Slurry ovarian cancer prognosis survival time retesting method based on Lasso regression construction

Abstract

The invention provides a serous ovarian cancer prognosis survival time back test method constructed based on Lasso regression, and relates to the technical field of biomedicine, and the serous ovarian cancer prognosis survival time back test method comprises the following steps: obtaining the prognosis survival time and final clinical factors of a serous ovarian cancer patient, associating the prognosis survival time with the final clinical factors, and constructing a sample data set, and establishing an initial model, carrying out Lasso regression test based on the test sample set, obtaining a trained back test model when the test passes, and otherwise, carrying out iterative training based on the training sample set and obtaining the back test model again. According to the method, the recognition accuracy of the back-test model can be improved, and the prognostic survival probability of the serous ovarian cancer patient is efficiently back-tested, so that a reference is provided for treatment of the serous ovarian cancer patient.

Description

Technical Field

[0001] The present invention relates to the field of biomedical technology, and in particular to a serous ovarian cancer prognosis survival period backtesting method based on Lasso regression. Background Art

[0002] Serous ovarian cancer is a malignant tumor and the most common type of epithelial ovarian cancer. Most patients with serous ovarian cancer are diagnosed at stage III (51%) and stage IV (29%), with a 5-year survival rate of 42% and 26%, respectively, and an extremely high mortality rate. Therefore, back-testing the prognosis and survival of patients with serous ovarian cancer is of great significance to improving the quality of life of patients and reducing the mortality rate.

[0003] In recent years, many researchers have applied various machine learning and statistical methods to clinical research on serous ovarian cancer, thereby establishing corresponding backtesting models to assess patient risks. For example, the filtering method evaluates the importance of each clinical factor by scoring it, sets a threshold or selects clinical factors based on the number of clinical factors, and discards clinical factor information with lower scores; the variance selection method calculates the variance of each clinical factor, and then discards clinical factors with variances below the threshold based on the threshold.

[0004] However, the above methods all establish backtest models for all clinical factors in the disease, and there are many redundant clinical factors. It is impossible to select the most important clinical factors for the backtest results according to the characteristics of different diseases, resulting in a relatively low accuracy rate of the established backtest model. Therefore, it is urgent to provide a solution to improve the above problems. Summary of the invention

[0005] The purpose of the present invention is to provide a serous ovarian cancer prognosis survival backtesting method based on Lasso regression, so as to improve the problem of low accuracy of the existing backtesting model due to excessive selection of clinical factors.

[0006] The present invention provides a serous ovarian cancer prognosis survival backtesting method based on Lasso regression, comprising the following steps:

[0007] Obtaining the prognostic survival period and final clinical factors of patients with serous ovarian cancer, and constructing a sample data set after correlating the prognostic survival period with the final clinical factors;

[0008] Build the initial model and perform LASSO regression test based on the test sample set. If the test passes, a trained backtest model is obtained. Otherwise, iterative training is performed based on the training sample set to obtain a trained backtest model.

[0009] Backtest the training sample data set based on the trained backtest model and obtain the backtest results.

[0010] The beneficial effect of the serous ovarian cancer prognosis survival backtest method provided by the present invention based on Lasso regression is that the method uses proportional risk test, univariate and multivariate regression analysis to gradually screen out the final clinical factors with higher correlation with the prognosis survival of serous ovarian cancer patients, reduce the influence of redundant factors on the accuracy of the backtest results, and optimize the coefficients of the final clinical factors by constructing a loss function, further improve the accuracy of the backtest model, and finally evaluate the performance of the backtest model from the consistency index, ensuring the feasibility of the model in clinical application, so as to facilitate doctors to provide individualized medical guidance to serous ovarian cancer patients in multiple periods such as preoperative and postoperative. In addition, screening relevant clinical factors can reduce redundant clinical factors, screen out clinical factors that have the greatest impact on the prognosis survival of serous ovarian cancer patients and have the most backtest value, prevent excessive clinical factors from causing the backtest model to overfit, and prevent the backtest effect from being poor on new data, thereby improving the accuracy and generalization ability of the backtest model.

