invasive micropapillary carcinoma and pre-process the data, build a data set, and regularly update the data set;

[0094] The sources of the chemotherapy-related data for invasive micropapillary carcinoma include various public databases and clinical databases;

[0095] The data included clinical characteristics, case characteristics, treatment characteristics, and follow-up characteristics;

[0096] The clinical characteristics include age, race, marital status, and side of invasive micropapillary carcinoma;

[0097] The case characteristics included histological grade, tumor size, lymph node status, estrogen receptor status, progesterone receptor status, and human epidermal growth factor receptor 2 status;

[0098] The treatment characteristics include whether the patient has received radiotherapy, surgery, or chemotherapy;

[0099] The follow-up characteristics include survival status, survival time;

[0100] The data preprocessing includes:

[0101] Clean and standardize the data to remove missing values;

[0102] The data were coded for categorical variables and standardized for continuous variables, such as age (<50 years = 1, ≥50 years = 2), race (white = 1, black = 2, other = 3), marital status (married = 1, single = 2), lesion side (left = 1, right = 2), histological grade (grade I-II = 1, grade III-IV = 2), tumor size (T1 = 1, T2 = 2, T3 = 3, T4 = 4), lymph node status (N0 = 0, N1 = 1, N2 = 3, N3 = 4), ER status (negative = 0, positive = 1), PR status (negative = 0, positive = 1), HER2 status (negative = 0, positive = 1), whether the patient had received radiotherapy (no = 0, yes = 1), whether the patient had received surgery (no = 0, yes = 1), whether the patient had received chemotherapy (no = 0, yes = 1), survival status (alive = 0, dead = 1), and survival time (months).

[0103] Screening module: used to screen independent prognostic factors and interaction variables based on the data set:

[0104] Propensity score matching and baseline correction were performed on the data to balance the baseline characteristics of the data that had undergone chemotherapy and the data that had not undergone chemotherapy, eliminating confounding bias. Compared with direct screening of prognostic factors for the included population, the PSM-matched data set significantly reduced the prediction bias caused by differences in baseline characteristics, providing scientific support for subsequent analysis. Specifically, it is based on the 1:1 proximity matching method, in which each data point that received chemotherapy is matched with a data point that did not receive chemotherapy with a set caliper value (for example, 0.2 times the standard deviation), and the Chi squared test is used to verify the balance of the matched characteristics:

[0105] Independent prognostic factors and interaction variables were screened. Compared with the traditional stepwise regression model, this step systematically introduced interaction terms to improve the dynamic adaptability of the model and the comprehensiveness of variable screening, avoiding the omission of complex relationships that are important for prognosis, including:

[0106] Univariate analysis was used to determine the variables significantly associated with the overall survival rate (OS), i.e., a threshold was set, and univariate values ​​less than the threshold (e.g., P ≤ 0.05) were considered to be relevant univariate variables. In this embodiment, the univariate variables determined based on Cox regression analysis were: marital status, tumor size, ER status, PR status, and whether radiotherapy had been received.

[0107] In view of the characteristics of IMPC with high lymph node metastasis rate and high-level histological grade, as well as clinical practice experience, in order to capture the dynamic relationship between features, the interaction term detection method is adopted to gradually introduce interaction terms between variables. The interaction terms determined in this embodiment include: N stage × histological grade, N stage × radiotherapy, N stage × T stage, T stage × radiotherapy, HER2 × surgery, age × histological grade. The specific meanings of different variables are as follows: T stage: refers to the local extension of the tumor, especially the size of the tumor, the scope of invasion, and whether it invades the surrounding tissues (T1: the maximum diameter of the primary tumor is ≤2 cm; T2: the maximum diameter of the primary tumor is >2 cm but ≤5 cm; T3: the maximum diameter of the primary tumor is >5 cm, or the tumor has invaded the chest wall or skin; T4: the primary tumor invades the chest wall or skin). N stage: refers to the lymph node stage of the tumor, often refers to the lymph node involvement of the tumor (N0: no regional lymph node metastasis, indicating that the tumor has not spread to the lymph nodes; N1: mild regional lymph node metastasis, usually 1-3 axillary or clavicular lymph nodes are involved; N2: moderate regional lymph node metastasis, usually 4-9 axillary lymph nodes or deeper lymph nodes are involved; N3: extensive regional lymph node metastasis, usually more than 10 axillary lymph nodes or other deep lymph nodes are metastatic). Histological grade: Tumors are graded according to the degree of differentiation of tumor cells (grade I: well differentiated; grade II: moderately differentiated; grade III: poorly differentiated; grade IV: undifferentiated). HER2 status: refers to the expression level or amplification status of HER2 in breast cancer. It is divided into HER2-positive and HER2-negative according to the results of immunohistochemistry and fluorescence in situ hybridization. Age: refers to the age of the patient, which can be divided into two categories: <50 years old and ≥50 years old. Radiotherapy: refers to whether the patient has received radiotherapy for breast tumors (yes, no). Surgery: refers to whether the patient has received surgical treatment for breast tumors (yes, no);

