Prediction method, device and equipment for adverse reaction of chemotherapeutic drug and medium

Abstract

The invention provides a chemotherapeutic drug adverse reaction prediction method, device, equipment and medium, and relates to the technical field of medical information, and the method comprises the steps: obtaining a historical case data set; a plurality of adverse reaction measured values are obtained through calculation; obtaining an initial prediction model; obtaining an expected value prediction sub-model; obtaining a standard deviation prediction sub-model; acquiring chemotherapy data of the target case; and obtaining an untoward effect expected value of the target case and a corresponding standard deviation. According to the method, the adverse reaction evaluation indexes are divided into three types of myelosuppression, hepatotoxicity and kidney toxicity, the numerical value and the change value of each index are comprehensively considered, and the adverse reaction evaluation method is constructed, so that accurate quantitative description is obtained; the adverse reaction prediction model covering the expected value and the standard deviation is constructed by dividing the influence factors of the adverse reaction into the chemotherapy drugs, the chemotherapy cycle and the influence factors, so that the expected adverse reaction probability distribution can be accurately described, and a technical support is provided for quantitative analysis and prediction of the adverse reaction.

Description

Technical Field

[0001] The present invention relates to the field of medical information technology, and in particular to a method, device, equipment and medium for predicting adverse reactions of chemotherapy drugs. Background Art

[0002] The adverse reactions of chemotherapy drugs mainly manifest in bone marrow suppression, liver toxicity, gastrointestinal adverse reactions, renal toxicity and pulmonary toxicity, which are related to the drug itself, as well as the patient's body and other factors. There are individual differences in adverse drug reactions. The adverse reactions of the same drug in different individuals are diverse, and the same adverse reaction can also be caused by different drugs. At present, there is a lack of systematic and complete analysis of the individualization of chemotherapy drug dosages, especially the research on the causes and influencing factors of adverse reactions is too simple, which makes doctors lack foresight of adverse reactions in clinical medication, and cannot achieve precise control of chemotherapy drugs, affecting the quality of life of patients. Summary of the invention

[0003] The purpose of the present invention is to provide a method, device, equipment and medium for predicting adverse reactions of chemotherapeutic drugs to improve the above problems. In order to achieve the above purpose, the technical solution adopted by the present invention is as follows:

[0004] In a first aspect, the present application provides a method for predicting adverse reactions of chemotherapeutic drugs, comprising:

[0005] Acquire a historical case data set, wherein the historical case data set includes chemotherapy data and adverse reaction evaluation data of multiple historical cases, wherein the chemotherapy data includes chemotherapy drugs, chemotherapy cycles, and patient body conditions, and the adverse reaction evaluation data includes bone marrow suppression indicators, liver toxicity indicators, and kidney toxicity indicators;

[0006] Based on multiple adverse reaction assessment data, multiple adverse reaction measured values ​​are calculated;

[0007] Initializing a preset prediction model based on the historical case data set to obtain an initial prediction model;

[0008] Based on the historical case data set, the initial prediction model is trained, and when the preset first loss function meets the set condition, the training is stopped to obtain the expected value prediction sub-model;

[0009] Based on the historical case data set and the expected value prediction sub-model, a second loss function is constructed to obtain an optimal solution and a standard deviation prediction sub-model is obtained;

[0010] Obtain chemotherapy datasets for target cases;

[0011] The chemotherapy data of the target case is input into an adverse reaction prediction model to obtain the expected value of adverse reactions of the target case and the corresponding standard deviation. The adverse reaction prediction model is composed of the expected value prediction sub-model and the standard deviation prediction sub-model.

[0012] In a second aspect, the present application also provides a device for predicting adverse reactions of chemotherapeutic drugs, comprising:

[0013] A first acquisition unit is used to acquire a historical case data set, wherein the historical case data set includes chemotherapy data and adverse reaction evaluation data of multiple historical cases, wherein the chemotherapy data includes chemotherapy drugs, chemotherapy cycles and patient body conditions, and the adverse reaction evaluation data includes bone marrow suppression indicators, liver toxicity indicators and kidney toxicity indicators;

[0014] A first calculation unit, configured to calculate a plurality of adverse reaction measured values ​​based on a plurality of adverse reaction assessment data;

[0015] An initialization unit, used to initialize a preset prediction model based on the historical case data set to obtain an initial prediction model;

[0016] A first training unit is used to train the initial prediction model based on the historical case data set, and when a preset first loss function meets a set condition, stop the training to obtain an expected value prediction sub-model;

[0017] A second training unit is used to construct a second loss function based on the historical case data set and the expected value prediction sub-model, find the optimal solution, and obtain a standard deviation prediction sub-model;

[0018] A second acquisition unit is used to acquire a chemotherapy data set of a target case;

[0019] An input unit is used to input the chemotherapy data of the target case into an adverse reaction prediction model to obtain the expected value of the adverse reaction of the target case and the corresponding standard deviation. The adverse reaction prediction model is composed of the expected value prediction sub-model and the standard deviation prediction sub-model.

[0020] In a third aspect, the present application also provides a device for predicting adverse reactions of chemotherapy drugs, comprising:

[0021] Memory for storing computer programs;

[0022] A processor is used to implement the steps of the method for predicting adverse reactions of chemotherapy drugs when executing the computer program.

[0023] In a fourth aspect, the present application further provides a readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the above-mentioned method for predicting adverse reactions of chemotherapy drugs are implemented.

[0024] The beneficial effects of the present invention are:

[0025] The present invention divides the adverse reaction evaluation indicators into three categories: bone marrow suppression, liver toxicity and kidney toxicity, and comprehensively considers the numerical value and change value of each indicator to construct an adverse reaction evaluation method, so that it can be accurately quantitatively described; and by dividing the influencing factors of adverse reactions into three categories: chemotherapy drugs, chemotherapy cycles and influencing factors, an adverse reaction prediction model covering expected values ​​and standard deviations is constructed, which can accurately describe the expected probability distribution of adverse reactions and provide technical support for the quantitative analysis and prediction of adverse reactions.