[0011] Optionally, the process of obtaining final clinical factors for patients with serous ovarian cancer includes:

[0012] Obtaining relevant clinical factors of the serous ovarian cancer patient;

[0013] Setting a threshold probability between the relevant clinical factors and the prognostic survival period, and screening the relevant clinical factors based on the threshold probability and univariate regression to obtain preliminary screening clinical factors;

[0014] Performing probability parameter standard analysis and screening on the preliminary screening clinical factors to obtain preset clinical factors;

[0015] The pre-set clinical factors were tested for independence based on multivariate regression to obtain the final clinical factors.

[0016] Optionally, the relevant clinical factors include: age, BMI, smoking, taking oral contraceptives, number of pregnancies, number of births, number of miscarriages, whether menopausal, whether there are comorbidities, family history, whether there are complications, HE4, CA125, ROMA, tumor location, tumor stage, tumor grade, size of residual tumor after surgery, chemotherapy regimen, whether targeted drugs are used, and whether a complete chemotherapy cycle is present.

[0017] Optionally, the process of derivatizing the mathematical expression of the loss function to obtain the coefficient value of the final clinical factor corresponding to when the derivative function of the loss function is 0 includes:

[0018] The process of obtaining the backtest model to be tested after training the initial model based on the training sample set includes:

[0019] Construct a loss function, the mathematical expression of which is:

[0020]

[0021] Among them, λ is the optimal penalty factor, θ j is the coefficient of the final clinical factor, X ij represents the jth final clinical factor of the i-th serous ovarian cancer patient in the training sample set, n is the number of final clinical factors, and m is the total number of serous ovarian cancer patients in the training sample set;

[0022] The mathematical expression of the loss function is differentiated to find the coefficient value of the final clinical factor corresponding to the derivative of the loss function being 0, the training process is completed, and the backtest model to be tested is obtained based on the coefficient value of the final clinical factor.

[0023] Optionally, the process of derivatizing the mathematical expression of the loss function to obtain the coefficient value of the final clinical factor corresponding to when the derivative function of the loss function is 0 includes:

[0024] The mathematical expression of the loss function is divided into L 1 (θ) and L 2 (θ) two parts, among which,

[0025]

[0026] Where n represents the number of final clinical factors, X ij represents each final clinical factor, θ j represents the parameters of each externally verified backtest model, λ represents the penalty factor, m represents the number of training sample sets, and y i represents the true value;

[0027] For L 1 (θ) can be derived as follows:

[0028]

[0029] in,

[0030] For L 2 (θ) can be derived as follows:

[0031]

[0032] The derivative of the loss function is The superposition of the two parts can be expressed as:

[0033]

[0034] Let the derivative of the loss function be 0, and get the coefficient value θ of the final clinical factor k , whose expression is:

[0035]

[0036] Optionally, when testing the backtest model to be tested, it includes:

[0037] The final clinical factors in the test sample set are backtested based on the trained backtest model to obtain the backtest results of the serous ovarian cancer patients in the sample set to be backtested, and after comparing the backtest results of the serous ovarian cancer patients to be backtested with the corresponding true results in the test sample set, a similarity score between the backtest results of the serous ovarian cancer patients to be backtested and the true results is obtained. When the similarity score is greater than the excellent score threshold, the test passes; otherwise, the number of training sample sets is increased, iterations are continued, the backtest model is re-obtained, and the test sample set is used again for testing until the test passes.