[0108] The Wald test was used to screen significant interaction terms. If the empirical value was less than the set threshold (such as P < 0.05), it indicated that the interaction term had a significant effect on prognosis.

[0109] The significant interaction terms screened in this example are: age × histological grade (P = 0.003); in addition, this example also uses the survival package and survcomp package of the R language to compare the model performance (such as AIC, BIC, C-index) before and after the introduction of the interaction term to confirm whether the interaction term improves the predictive ability of the model. The results of AIC (858.68 vs 851.05), BIC (870.20 vs 864.87), and C-index (0.76 vs 0.77) before and after the introduction of the interaction term suggest that the introduction of the interaction term (age × histological grade) can improve the predictive ability of the model;

[0110] The screened univariate variables and interaction terms were included in the multivariate Cox regression model to determine the final independent prognostic factors and interaction terms:

[0111] First, the regression coefficients of the selected single variables and interaction terms are determined based on the multivariate Cox regression model:

[0112] h(t)=h0(t)exp(β1X1+β2X2+…+β i X i )

[0113] Among them, β i represents the selected single variable or interaction term, β i represents the regression coefficient of the single variable or interaction term to be determined, indicating the strength and direction of the interaction effect, h(t) is the risk function, which refers to the risk of an individual experiencing an event at a specific time t, and h0(t) is the baseline risk function, which represents the basic risk of an individual experiencing an event without the influence of any predictor variable;

[0114] The Wald test is then used to determine whether the selected single variables and interaction terms are significant based on the determined regression coefficient, and the P value is obtained based on the regression coefficient (β) and standard error (SE). If the P value is less than the set significance level threshold (usually 0.05), the regression coefficient is significant, indicating that the variable has a significant impact on the risk, thereby determining the final selected single variables and interaction terms.

[0115] The final independent prognostic factors and interaction terms determined in this example are: marital status, tumor size, ER status, PR status, whether or not radiotherapy was received, and age × histological grade;

[0116] Risk assessment module: used to build a risk assessment model based on the screened independent prognostic factors and interaction variables, and train the model based on the data set:

[0117] Divide the data set into a training set and a validation set in a certain ratio, for example, the ratio is 7:3;

[0118] The Nomogram model was constructed using the rms package of the R language, and the risk assessment model was determined as follows:

[0119] Risk Score=α1Y1+α2Y2+…+α i Y i

[0120] Among them, β i is the regression coefficient of the ith variable, X i is the value of the variable;

[0121] In this embodiment, based on the single variables and interaction terms determined in step 2, the risk assessment model is obtained as follows:

[0122] Risk score=P martial status ×a+P T stage ×b+P ER ×c+P PR ×d+P radiotherapy ×e

[0123] +P age_grade ×f;

[0124] The value of P is determined according to the classification of different factors, among which married P martial status =0, unmarried martial status =1; T1 P T stage =0, T2-4 P T stage =1; ER positive P ER =0, ER negative P ER =1; PR positive P PR =0, PR negative PR =1; not received radiotherapy P radiotherapy =0, received radiotherapy P radiotherapy =1; <50 years old and histological grade I-II P age_grade =0, otherwise P age_grade =1;

[0125] a, b, c, d, e, and f in the risk score are preset constant values, i.e., the corresponding goose coefficients, a=46, c=35, d=15, e=20. The preset constant value of b varies according to different T stages, b=34 at T2, b=75 at T3, and b=100 at T4. The preset constant value of f varies according to different ages and histological grades, f=62 when <50 years old and histological grade III-IV, f=92 when ≥50 years old and histological grade I-II, and f=100 when ≥50 years old and histological grade III-IV.