[0026] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or be understood by implementing the embodiments of the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description, claims, and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0028] Figure 1 Schematic diagram of the process of predicting adverse reactions of chemotherapy drugs described in an embodiment of the present invention;

[0029] Figure 2 Schematic diagram of the structure of the device for predicting adverse reactions of chemotherapy drugs described in an embodiment of the present invention;

[0030] Figure 3 Schematic diagram of the structure of the device for predicting adverse reactions of chemotherapy drugs described in an embodiment of the present invention.

[0031] Markings in the figure:

[0032] 10. First acquisition unit; 20. First calculation unit; 30. Initialization unit; 40. First training unit; 50. Second training unit; 60. Second acquisition unit; 70. Input unit; 800. Device for predicting adverse reactions of chemotherapy drugs; 801. Processor; 802. Memory; 803. Multimedia component; 804. I / O interface; 805. Communication component. DETAILED DESCRIPTION

[0033] 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 in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0034] It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in the subsequent drawings. At the same time, in the description of the present invention, the terms "first", "second", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0035] Embodiment 1:

[0036] This embodiment provides a method for predicting adverse reactions of chemotherapy drugs.

[0037] See also Figure 1 , the figure shows that the method includes step S1, step S2, step S3, step S4, step S5, step S6 and step S7.

[0038] Step S1. Acquire a historical case data set, the historical case data set includes chemotherapy data and adverse reaction evaluation data of multiple historical cases, the chemotherapy data includes chemotherapy drugs, chemotherapy cycles and patient body conditions, and the adverse reaction evaluation data includes bone marrow suppression index, liver toxicity index and kidney toxicity index;

[0039] Specifically, considering that the adverse reactions caused by chemotherapy are related to chemotherapy drugs, chemotherapy cycles and influencing factors of the patient's body, the adverse reaction prediction model established in this application is also established based on the above three categories of influencing factors.

[0040] For chemotherapy drug data, including chemotherapy regimens and dosage. Considering that the chemical composition, chemical structure, physicochemical properties and mechanism of action of drugs are different, their therapeutic effects are different from therapeutic effects, and the manifestations of adverse reactions are also different. Most current chemotherapy regimens use combined medication, which also has a significant impact on adverse reactions due to the interaction between drugs. Therefore, the adverse reaction prediction model constructed by this application should correspond to a specific chemotherapy regimen one by one. The present invention only considers constructing a prediction model under the premise of the same chemotherapy regimen, so the dosage is closely related to adverse reactions. Generally, the higher the dosage, the stronger the efficacy, and the more serious the adverse reactions. Therefore, this application uses the dosage as one of the input variables of the prediction model.

[0041] Regarding chemotherapy cycle data, usually as patients undergo multiple cycles of chemotherapy, their body immunity tends to decrease, and adverse reactions at the same medication dose tend to increase. Therefore, this application uses the chemotherapy cycle as one of the input variables of the prediction model.

[0042] The factors affecting the patient's physical condition include six factors, including gene polymorphism, gender, age, physical condition, nutritional status and pathological status of the body. Gene polymorphism refers to the differences in genotype and phenotype between different individuals of the same gene, which may affect the metabolism and treatment of chemotherapeutic drugs, thereby causing different adverse reactions when the same conditions are used, especially for special genotypes that are clearly required to be noted in the use of drugs. Gender factors have a certain influence on the metabolism and effect of drugs. Generally, women are more sensitive to drugs and are more prone to adverse reactions. People of different age groups have their own physiological characteristics, and the absorption, distribution, metabolism and excretion of drugs are different. At the same time, the elderly have a tendency to have many diseases, many medications and poor nutritional status. Children have slow metabolism and excretion of drugs due to the incomplete development of organs such as liver and kidney. These two groups of people are more prone to adverse reactions. Physical condition is an indicator of the patient's general health status and tolerance based on his physical strength. Generally, the lower the physical condition, the more serious the adverse reactions. Nutritional status refers to an individual's intake, absorption, metabolism and utilization of food nutrients. Long-term poor nutritional status can lead to decreased microsomal enzyme activity in liver cells, slower drug metabolism, and more likely to cause adverse reactions. The pathological state of the body refers to the signs of disease in the patient's body before chemotherapy, such as liver and kidney dysfunction, which will cause slower drug excretion and longer half-life, making adverse reactions more likely. This application uses the above six factors as model input variables.

[0043] Step S2. Based on multiple adverse reaction assessment data, multiple adverse reaction measured values ​​are calculated;

[0044] Specifically, in this application, the evaluation indicators are divided into three categories: bone marrow suppression, hepatotoxicity and renal toxicity. Bone marrow suppression indicators include neutrophil content, hemoglobin content and platelet content, hepatotoxicity indicators include alanine aminotransferase content, aspartate aminotransferase content and bilirubin content, and renal toxicity indicators include urea nitrogen content and creatinine content. Comprehensively considering the changes in the values ​​of each indicator before and after chemotherapy, a continuous distribution value between 0 and 1 is obtained, which can accurately and quantitatively evaluate the degree of adverse reactions.

[0045] Specifically, step S2 specifically includes step S21, step S22, step S23, step S24, step S25, step S26 and step S27:

[0046] Step S21. Obtaining the neutrophil content, hemoglobin content, platelet content, alanine aminotransferase content, aspartate aminotransferase content, bilirubin content, urea nitrogen content, and creatinine content of historical cases before chemotherapy from the evaluation data as the first former index, the second former index, the third former index, the fourth former index, the fifth former index, the sixth former index, the seventh former index, and the eighth former index, respectively;

[0047] Step S22. Obtaining the neutrophil content, hemoglobin content, platelet content, alanine aminotransferase content, aspartate aminotransferase content, bilirubin content, urea nitrogen content, and creatinine content of historical cases after chemotherapy from the evaluation data, as the first post-indicator, the second post-indicator, the third post-indicator, the fourth post-indicator, the fifth post-indicator, the sixth post-indicator, the seventh post-indicator, and the eighth post-indicator, respectively;