[0038] Optionally, patients with serous ovarian cancer were screened based on inclusion and exclusion criteria, among which,

[0039] The inclusion criteria must be met: patients diagnosed through clinical symptoms, gynecological examination, pathological and cytological examinations; patients with complete medical records; patients undergoing first formal treatment; patients receiving postoperative adjuvant formal chemotherapy;

[0040] The exclusion criteria must be met: patients with other gynecological diseases and tumors of other organs and tissues; patients who have not undergone surgical treatment; patients with a history of other surgeries; patients with concurrent severe temperamental lesions and incomplete clinical data. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 This is a basic step flow chart of a serous ovarian cancer prognosis survival backtesting method based on Lasso regression provided by the present invention;

[0042] Figure 2 It is a Lasso regression diagram for screening variables provided by the present invention;

[0043] Figure 3 It is a Lasso coefficient path diagram for feature screening provided by the present invention;

[0044] Figure 4 The present invention provides a nomogram of a serous ovarian cancer prognosis survival backtest method based on Lasso regression;

[0045] Figure 5 It is a consistency index curve diagram of the training sample set and the backtest sample set provided by the present invention; DETAILED DESCRIPTION

[0046] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are 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. Unless otherwise defined, the technical terms or scientific terms used herein should be understood by people with general skills in the field to which the present invention belongs. "Including" and similar words used in this article mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects.

[0047] See also Figure 1 The embodiment of the present invention provides a serous ovarian cancer prognosis survival backtesting method based on Lasso regression, the method comprising the following steps:

[0048] S1. Obtain the prognosis survival period and final clinical factors of patients with serous ovarian cancer, and construct a sample data set after associating the prognosis survival period with the final clinical factors;

[0049] S2. Build the initial model and perform LASSO regression test based on the test sample set. If the test passes, a trained backtest model is obtained. Otherwise, iterative training is performed based on the training sample set to obtain a trained backtest model.

[0050] S3. Backtest the training sample data set based on the trained backtest model and obtain the backtest result.

[0051] In some embodiments, when executing step S1, the process of constructing the sample data set includes:

[0052] According to the inclusion and exclusion criteria, a total of 100 patients diagnosed with serous ovarian cancer in Jingdezhen Second People's Hospital from January 2016 to December 2023 were selected for inclusion in the study as the sample data set;

[0053] Specifically, the inclusion criteria must meet the following four standards: (1) patients diagnosed through clinical symptoms, gynecological examination, pathological and cytological examinations; (2) patients with complete medical records; (3) patients who received their first formal treatment in this hospital; and (4) patients who received postoperative adjuvant formal chemotherapy.

[0054] Furthermore, the exclusion criteria had to meet the following four criteria: (1) patients with other gynecological diseases and tumors of other organs and tissues; (2) patients who did not undergo surgical treatment; (3) patients with a history of other surgeries; and (4) patients with concurrent severe temperamental lesions and incomplete clinical data.

[0055] In some embodiments, when executing step S1, the process of obtaining the final clinical factors of the serous ovarian cancer patient includes:

[0056] S1-1. Obtain relevant clinical factors of the patient with serous ovarian cancer;

[0057] S1-2, setting a threshold probability of the relevant clinical factors and the prognosis survival period, and screening the relevant clinical factors based on the threshold probability and univariate regression to obtain preliminary screening clinical factors;

[0058] S1-3, performing probability parameter standard analysis and screening on the preliminary screening clinical factors to obtain preset clinical factors;

[0059] S1-4. Perform independence test on the preset clinical factors based on multivariate regression to obtain the final clinical factors.

[0060] Specifically, the relevant clinical factors in step S1-1 refer to the factors that may affect the prognosis and survival of patients with serous ovarian cancer clinically, including: age, BMI, smoking, oral contraceptives, number of pregnancies, number of births, number of miscarriages, menopause, comorbidities, family history, complications, HE4, CA125, ROMA, tumor location, tumor stage, tumor grade, size of residual tumor after surgery, chemotherapy regimen, whether targeted drugs are used, and whether a complete chemotherapy cycle is present. Table 1 shows a quantitative scheme for relevant clinical factors.

[0061] Table 1 Quantification scheme of relevant clinical factors

[0062]

[0063]

[0064] Specifically, when executing step S1-2, the steps for obtaining the initial screening clinical factors are as follows:

[0065] The threshold probability of the initial inclusion of clinical factors was set at α = 12%, and the P value of each relevant clinical factor was calculated. The relevant clinical factors corresponding to P ≤ α were screened out from the relevant clinical factors as the initial screening clinical factors; the initial screening clinical factors were HE4, ROAM, tumor stage, tumor grade, postoperative residual tumor size, whether targeted drugs were used, and whether there was a complete chemotherapy cycle. Table 2 shows the quantitative scheme of the initial screening clinical factors.