[0126] The coefficients in the risk assessment model were trained and adjusted based on the data set, and the risk assessment model was verified using ROC curve analysis (using the AUC value to evaluate the predictive ability of the model), calibration curve (testing the consistency between the predicted results of the Nomogram model and the actual results) and decision curve analysis (evaluating the net benefit of the model under different thresholds).

[0127] Benefit prediction and evaluation module: used to predict and evaluate risks based on the trained risk assessment model, and complete the prediction and evaluation of the benefits of chemotherapy for invasive micropapillary carcinoma. That is, based on the trained risk assessment model, the risk assessment score is determined. If the risk assessment score is higher than the determined cutoff value, it is considered that there is a significant benefit from chemotherapy. Otherwise, it is considered that there is no significant benefit from chemotherapy.

[0128] The cutoff value is obtained by analyzing the continuous variables in the risk assessment model according to X-tile, and the specific steps include:

[0129] ① Input data: Input clinical data and prognosis results (survival time and survival status).

[0130] ② Split variables: Select a continuous variable (risk score), and then the X-tile software will automatically calculate the different split points of the variable for survival analysis.

[0131] ③ Selection of the optimal cutoff value: X-tile compares the Log-rank test results of different cutoff points to select the optimal cutoff point that can significantly distinguish the survival or prognosis differences between the two groups (high-risk group and low-risk group). These optimal cutoff values ​​are usually determined by maximizing statistical significance or minimizing the difference between groups. The optimal cutoff value in this example is 154. The optimal cutoff value in this example is 154.

[0132] In addition, this scheme regularly updates the database and the data and follow-up results of IMPC patients in clinical practice, repeats the above modeling and verification steps 1-4 to retrain the model parameters, so that the model can adapt to the changing characteristics of the patient population. This dynamic update mechanism avoids the problem of model aging and ensures the real-time and reliability of the prediction results. Compared with the traditional static model, this real-time update method significantly improves the flexibility and adaptability of the model, and can maintain the real-time and accuracy of the prediction results, thus being closer to the needs of individualized treatment.

[0133] For example, a patient was diagnosed with IMPC, was 48 years old, married, histological grade II, T2, ER positive, PR positive, and had not received radiotherapy.

[0134] Risk score calculation: Risk score = P martial status ×a+P T stage ×b+P ER ×c+P PR ×d+P radiotherapy ×e+P age_grade ×f, substitute the variables and get: Risk Score = 54.

[0135] Result: The patient belongs to the low-risk group and chemotherapy is not recommended.

[0136] For example, a patient was diagnosed with IMPC, he was 65 years old, married, histological grade II, T3, ER negative, PR negative, and received radiotherapy.

[0137] Risk score calculation: Risk score = P martial status ×a+P T stage ×b+P ER ×c+P PR ×d+P radiotherapy ×e+P age_grade ×f, substitute the variables and get: Risk Score = 217.

[0138] Result: The patient belongs to the high-risk group and chemotherapy is recommended.