[0048] Step S23. Obtaining the neutrophil content, hemoglobin content, platelet content, alanine aminotransferase content, aspartate aminotransferase content, bilirubin content, urea nitrogen content and creatinine content in a normal state of a human body as the first indicator, the second indicator, the third indicator, the fourth indicator, the fifth indicator, the sixth indicator, the seventh indicator and the eighth indicator, respectively;

[0049] Step S24. Calculate the bone marrow suppression assessment value based on the first index before, the second index before, the third index before, the first index after, the second index after, the third index after, the first index, the second index, the third index and the first weight adjustment coefficient;

[0050] Specifically, bone marrow suppression is an adverse reaction of the hematopoietic system caused by chemotherapy, with neutropenia being the most common, followed by thrombocytopenia, and in severe cases, hemoglobin will also be affected. In this application, the three indicators of neutrophil content, hemoglobin content and platelet content are used as the basis, and a continuous value between 0 and 1 is used to represent the degree of bone marrow suppression. The larger the value, the more severe the adverse reaction.

[0051] Specifically, step S24 specifically includes step S241, step S242, step S243 and step S244:

[0052] Step S241. Calculate the neutrophil evaluation value based on the first front index, the first rear index, the first index, the first weight adjustment coefficient and the preset first normalization parameter;

[0053] Specifically, step S241 specifically includes step S2411, step S2412, step S2413, step S2414 and step S2415;

[0054] Step S2411. Calculate the difference between the first index and the previous first index and the next first index, respectively, to obtain a first difference and a second difference;

[0055] Step S2412. Calculate the ratios of the first difference and the second difference to the preset first normalization parameter to obtain a first ratio and a second ratio;

[0056] Step S2413. Calculate the product of the first ratio and the first weight adjustment coefficient to obtain a first product;

[0057] Step S2414: Calculate the difference between the preset fixed value and the first weight adjustment coefficient, and multiply the difference by the second ratio to obtain a second product; wherein the preset fixed value is 1;

[0058] Step S2415. Calculate the sum of the first product and the second product to obtain a neutrophil evaluation value.

[0059] Step S242. Calculate the hemoglobin evaluation value based on the second index, the front second index, the rear second index, the first weight adjustment coefficient and the preset second normalization parameter;

[0060] Specifically, the calculation method of the hemoglobin evaluation value is consistent with the above-mentioned neutrophil evaluation method.

[0061] Step S243. Calculate the platelet evaluation value based on the third indicator, the front third indicator, the rear third indicator, the first weight adjustment coefficient and the preset third normalization parameter;

[0062] Specifically, the calculation method of the platelet evaluation value is consistent with the above-mentioned neutrophil evaluation method.

[0063] Step S244. Determine the maximum value from the neutrophil evaluation value, the hemoglobin evaluation value and the platelet evaluation value as the bone marrow suppression evaluation value;

[0064] Specifically, remember is the neutrophil content, hemoglobin content and platelet content in the normal state of the human body. For the same case, the above indicators before and after chemotherapy are [a0,1 ,a 0,2 ,a 0,3 ] and [a 1,1 ,a 1,2 ,a 1,3 ], the bone marrow suppression assessment value calculation formula is:

[0065]

[0066] Among them, Λ i is the normalization parameter of the ith indicator; a i is the value of the ith indicator under normal human condition; a 0,i is the i-th index value before chemotherapy; a 1,i is the ith index value after chemotherapy; r1 is the first weight adjustment coefficient; A is the bone marrow suppression assessment value.

[0067] In this application, the value of r1 is set to 0.6 based on clinical data, and the common adverse event evaluation criteria (CTCAE) proposed by the National Cancer Institute (NCI) of the United States is used. and The values ​​are 2.5×10 9 / L, 125g / L and 1.25×10 11 / L, Λ1, Λ2 and Λ3 are 2.5×10 9 / L, 75g / L and 1.25×10 11 / L.

[0068] Step S25. Calculate the liver toxicity assessment value based on the front fourth index, the front fifth index, the front sixth index, the rear fourth index, the rear fifth index, the rear sixth index, the fourth index, the fifth index, the sixth index and the second weight adjustment coefficient;

[0069] Specifically, liver toxicity is an adverse metabolic reaction caused by chemotherapy, which is manifested as elevated transaminase, elevated bilirubin, chemical hepatitis, etc. This application uses three indicators, namely, alanine aminotransferase content, aspartate aminotransferase content, and bilirubin content, as the basis, and uses a continuous value between 0 and 1 to represent the degree of liver toxicity. The larger the value, the more severe the adverse reaction.

[0070] remember is the level of alanine aminotransferase, aspartate aminotransferase and bilirubin in the normal state of the human body. For the same case, the above indicators before and after chemotherapy are [b 0,1 ,b 0,2 ,b 0,3 ] and [b 1,1 ,b 1,2 ,b 1,3 ], the liver toxicity assessment value calculation formula is:

[0071]

[0072] in, is the jth index value under normal human condition; b 0,j is the jth index value before chemotherapy; b 1,j is the jth index value after chemotherapy; r2 is the second weight adjustment coefficient; B is the liver toxicity assessment value.

[0073] In this application, the r2 value is set as 0.6 based on clinical data, and the common adverse event evaluation criteria (CTCAE) proposed by the National Cancer Institute (NCI) of the United States is used. and The values ​​are 20.0U / L, 20.0U / L and 11.1μmol / L respectively.

[0074] Step S26. Based on the first seventh index, the first eighth index, the last seventh index, the last eighth index, the seventh index, the eighth index and the third weight adjustment coefficient, a kidney toxicity assessment value is calculated;

[0075] Specifically, renal toxicity is an adverse reaction to the urinary system caused by chemotherapy. Mild damage may have no obvious symptoms, while severe cases may manifest as oliguria, anuria, hematuria, acute necrosis of renal tubular epithelial cells and uremia. This application uses urea nitrogen content and creatinine content as the two indicators, and uses a continuous value between 0 and 1 to represent the degree of renal toxicity. The larger the value, the more severe the adverse reaction.