[0066] Table 2 Quantification scheme of clinical factors for initial screening

[0067]

[0068] Further, in step S1-3, the probability parameter standard analysis and screening of the primary screening clinical factors are performed, and the screening variable results are as follows: Figure 2 As shown, the process of obtaining pre-specified clinical factors includes:

[0069] After performing a proportional risk test on the initial clinical factors, the P value of each initial screening clinical factor was calculated. The initial screening clinical factor ROMA with P>0.05 was eliminated, and the remaining 6 factors were the preset clinical factors, which were HE4, tumor stage, tumor grade, postoperative residual tumor size, whether targeted drugs were used, and whether there was a complete chemotherapy cycle.

[0070] Furthermore, the Lasso regression characteristic screening results are as follows Figure 3 As shown, the process of obtaining the final clinical factors in executing step S1-4 includes:

[0071] The pre-set clinical factors were tested for independence based on multivariate regression, and the final clinical factors were obtained after eliminating the associated pre-set clinical factors. Since each pre-set clinical factor was independent of each other, the final clinical factors were still: HE4, tumor stage, tumor grade, postoperative residual tumor size, whether targeted drugs were used, and whether a complete chemotherapy cycle was present.

[0072] In some embodiments, the process of executing step S2 to perform iterative training based on the training sample set includes a cross-validation process and a loss function construction process; in the cross-validation process, the optimal penalty parameter value is determined through error analysis, and in the process of constructing the loss function, the coefficient value of the final clinical factor corresponding to the minimum loss function is obtained based on the optimal penalty parameter value, and a trained backtest model is constructed based on the coefficient value.

[0073] Furthermore, the process of obtaining the coefficient value of the final clinical factor corresponding to the minimum loss function based on the optimal penalty parameter value includes:

[0074] A loss function is constructed based on the optimal penalty parameter value, and the mathematical expression of the loss function is:

[0075]

[0076] Among them, λ is the optimal penalty factor, θ j is the coefficient of the final clinical factor, X ij Indicates the first iThe jth final clinical factor of the patients with serous ovarian cancer, n is the number of final clinical factors, and m is the total number of patients with serous ovarian cancer in the training sample set;

[0077] The mathematical expression of the loss function is derived to find the coefficient value of the final clinical factor corresponding to when the derivative function of the loss function is 0.

[0078] Furthermore, the mathematical expression of the loss function is divided into L 1 (θ) and L 2 (θ) two parts, among which,

[0079]

[0080] Where n represents the number of final clinical factors, X ij represents each final clinical factor, θ j represents the parameters of each externally verified backtest model, λ represents the penalty factor, m represents the number of training sample sets, and y i represents the true value;

[0081] For L 1 (θ) can be derived as follows:

[0082]

[0083] in,

[0084] For L 2 (θ) can be derived as follows:

[0085]

[0086] Furthermore, the derivative of the loss function is The superposition of the two parts can be expressed as:

[0087]

[0088] Furthermore, the derivative of the loss function is set to 0, and the coefficient value of the final clinical factor is obtained as θ k , whose expression is:

[0089]

[0090] Further, based on the coefficient θ k Construct a trained backtest model, the mathematical expression of the trained backtest model is:

[0091] Y=θ 1 X 1 +θ 2 X 2 +θ3 X 3 +....+θ 6 X 6 ;

[0092] Among them, Y is the future survival period of patients with serous ovarian cancer, X is 1 , X 2 , …, X 6 represents the final clinical factors, namely: HE4, tumor stage, tumor grade, postoperative residual tumor size, whether targeted drugs are used, and whether there is a complete chemotherapy cycle, θ 1 ,θ 2 ,…θ 6 Represents the parameters of the externally validated backtest model.