[0139] like Figure 3The risk assessment model of this scheme is shown in the figure. It is based on the results of multivariate analysis and integrates clinical and pathological characteristics, including marital status (Marital_status), tumor stage (T_stage), estrogen receptor (ER), progesterone receptor (PR), radiotherapy (Radiotherapy), and the interaction variable of age and histological grade (Age_Grade). It is used to evaluate the 1-year, 3-year and 5-year survival probability of patients and perform risk stratification according to the total score (Total Points);

[0140] like Figure 4 The ROC curves of 1-year, 3-year and 5-year OS of the training set (A) and the validation set (B) are shown. The risk assessment model constructed based on this scheme has an ROC curve area (AUC) of 1-year, 3-year and 5-year OS of IMPC patients in the training set of 0.746 (95% CI: 0.622-0.869), 0.811 (95% CI: 0.748-0.874) and 0.774 (95% CI: 0.708-0.840), respectively; in the validation set, the AUC of 1-year, 3-year and 5-year OS are 0.849 (95% CI: 0.639-1.058), 0.680 (95% CI: 0.521-0.839) and 0.619 (95% CI: 0.661-0.840), respectively. CI: 0.444-0.794). The above data show that the risk assessment model constructed by this scheme has a high prediction accuracy and can effectively predict the prognosis of IMPC patients, showing a relatively ideal clinical application potential;

[0141] like Figure 5 The following is the calibration curve analysis of OS in the training set (A) and validation set (B) at 1, 3, and 5 years. In the training set and validation set, the prediction calibration curves of the risk assessment model constructed by this scheme for 1-year, 3-year, and 5-year OS are all close to the diagonal line, which indicates that the model shows good calibration in actual prediction and can accurately reflect the actual survival of patients.

[0142] like Figure 6 The figure shows the decision curve analysis of OS at 1, 3, and 5 years for the training set (A) and validation set (B). In both the training and validation sets, the risk assessment model constructed by this scheme was able to show a high net benefit value under different threshold settings, further verifying the practicality and effectiveness of the model in clinical decision-making, and providing important reference value in clinical practice.

[0143] like Figure 7The figure shows the Kaplan-Meier analysis of OS of IMPC patients in different risk subgroups. (A) Training set, (B) Validation set, (C) Total population. In the training set, validation set, and total population, the hazard ratio HR (95% CI) of OS of patients in the high-risk group was 4.778 (95% CI: 2.951-7.737), 3.320 (95% CI: 1.519-7.259), and 4.309 (95% CI: 2.864-6.484), respectively, compared with patients in the low-risk group, and the P values ​​were all less than 0.05, indicating that the risk stratification based on the risk assessment model constructed in this scheme can effectively distinguish the prognostic risk of IMPC patients and provide an accurate basis for subsequent treatment strategies.

[0144] like Figure 8 Figure 2 shows the Kaplan-Meier analysis of OS of low-risk IMPC patients with or without chemotherapy, including (A) training set, (B) validation set, and (C) total population. In the training set, validation set, and total population, the hazard ratio HR (95% CI) of the chemotherapy group and the non-chemotherapy group of low-risk patients was 0.607 (95% CI: 0.274-1.341), 0.369 (95% CI: 0.097-1.394), and 0.531 (95% CI: 0.271-1.041), respectively, and the P values ​​were all greater than 0.05, indicating that chemotherapy did not significantly improve the OS of low-risk IMPC patients, suggesting that chemotherapy does not need to be given priority in the treatment of low-risk patients.

[0145] like Fig. 9 Figure 2 shows the Kaplan-Meier analysis of OS of high-risk IMPC patients with or without chemotherapy, including (A) training set, (B) validation set, and (C) total population. In the training set, validation set, and total population, the hazard ratio HR (95% CI) of the chemotherapy group and the non-chemotherapy group of high-risk patients was 0.534 (95% CI: 0.314-0.907), 0.139 (95% CI: 0.039-0.495), and 0.434 (95% CI: 0.263-0.715), respectively, and the P values ​​were all less than 0.05, indicating that the OS of high-risk IMPC patients was significantly improved after chemotherapy, suggesting that chemotherapy has a significant effect in improving the prognosis of high-risk patients.