[0076] remember is the urea nitrogen content and creatinine content in the normal state of the human body. For the same case, the above indicators before and after chemotherapy are [c 0,1 ,c 0,2 ] and [c 1,1 ,c 1,2 ], the calculation formula for the renal toxicity assessment value is:

[0077]

[0078] in, is the kth index value under normal human condition; c 0,k is the kth index value before chemotherapy; c 1,k is the kth index value after chemotherapy; r3 is the third weight adjustment coefficient; C is the renal toxicity assessment value;

[0079] In this application, the r3 value is set to 0.6 based on clinical data, and the common adverse event evaluation criteria (CTCAE) proposed by the National Cancer Institute (NCI) of the United States is used. and The values ​​are 3.6mmol / L and 50.8μmol / L respectively.

[0080] Step S27. Determine the maximum value from the bone marrow suppression assessment value, the liver toxicity assessment value and the kidney toxicity assessment value as the adverse reaction actual measured value;

[0081] Specifically, for any case, the adverse reaction of chemotherapy is defined as the maximum value of the bone marrow suppression assessment value, the liver toxicity assessment value and the kidney toxicity assessment value, that is, the actual measured value of the adverse reaction of chemotherapy is:

[0082] Y=max{A,B,C}

[0083] Among them, Y is the actual measured value of adverse reaction; A is the bone marrow suppression assessment value; B is the liver toxicity assessment value; C is the kidney toxicity assessment value.

[0084] Step S3. Initialize the preset prediction model based on the historical case data set to obtain an initial prediction model;

[0085] Specifically, considering that when various influencing factors are the same, the adverse reactions produced by different individuals satisfy a certain probability distribution law, in order to reasonably simplify the model, this application assumes that the adverse reactions under the same influencing factors satisfy a normal distribution.

[0086] Let y be the expected value of the normal distribution of adverse reactions, and θ be the standard deviation of the normal distribution of adverse reactions. For the same chemotherapy regimen, the following adverse reaction prediction model is constructed, which consists of an expected value prediction sub-model and a standard deviation prediction sub-model:

[0087]

[0088] The former represents the expected value prediction submodel, and the latter represents the standard deviation prediction submodel; y is the expected value of the normal distribution of adverse reactions, θ is the standard deviation of the normal distribution of adverse reactions; x0 is the special genotype number, is the additional influencing parameter of the gene polymorphism corresponding to the special genotype number x0; x1 is the dosage; α1 and β1 are the influencing parameters of the dosage; x2 is the chemotherapy cycle; α2 and β2 are the influencing parameters of the chemotherapy cycle; x3 is gender, -1 represents male, 1 represents female; α3 is the influencing parameter of gender; x4 is age; x5 is physical condition; x6 is nutritional status; x7 is pathological state of the body; α i and β i is the influencing parameter corresponding to the i-th influencing factor, and the values ​​of i are 4, 5, 6 and 7; is the mean of the i-th influencing factor, i takes values ​​of 4, 5, 6 and 7; λ0 is the first standard deviation parameter; λ1 is the second standard deviation parameter; λ2 is the third standard deviation parameter.

[0089] The value of x0 ranges from 0 to K, where K is the total number of special genotypes that need to be considered for the current treatment plan. Among them, 0 means that the current case has no special genotype, and δ0 is always 0, indicating that the case is not affected by the additional effects of gene polymorphism. If the genotypes that need to be paid attention to are not clearly required in the chemotherapy drug use guidelines, this model does not need to consider the impact of gene polymorphism.

[0090] x1 represents the ratio of the dose per unit body surface area to the standard dose, and its value range is a continuous value between 0 and 1, where the body surface area is calculated based on two parameters: the patient's height and weight.

[0091] x5 is the physical condition. This application adopts the internationally commonly used KPS scoring method, with a value range of integers between 0 and 100. The evaluation criteria are shown in Table 1:

[0092] Table 1

[0093] standard Points Normal, no symptoms and signs 100 Able to carry out normal activities with mild symptoms and signs 90 Barely able to carry out normal activities, with some symptoms or signs 80 Can take care of themselves, but cannot maintain a normal life and work 70 Able to take care of most of life by oneself, but occasionally needs help from others 60 Often need care 50 Unable to take care of themselves and need special care and assistance 40 Severely unable to take care of oneself 30 Serious illness, requiring hospitalization and aggressive supportive care 20 Seriously ill, close to death 10 die 0

[0094] x6 is the nutritional status, which is converted according to the body mass index (BMI) in the present invention, and the value range is a continuous value from 0 to 1. The larger the value, the better the nutritional status. The calculation method of the nutritional status is:

[0095]

[0096] Among them, x6 is the nutritional status; v is the BMI value of the case.

[0097] x7 is the pathological state of the body. Similar to the evaluation method of adverse reactions, this application also divides the determination factors into three categories: bone marrow suppression, liver toxicity and kidney toxicity. The calculation method is as follows:

[0098]

[0099] in, is the content of the ith indicator related to the bone marrow suppression assessment value in the normal state of the human body; a 0,i is the content of the ith index related to the bone marrow suppression assessment value before chemotherapy; is the content of the jth index related to the liver toxicity assessment value under normal human conditions; b 0,j is the content of the jth index related to the liver toxicity assessment value before chemotherapy; is the content of the kth index related to the renal toxicity assessment value under normal human conditions; c 0,k is the content of the kth index related to the renal toxicity assessment value before chemotherapy.

[0100] {λ0,λ1,λ2} is used to express the relationship between the expected value and the standard deviation. The smaller the standard deviation, the smaller the difference between the actual adverse reaction value and the expected value, and the higher the stability of the constructed model.

[0101] The historical case data set containing N data is Among them, x i , is the chemotherapy data of the i-th historical case data; is the measured value of adverse reactions in the i-th historical case data. The first loss function of the above model is defined as:

[0102]

[0103] Among them, J is the first loss function; y i is the expected value of adverse reactions of the i-th historical case data; is the measured value of the adverse reaction of the i-th historical case data; N is the number of case data contained in the historical case data set Ψ; i≤N.