[0093] In some embodiments, the process of executing step S3 to obtain the backtest result includes: drawing a nomogram according to the trained backtest model mathematical expression and the contribution of each final clinical factor to the backtest result, calculating the score of the nomogram corresponding to the value of each final clinical factor and performing weighted summation to obtain a total score, obtaining the survival rate of serous ovarian cancer patients according to the total score, and obtaining the backtest result.

[0094] Specifically, Figure 4 As shown, each final clinical factor is mapped to the top score of the nomogram to obtain the corresponding score of each final clinical factor, and the score corresponding to each final clinical factor is accumulated and mapped to the total score, and the total score is mapped downward to obtain the one-year, three-year, and five-year survival rates of patients with serous ovarian cancer.

[0095] In some embodiments, when executing step S3, the process of backtesting the test sample set includes the following steps:

[0096] The final clinical factors in the test sample set are backtested based on the trained backtest model to obtain the backtest results of the serous ovarian cancer patients to be backtested, and after comparing the backtest results of the serous ovarian cancer patients to be backtested, a similarity score curve between the backtest results of the serous ovarian cancer patients to be backtested and the true results is obtained; when the similarity score is greater than the excellent score threshold, the backtest is passed.

[0097] Furthermore, it can be seen from the similarity curve that the test passes and the backtest results can be obtained based on the trained backtest model.

[0098] Specifically, the consistency index is used to estimate the similarity between the real result score and the backtest result score ranking. Its theoretical value is between 0 and 1. When the consistency index is > 0.5, it means that the trained backtest model has better performance. The calculated consistency index curve of the training sample set is as follows: Figure 5As shown, the results show that the consistency index of the training sample set is greater than 0.5, which means that the trained backtest model fits the training sample set well.

[0099] The Chinese and English explanations corresponding to the drawings of the present invention are as follows:

[0100] Points: The score corresponding to each final clinical factor

[0101] HE4: human epididymis protein 4

[0102] Stage: Tumor stage

[0103] Class: Tumor grade

[0104] Size: The size of the residual tumor after surgery

[0105] Target: Whether to use targeted drugs

[0106] Course: Is there a complete chemotherapy cycle?

[0107] Total Points: Total score

[0108] 1-year Survival: One-year survival rate

[0109] 3-year Survival: Three-year survival rate

[0110] 5-year Survival: Five-year survival rate

[0111] Partial likelihood deviance: partial likelihood deviation

[0112] Coefficient: coefficient

[0113] Concordance index: consistency index

[0114] variables: clinical factor variables

[0115] Months: Month

[0116] Although the embodiments of the present invention are described in detail above, it is obvious to those skilled in the art that various modifications and variations can be made to these embodiments. However, it should be understood that such modifications and variations are within the scope and spirit of the present invention as described in the claims. Moreover, the present invention described herein may have other embodiments and may be implemented or realized in a variety of ways.

Claims

1. A serous ovarian cancer prognosis survival backtesting method based on Lasso regression, characterized in that: include: Obtaining the prognostic survival period and final clinical factors of patients with serous ovarian cancer, and constructing a sample data set after correlating the prognostic survival period with the final clinical factors; Build the initial model and perform Lasso regression test based on the test sample set. If the test passes, a trained backtest model is obtained. Otherwise, iterative training is performed based on the training sample set to obtain a trained backtest model. Backtest the training sample data set based on the trained backtest model and obtain the backtest results.

2. According to claim 1, a serous ovarian cancer prognosis survival backtesting method based on Lasso regression is characterized in that: The process of obtaining the final clinical factors for patients with serous ovarian cancer includes: Obtaining relevant clinical factors of the serous ovarian cancer patient; Setting a threshold probability between the relevant clinical factors and the prognostic survival period, and screening the relevant clinical factors based on the threshold probability and univariate regression to obtain preliminary screening clinical factors; Performing probability parameter standard analysis and screening on the preliminary screening clinical factors to obtain preset clinical factors; The pre-set clinical factors were tested for independence based on multivariate regression to obtain the final clinical factors.