[0146] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A method for predicting and evaluating the benefit of chemotherapy for invasive micropapillary carcinoma, characterized in that: The following steps are involved: Step 1: Obtain chemotherapy-related data for invasive micropapillary carcinoma, preprocess the data, construct a data set, and regularly update the data set; Step 2: Screening of independent prognostic factors and interaction variables based on the data set; Step 3: Based on the screened independent prognostic factors and interaction variables, a risk assessment model is constructed, and the model is trained based on the data set; Step 4: Perform risk prediction and assessment based on the trained risk assessment model to complete the prediction and assessment of chemotherapy benefits for invasive micropapillary carcinoma.

2. The method for predicting and evaluating the chemotherapy benefit of invasive micropapillary carcinoma according to claim 1, characterized in that: The sources of the chemotherapy-related data for invasive micropapillary carcinoma in step 1 include various public databases and clinical databases; The data included clinical characteristics, case characteristics, treatment characteristics, and follow-up characteristics; The clinical characteristics include age, race, marital status, and side of invasive micropapillary carcinoma; The case characteristics included histological grade, tumor size, lymph node status, estrogen receptor status, progesterone receptor status, and human epidermal growth factor receptor 2 status; The treatment characteristics include whether the patient has received radiotherapy, surgery, or chemotherapy; The follow-up characteristics include survival status and survival time.

3. The method for predicting and evaluating the chemotherapy benefit of invasive micropapillary carcinoma according to claim 1, characterized in that: The data preprocessing in step 1 includes: Clean and standardize the data to remove missing values; The data were coded for categorical variables and standardized for continuous variables.

4. The method for predicting and evaluating the chemotherapy benefit of invasive micropapillary carcinoma according to claim 1, characterized in that: The screening of independent prognostic factors and interaction variables in step 2 is specifically as follows: Step 2-1, perform propensity score matching and baseline correction on the data to balance the baseline characteristics of the data that have undergone chemotherapy and the data that have not undergone chemotherapy, that is, based on the 1:1 neighbor matching method, each data that has undergone chemotherapy is matched with a data that has not undergone chemotherapy with a set caliper value, and the chi-square test is used to verify the balance of the characteristics after matching; Step 2-2: Screening of independent prognostic factors and interaction variables.

5. The method for predicting and evaluating the chemotherapy benefit of invasive micropapillary carcinoma according to claim 4, characterized in that: The screening of independent prognostic factors and interaction variables in step 2-2 is specifically as follows: Step 2-2-1, determine the single variable based on Cox regression analysis, and screen the variables significantly correlated with overall survival (OS); Step 2-2-2: Use the interaction term detection method to introduce interaction terms between variables, and screen significant interaction terms through the Wald test. If the empirical value is less than the set threshold, it means that the interaction term has a significant effect on prognosis; Step 2-2-3: Incorporate the selected univariate variables and interaction terms into the multivariate Cox regression model, and determine the final independent prognostic factors and interaction terms based on the Wald test: First, the regression coefficients of the selected single variables and interaction terms are determined based on the multivariate Cox regression model: h(t)=h0(t)exp(β1X1+β2X2+…+β i X i ) Among them, β i represents the selected single variable or interaction term, β i represents the regression coefficient of the single variable or interaction term to be determined, indicating the strength and direction of the interaction effect, h0(t) represents the baseline risk function, and h(t) is the risk function; The Wald test is then used to determine whether the selected single variables and interaction terms are significant based on the determined regression coefficients, and the final selected single variables and interaction terms are determined based on the set thresholds.

6. The method for predicting and evaluating the chemotherapy benefit of invasive micropapillary carcinoma according to claim 5, characterized in that: The risk assessment model in step 3 is: Risk Score=α1Y1+α2Y2+…+α i Y i Among them, α i is the regression coefficient of the i-th variable, Y i is the value of the variable; The coefficients in the risk assessment model were trained and adjusted based on the data set, and the risk assessment model was verified using ROC curve analysis, calibration curve and decision curve analysis.

7. The method for predicting and evaluating the chemotherapy benefit of invasive micropapillary carcinoma according to claim 6, characterized in that: In the risk prediction and assessment in step 4, a risk assessment score is determined based on the trained risk assessment model. If the risk assessment score is higher than the determined cutoff value, it is considered that chemotherapy is predicted to have a significant benefit, otherwise it is considered that chemotherapy is predicted to have no significant benefit; The cutoff value is obtained by analyzing the continuous variables in the risk assessment model based on X-tile.