[0104] The initial values ​​of the above-mentioned influencing parameters play an important role in solving the prediction model parameters. The appropriate initial values ​​can not only avoid falling into local extreme values, but also improve the calculation efficiency. The importance and correlation analysis of each influencing factor of adverse reactions are carried out. The present invention determines the initial values ​​one by one according to the order of the influencing parameters of the dosage, the influencing parameters of the chemotherapy cycle, the additional influencing parameters of the gene polymorphism, the influencing parameters of gender and other influencing parameters, and makes the initial values ​​as close to the optimal value as possible.

[0105] Specifically, step S3 specifically includes step S31, step S32, step S33, step S34, step S35, step S36, step S37 and step S38:

[0106] Step S31. Determine a first data subset from the historical case data set, wherein the chemotherapy cycle corresponding to the first data subset is a first preset value, and the corresponding special genotype serial number is a second preset value;

[0107] Step S32. Simplify the prediction model based on the first data subset to obtain a first model, where the first model only considers the effect of the dosage in the chemotherapy drug data;

[0108] Specifically, for the calculation of dose-effect parameters, a subset Ψ'1 is selected from the data set Ψ, and the chemotherapy cycle x of the jth historical case data in the subset Ψ'1 is j,2 is 1, the special genotype number x of the jth historical case data j,0 is 0, that is:

[0109]

[0110] Let the number of historical cases contained in subset Ψ'1 be P1, then the first loss function J adapted to the subset Ψ'1 can be simplified as:

[0111]

[0112] Among them, α1 and β1 are the parameters affecting the dosage of medication; x j,1 is the dosage of the jth historical case data in subset Ψ'1; is the measured value of the adverse reaction of the jth historical case data in the subset Ψ'1; J1 is the simplified loss function; j≤P1.

[0113] The optimal solution corresponding to the above loss function can be analyzed as:

[0114]

[0115] Among them, X1 is the matrix composed of the dosage of P1 historical case data; Y1 is the vector composed of the measured values ​​of adverse reactions of P1 historical case data; α1 and β1 are the parameters affecting the dosage; j,1 is the dosage of the jth historical case data in subset Ψ'1; The dosage of the historical case data in item P1 in subset Ψ'1; is the measured value of the adverse reaction of the jth historical case data in the subset Ψ'1; j≤P1.

[0116] Step S33. Determine a second data subset from the historical case data set, wherein the special genotype number corresponding to the second data subset is a second preset value;

[0117] Step S34. Expanding the first model based on the second data subset to obtain a second model, wherein the second model further considers the influence of chemotherapy cycles compared to the first model;

[0118] Specifically, for the calculation of the parameters affecting the chemotherapy cycle, a subset Ψ'2 is selected from the data set Ψ, and the special genotype number x of the jth historical case data in the subset Ψ'2 is j,0 is 0, that is:

[0119]

[0120] The number of historical cases contained in the subset Ψ'2 is P2. Given the parameters affecting the dosage, the first loss function J adapted to the subset Ψ'2 can be simplified as:

[0121]

[0122] Among them, α2 and β2 are the parameters affecting the chemotherapy cycle; x j,1is the dosage of the jth historical case data in subset Ψ'2; j,2 is the chemotherapy cycle of the jth historical case data in subset Ψ'2; is the measured value of adverse reaction of the jth historical case data in subset Ψ'2; J2 is the simplified loss function; g1(x j,1 ) is the effect value calculation function of the dosage on the adverse reaction of the j-th historical case data in the subset Ψ'2, and its parameters are obtained in step S32; j≤P2.

[0123] The optimal solution corresponding to the above first loss function can be analyzed as:

[0124]

[0125] Among them, X2 is the matrix composed of the product of the chemotherapy cycle of P2 historical case data and the calculation function of the effect of drug dosage on adverse reactions; Y2 is the vector composed of the difference between the measured value of adverse reactions of P2 historical case data and the calculation function of the effect of drug dosage on adverse reactions; α2 and β2 are the influencing parameters of chemotherapy cycle; x j,1 is the dosage of the jth historical case data in subset Ψ'2; is the dosage of the historical case data of item P2 in subset Ψ'2; j,2 is the chemotherapy cycle of the jth historical case data in subset Ψ'2; The chemotherapy cycle of the historical case data of item P2 in subset Ψ'2; is the measured value of adverse reaction of the jth historical case data in subset Ψ'2; g1(x j,1 ) is the effect value calculation function of the dosage on the adverse reaction of the j-th historical case data in the subset Ψ'2, and its parameters are obtained in step S32; j≤P2.

[0126] Step S35. Determine multiple third data subsets from the historical case data set, and the special genotype serial number corresponding to each third data subset is the same value;

[0127] Step S36. Expanding the second model based on multiple third data subsets to obtain a third model, wherein the third model further considers the additional impact of gene polymorphism compared to the second model;

[0128] Specifically, for the calculation of the additional influencing parameters of gene polymorphism, a subset Ψ' is selected from the data set Ψ S , subset Ψ' S The special genotype serial numbers of the historical case data are all non-zero integers k. S The number of case data contained is M' k ,Right now:

[0129]

[0130] Given the influencing parameters of drug dosage and chemotherapy cycle, the first loss function J can be simplified as:

[0131]

[0132] Among them, J S is the simplified loss function; k is the special genotype number of the historical case data; δ k is the additional influencing parameter of the gene polymorphism corresponding to the special genotype number k; j,1 For the subset Ψ' S The dosage of the historical case data in the jth article; x j,2 For the subset Ψ' S The chemotherapy cycle of the jth historical case data; g1(x j,1 ) is a subset Ψ' S The influence value calculation function of the dosage of the j-th historical case data on the adverse reaction, whose parameters are obtained in step S32; g2(x j,2 ) is a subset Ψ' S The calculation function of the influence value of chemotherapy cycle on adverse reactions of the j-th historical case data, whose parameters are obtained in step S34; For the subset Ψ' S The measured value of adverse reactions in the jth historical case data; j≤M' k .