3. The method for backtesting the prognosis and survival period of serous ovarian cancer based on Lasso regression according to claim 2, characterized in that: The relevant clinical factors include: age, BMI, smoking, taking oral contraceptives, number of pregnancies, number of births, number of miscarriages, menopause, comorbidities, family history, complications, HE4, CA125, ROMA, tumor location, tumor stage, tumor grade, size of residual tumor after surgery, chemotherapy regimen, whether targeted drugs are used, and whether a complete chemotherapy cycle is present.

4. The method for backtesting the prognosis and survival period of serous ovarian cancer based on Lasso regression according to claim 1, characterized in that: The process of obtaining the backtest model to be tested after training the initial model based on the training sample set includes: Construct a loss function, the mathematical expression of which is: Among them, λ is the optimal penalty factor, θ j is the coefficient of the final clinical factor, X ij represents the jth final clinical factor of the i-th serous ovarian cancer patient in the training sample set, n is the number of final clinical factors, and m is the total number of serous ovarian cancer patients in the training sample set; The mathematical expression of the loss function is differentiated to find the coefficient value of the final clinical factor corresponding to the derivative of the loss function being 0, the training process is completed, and the backtest model to be tested is obtained based on the coefficient value of the final clinical factor.

5. The method for backtesting the prognosis and survival period of serous ovarian cancer based on Lasso regression according to claim 4, characterized in that: The process of derivatizing the mathematical expression of the loss function to obtain the coefficient value of the final clinical factor corresponding to when the derivative function of the loss function is 0 includes: The mathematical expression of the loss function is divided into two parts: L1(θ) and L2(θ), where: Where n represents the number of final clinical factors, X ij represents each final clinical factor, θ j represents the parameters of each externally verified backtest model, λ represents the penalty factor, m represents the number of training sample sets, and y i represents the true value; Taking the derivative of L1(θ) we get: in, Taking the derivative of L2(θ) we get: The derivative of the loss function is The superposition of the two parts can be expressed as: Let the derivative of the loss function be 0, and get the coefficient value θ of the final clinical factor k , whose expression is:

6. The method for backtesting the prognosis and survival period of serous ovarian cancer based on Lasso regression according to claim 1, characterized in that: The expression of the trained backtest model is: Y=θ1X i1 +θ2X i2 +θ3X i3 +....+θ k X ik ; Among them, Y is the future survival period of patients with serous ovarian cancer, X is i1 , X i2 , …, X ik Represents each final clinical factor, θ1, θ2, …θ k represents the parameters of the externally validated backtest model, and k represents the number of final clinical factors.

7. The method for backtesting the prognosis and survival period of serous ovarian cancer based on Lasso regression according to claim 1, characterized in that: When testing the backtest model to be tested, it includes: The final clinical factors in the test sample set are backtested based on the trained backtest model to obtain the backtest results of the serous ovarian cancer patients in the sample set to be backtested, and after comparing the backtest results of the serous ovarian cancer patients to be backtested with the corresponding true results in the test sample set, a similarity score between the backtest results of the serous ovarian cancer patients to be backtested and the true results is obtained. When the similarity score is greater than the excellent score threshold, the test passes; otherwise, the number of training sample sets is increased, iterations are continued, the backtest model is re-obtained, and the test sample set is used again for testing until the test passes.

8. The method for backtesting the prognosis and survival period of serous ovarian cancer based on Lasso regression according to claim 1, characterized in that: Patients with serous ovarian cancer were screened based on the inclusion and exclusion criteria, among which, The inclusion criteria must be met: patients diagnosed through clinical symptoms, gynecological examination, pathological and cytological examinations; patients with complete medical records; patients undergoing first formal treatment; patients receiving postoperative adjuvant formal chemotherapy; The exclusion criteria must be met: patients with other gynecological diseases and tumors of other organs and tissues; patients who have not undergone surgical treatment; patients with a history of other surgeries; patients with concurrent severe temperamental lesions and incomplete clinical data.

Images