8. A system for predicting and evaluating the benefit of chemotherapy for invasive micropapillary carcinoma, characterized in that: Includes the following modules: Data processing module: used to obtain chemotherapy-related data of invasive micropapillary carcinoma and pre-process the data, build a data set, and regularly update the data set; The sources of the chemotherapy-related data for invasive micropapillary carcinoma include various public databases and clinical databases; The data included clinical characteristics, case characteristics, treatment characteristics, and follow-up characteristics; The clinical characteristics include age, race, marital status, and side of invasive micropapillary carcinoma; The case characteristics included histological grade, tumor size, lymph node status, estrogen receptor status, progesterone receptor status, and human epidermal growth factor receptor 2 status; The treatment characteristics include whether the patient has received radiotherapy, surgery, or chemotherapy; The follow-up characteristics include survival status, survival time; The data preprocessing includes: Clean and standardize the data to remove missing values; Data were coded for categorical variables and standardized for continuous variables; Screening module: used to screen independent prognostic factors and interaction variables based on the data set; Risk assessment module: used to build a risk assessment model based on the screened independent prognostic factors and interaction variables, and train the model based on the data set; Benefit prediction and evaluation module: used to predict and evaluate risks based on the trained risk assessment model, and complete the prediction and evaluation of the benefits of chemotherapy for invasive micropapillary carcinoma.

9. The system for predicting and evaluating the benefit of chemotherapy for invasive micropapillary carcinoma according to claim 8, characterized in that: The screening module screens independent prognostic factors and interactive variables based on the data set, specifically: Propensity score matching and baseline correction were performed on the data to balance the baseline characteristics of the data that had undergone chemotherapy and the data that had not undergone chemotherapy. That is, based on the 1:1 neighbor matching method, each data point that had undergone chemotherapy was matched with a data point that had not undergone chemotherapy with a set caliper value, and the chi-square test was used to verify the balance of the characteristics after matching; Screening of independent prognostic factors and interaction variables, including: Based on Cox regression analysis, univariate analysis was performed to identify variables that were significantly associated with overall survival (OS). The interaction term detection method was used to introduce interaction terms between variables, and the significant interaction terms were screened by Wald test. If the empirical value was less than the set threshold, it indicated that the interaction term had a significant effect on prognosis. The screened univariate and interaction terms were included in the multivariate Cox regression model, and the final independent prognostic factors and interaction terms were determined based on the Wald test: First, the regression coefficients of the selected single variables and interaction terms are determined based on the multivariate Cox regression model: h(t)=h0(t)exp(β1X1+β2X2+…+β i X i ) Among them, β i represents the selected single variable or interaction term, β i represents the regression coefficient of the single variable or interaction term to be determined, indicating the strength and direction of the interaction effect, h0(t) represents the baseline risk function, and h(t) is the risk function; The Wald test is then used to determine whether the selected single variables and interaction terms are significant based on the determined regression coefficients, and the final selected single variables and interaction terms are determined based on the set thresholds.

10. The system for predicting and evaluating the benefit of chemotherapy for invasive micropapillary carcinoma according to claim 8, characterized in that: The risk assessment module constructs a risk assessment model based on the screened independent prognostic factors and interactive variables, and the process of training the model based on the data set is as follows: The risk assessment model is: Risk Score=α1Y1+α2Y2+…+α i Y i Among them, α i is the regression coefficient of the i-th variable, Y i is the value of the variable; The coefficients in the risk assessment model are trained and adjusted based on the data set, and the risk assessment model is verified using ROC curve analysis, calibration curve and decision curve analysis; Based on the trained risk assessment model, a risk assessment score is determined. If the risk assessment score is higher than the determined cutoff value, it is considered that chemotherapy is predicted to have a significant benefit; otherwise, it is considered that chemotherapy is predicted to have no significant benefit; The cutoff value is obtained by analyzing the continuous variables in the risk assessment model based on X-tile.

Images