[0133] The optimal solution corresponding to the above loss function is:

[0134]

[0135] Among them, k is the special genotype number of historical case data; δ k is the additional influencing parameter of the gene polymorphism corresponding to the special genotype number k; j,1 For the subset Ψ' S The dosage of the historical case data in the jth article; x j,2 For the subset Ψ' S The chemotherapy cycle of the jth historical case data; g1(x j,1 ) is a subset Ψ' S The influence value calculation function of the dosage of the j-th historical case data on the adverse reaction, whose parameters are obtained in step S32; g2(x j,2 ) is a subset Ψ' S The calculation function of the influence value of chemotherapy cycle on adverse reactions of the j-th historical case data, whose parameters are obtained in step S34; For the subset Ψ' SThe measured value of adverse reactions in the historical case data of the jth item; M' k For the subset Ψ' S The number of case data contained; j≤M' k .

[0136] Step S37. Expand the third model based on the historical case data set to obtain a fourth model. Compared with the third model, the fourth model also considers the influence of gender;

[0137] Specifically, for the gender influence parameter, all the data of the data set Ψ are used as the calculation basis. Given the medication dosage influence parameter, chemotherapy cycle influence parameter and gene polymorphism additional influence parameter, the first loss function J is simplified to:

[0138]

[0139] Among them, J3 is the simplified loss function; α3 is the gender influence parameter; g0(x i,0 ) is the calculation function of the influence value of the special genotype number of the i-th historical case data on adverse reactions in the data set Ψ, and its parameters are obtained in step S36; g1(x i,1 ) is the calculation function of the effect of the dosage on the adverse reaction of the j-th historical case data in the data set Ψ, and its parameters are obtained in step S32; g2(x i,2 ) is a calculation function for the effect of chemotherapy cycles on adverse reactions of the i-th historical case data in the data set Ψ, and its parameters are obtained in step S34; is the measured value of the adverse reaction of the i-th historical case data in the data set Ψ; N is the number of historical cases contained in the data set Ψ; x i,1 is the dosage of the i-th historical case data in the dataset Ψ; i,2 is the chemotherapy cycle of the i-th historical case data in the data set Ψ; i,3 is the gender of the i-th historical case data in the dataset Ψ; i≤N.

[0140] The optimal solution corresponding to the above loss function is:

[0141]

[0142] Among them, α3 is the gender influence parameter; g0(x i,0 ) is the calculation function of the influence value of the special genotype number of the i-th historical case data on adverse reactions in the data set Ψ, and its parameters are obtained in step S36; g1(x i,1 ) is the calculation function of the effect of the dosage on the adverse reaction of the j-th historical case data in the data set Ψ, and its parameters are obtained in step S32; g2(x i,2) is a calculation function for the effect of chemotherapy cycles on adverse reactions of the i-th historical case data in the data set Ψ, and its parameters are obtained in step S34; is the measured value of the adverse reaction of the i-th historical case data in the data set Ψ; N is the number of historical cases contained in the data set Ψ; x i,1 is the dosage of the i-th historical case data in the dataset Ψ; i,2 is the chemotherapy cycle of the i-th historical case data in the data set Ψ; i,3 is the gender of the i-th historical case data in the dataset Ψ; i≤N.

[0143] Step S38. Expand the fourth model based on the historical case data set to obtain an initial prediction model. Compared with the fourth model, the initial prediction model also considers the influence of age, physical condition, nutritional status and pathological state of the body;

[0144] For the four influencing factors of age, physical condition, nutritional status and pathological status of the body, all the data of the data set Ψ are used as the calculation basis, and their average value is used as the mean of each influencing factor, which is:

[0145]

[0146] in, is the mean age; is the mean value of physical condition; is the mean nutritional status; is the mean value of the pathological state of the body; x i,k is the kth influencing factor to be processed of the i-th historical case data in the data set Ψ, and the values ​​of k are 4, 5, 6 and 7; N is the number of historical cases contained in the data set Ψ; i≤N.

[0147] Based on all the data in the data set Ψ, given the influencing parameters of drug dosage, chemotherapy cycle, additional influencing parameters of gene polymorphism and gender, the four influencing factors of age, physical condition, nutritional status and pathological state of the body are processed respectively, and the first loss function J is simplified to:

[0148]

[0149] Among them, J k is the loss function corresponding to the kth influencing factor; x i,k is the kth influencing factor to be processed for the i-th historical case data in the data set Ψ; i,j is the jth processed influencing factor of the i-th historical case data in the data set Ψ; α k and β k is the influencing parameter of the kth influencing factor to be processed; is the mean value of the kth influencing factor to be processed; gj (x i,j ) is a function for calculating the impact value of the j-th processed influencing factor on the adverse reaction, and its parameters can be obtained by steps S32, S34, S36 and S37; is the measured value of the adverse reaction of the jth historical case data in the data set Ψ; N is the number of historical cases contained in the data set Ψ; the influencing factors to be processed are age, physical condition, nutritional status and pathological status of the body, and the influencing factors that have been processed are the influencing parameters of medication dosage, chemotherapy cycle, gene polymorphism additional parameters and gender influencing parameters; k takes values ​​of 4, 5, 6 and 7; i≤N.

[0150] The optimal solution corresponding to the above first loss function can be analyzed as:

[0151]

[0152] Among them, X k Y is the matrix related to the kth influencing factor to be processed in N historical cases; k is the vector of the measured values ​​of adverse reactions for N historical case data; α k and β k is the influencing parameter of the kth influencing factor to be processed; x N,k is the kth influencing factor to be processed for the Nth historical case data in the data set Ψ; N,j is the jth processed influencing factor of the Nth historical case data in the dataset Ψ; is the measured value of adverse reaction of the Nth historical case data in the dataset Ψ; j (x N,j ) is a function for calculating the impact value of the j-th processed influencing factor on the adverse reaction, and its parameters can be obtained by steps S32, S34, S36 and S37; is the mean value of the kth influencing factor to be processed; N is the number of historical cases contained in the data set Ψ; k takes values ​​of 4, 5, 6 and 7; i≤N.

[0153] When k=4, α4 and β4 are age influencing parameters; when k=5, α5 and β5 are physical fitness influencing parameters; when k=6, α6 and β6 are nutritional status influencing parameters; when k=7, α7 and β7 are pathological status influencing parameters.

[0154] Step S4. Based on the historical case data set, the initial prediction model is trained. When the preset first loss function meets the set conditions, the training is stopped to obtain the expected value prediction sub-model;

[0155] Specifically, the present application solves the parameters of the prediction model based on historical cases. Note that x = [x0, x1, ..., x7], where x0 is the special genotype number; x1 is the medication dosage; x2 is the chemotherapy cycle; x3 is the gender; x4 is the age; x5 is the physical condition; x6 is the nutritional status; and x7 is the pathological state of the body.

[0156] The constructed prediction model and influencing parameters are organized into the following expression form:

[0157] Φ=[δ0,δ1,…,δ K ]

[0158] Γ=[α1,β1,α2,β2,α3,α4,β4,…,α7,β7]

[0159] λ=[λ0,λ1,λ2]

[0160] Among them, Φ and Γ are the expected value influencing parameter sequences; λ is the standard deviation influencing parameter sequence; δ k is the additional influencing parameter of the gene polymorphism corresponding to the special genotype number k; α1 and β1 are the influencing parameters of the medication dosage; α2 and β2 are the influencing parameters of the chemotherapy cycle; α3 is the influencing parameter of gender; α4 and β4 are the influencing parameters of age; α5 and β5 are the influencing parameters of physical state; α6 and β6 are the influencing parameters of nutritional status; α7 and β7 are the influencing parameters of the pathological state of the body; λ0 is the first standard deviation parameter; λ1 is the second standard deviation parameter; λ2 is the third standard deviation parameter.

[0161] This application first obtains the expected value influencing parameter sequence, and then obtains the standard deviation influencing parameter sequence.

[0162] By selecting appropriate parameter sequences Φ and Γ to minimize the value of the above-mentioned first loss function, the expected value can be solved to affect the parameter sequence. Considering that in the N summation terms, each term contains all the parameters in Γ and only one parameter in Φ, it is impossible to solve these two types of parameters at the same time. Therefore, this application adopts an iterative solution method, fixes one type of parameters, optimizes and solves another type of parameters, and iterates repeatedly until the end condition is met.

[0163] Specifically, step S4 specifically includes step S41, step S42, step S43 and step S44:

[0164] Step S41. First training operation: inputting the historical case data set into the initial prediction model for first training, updating the initial prediction model when the value of the first loss function is minimum, and obtaining an adjusted prediction model; the first training is used to characterize the influence of fixed medication dosage, chemotherapy cycle, gender, age, physical condition, nutritional condition and pathological condition of the body, and optimizing the additional influence of gene polymorphism based on the derivation algorithm;

[0165] Specifically, when Γ is fixed, the optimal solution of Φ can be directly obtained.

[0166] By selecting a subset Ψ from the historical case data set Ψ S , subset Ψ S The special genotype serial number of the historical case data is fixed to a non-zero integer k. S The number of case data contained is M k ,Right now:

[0167]

[0168] The first loss function JCPC M k The corresponding sub-terms are subject to δ k Impact, remember Among them, g i (x i ) is the calculation function of the impact value of the ith influencing factor on adverse reactions; Π is the multiplication symbol. Derivative of the first loss function J yields:

[0169]

[0170] Among them, G(x j ) is a subset Ψ S The multiplication result of the impact values ​​of multiple influencing factors on adverse reactions in the jth historical case; x j For the data set Ψ S The chemotherapy data of the jth historical case data; k is the special genotype number of the historical case data; δ k It is the additional influencing parameter of the gene polymorphism corresponding to the special genotype number k; For the subset Ψ S The measured value of the adverse reaction in the jth historical case data; j≤M k .

[0171] Therefore, the optimal solution of Φ when Γ is fixed is:

[0172]

[0173] Among them, G(x j ) is a subset Ψ S The product of the multiple influencing factors of the j-th historical case data on the adverse reaction value function; x j For the data set Ψ S Chemotherapy data of historical case data in Article j; For the data set Ψ S The measured value of adverse reactions in the historical case data of Article j; M k For the data set Ψ SThe number of historical cases contained; δ k is the additional influencing parameter of the gene polymorphism corresponding to the special genotype number k; j≤M k .

[0174] Step S42. Second training operation: inputting the historical case data set into the adjusted prediction model for second training, and obtaining an updated initial prediction model when the value of the first loss function decreases; the second training is used to characterize the influence of the fixed gene polymorphism, based on the optimal gradient algorithm, on the influence of the medication dosage, the chemotherapy cycle, the gender, the age, the physical condition, the nutritional condition and the pathological state of the body for optimization;

[0175] Specifically, when Φ is fixed, the present invention solves Γ by an optimal gradient method.

[0176] The optimal gradient algorithm gradually updates the optimization parameters along the negative gradient direction of the first loss function J with respect to Γ. By reasonably determining the step size factor, the step size is shortened as the solution approaches the optimal solution. The gradient vector of J with respect to Γ is calculated as:

[0177]

[0178] in, is the gradient vector of the first loss function J with respect to Γ; J is the first loss function; α1 and β1 are the influencing parameters of medication dosage; α2 and β2 are the influencing parameters of chemotherapy cycle; α3 is the influencing parameter of gender; α4 and β4 are the influencing parameters of age; α5 and β5 are the influencing parameters of physical state; α6 and β6 are the influencing parameters of nutritional status; α7 and β7 are the influencing parameters of pathological state of the body.

[0179] Each component in the gradient vector is:

[0180]

[0181] Among them, J is the first loss function; α j and β j is the influencing parameter of the jth influencing factor; is the measured value of adverse reaction of the i-th historical case data in the data set Ψ; i is the expected value of the i-th historical case data in the data set Ψ; x i,j is the jth influencing factor of the i-th historical case data in the data set Ψ; g j (x i,j ) is the impact value of the jth influencing factor on the adverse reaction of the i-th historical case data in the data set Ψ; is the mean value of the jth influencing factor; N is the number of historical cases contained in the data set Ψ; and Variables are introduced for the convenience of expression, and their values ​​are 0, 1 and 0 respectively; i≤N.

[0182] The second-order partial derivative matrix of J with respect to Γ is calculated as:

[0183]

[0184] in, is the second-order partial derivative matrix of the first loss function J with respect to Γ; J is the first loss function; α1 and β1 are the influencing parameters of medication dosage; α2 and β2 are the influencing parameters of chemotherapy cycle; α3 is the influencing parameter of gender; α4 and β4 are the influencing parameters of age; α5 and β5 are the influencing parameters of physical condition; α6 and β6 are the influencing parameters of nutritional status; α7 and β7 are the influencing parameters of pathological state of the body.

[0185] Each component in the above second-order partial derivative matrix is ​​expressed as:

[0186]

[0187] Among them, J is the first loss function; α j and β j is the influencing parameter of the jth influencing factor; is the measured value of adverse reaction of the i-th historical case data in the data set Ψ; i is the expected value of the i-th historical case data in the data set Ψ; x i,j is the jth influencing factor of the i-th historical case data in the data set Ψ; g j (x i,j ) is the impact value of the jth influencing factor on the adverse reaction of the i-th historical case data in the data set Ψ; is the mean value of the jth influencing factor; x j,k is the kth influencing factor of the jth historical case in the data set Ψ; g k (x j,k ) is the impact value of the kth influencing factor of the jth historical case on the adverse reaction in the data set Ψ; N is the number of historical cases contained in the data set Ψ; k≠j; i≤N.

[0188] Based on the parameter vector sequence Γ k , calculate the updated value Γ according to the optimal gradient method k+1 for

[0189]

[0190] Among them, Γ k The parameter sequence affected by the predicted value after the k-th gradient algorithm update; is the first loss function J after the k-th gradient algorithm update Γ k The gradient vector of is the first loss function J after the k-th gradient algorithm update Γ k The second-order partial derivative matrix of .

[0191] Step S43. Calculation operation: input the historical case data set into the updated initial prediction model and calculate the first loss function to obtain the loss function value;

[0192] Step S44: Repeat the first training operation, the second training operation and the calculation operation until the difference between two adjacent loss function values ​​is less than a set threshold, and use the updated initial prediction model as the expected value prediction sub-model;

[0193] Specifically, for the expected value influencing parameter sequences Φ and Γ, under the premise that one of the parameters is fixed, the other parameter is updated according to the above calculation steps, and this iteration is repeated until the difference between the first loss function twice is less than the set threshold. The convergence condition is met and the parameter updating process is exited. At this time, the corresponding model is the expected value prediction sub-model.

[0194] The comparison formula of the first loss function between two adjacent times is:

[0195]

[0196] Among them, J(Φ k ,Γ k ) is the first loss function after the k-th gradient algorithm update; J(Φ k+1 ,Γ k+1 ) is the first loss function after the k+1th gradient algorithm update; N is the number of historical cases contained in the data set Ψ; ε is the set threshold.

[0197] In this application, the threshold value is set to 1×10 -6 The corresponding threshold values ​​may vary depending on the scenario and are not particularly limited here.

[0198] Step S5. Based on the historical case data set and the expected value prediction sub-model, a second loss function is constructed to obtain the optimal solution and the standard deviation prediction sub-model;

[0199] Specifically, step S5 specifically includes step S51, step S52, step S53, step S54, step S55 and step S56:

[0200] Step S51. Based on the expected value prediction sub-model, multiple adverse reaction expected values ​​corresponding to the historical case data set are calculated;

[0201] Step S52. Combining a plurality of adverse reaction expected values ​​with a plurality of corresponding adverse reaction measured values ​​to obtain adverse reaction sample pairs;

[0202] Step S53. Dividing the multiple adverse reaction sample pairs into multiple subsets based on the sizes of the multiple adverse reaction expected values, to obtain multiple adverse reaction subsets;

[0203] Step S54. Calculate the mean and standard deviation of the adverse reaction expected value in each adverse reaction subset to obtain multiple adverse reaction means and adverse reaction standard deviations;

[0204] Step S55. constructing a second loss function based on multiple adverse reaction means and multiple adverse reaction standard deviations, and obtaining a standard deviation influence parameter when the second loss function is minimized;

[0205] Step S56. Based on the standard deviation influencing parameter, obtain a standard deviation prediction sub-model;

[0206] Specifically, the expected value prediction sub-model is used to calculate the expected value of adverse reactions of N historical case data, and the expected value and the corresponding measured value of adverse reactions are combined into sample pairs, which are sorted from small to large according to the expected value and divided into Q subsets. Each subset contains R sample pairs, so any subset is:

[0207]

[0208] Among them, Ω q is the qth subset, q≤Q; y q,i is the expected value of adverse reactions in the ith sample pair in the qth subset; is the measured value of adverse reaction in the i-th sample pair in the q-th subset; R is the number of sample pairs in each subset.

[0209]

[0210] Among them, N is the number of historical cases, Q is the number of subsets, is the floor function.

[0211] In this application, the Q value is set to 15. The Q value in different scenarios may not be exactly the same, and there is no special limitation here.

[0212] Calculate each subset Ω q The expected value mean and corresponding standard deviation of :

[0213]

[0214] in, is the mean of the predicted values ​​of the qth subset; is the standard deviation of the qth subset; R is the number of sample pairs in each subset; y q,i, is the expected value of adverse reaction of the ith sample pair in the qth subset; is the measured value of the adverse reaction in the i-th sample pair in the q-th subset; i≤R.

[0215] The second loss function is defined as:

[0216]

[0217] Among them, I(λ) is the second loss function; is the expected value mean of the qth subset; is the standard deviation of the qth subset; Q is the number of subsets; λ0 is the first standard deviation parameter; λ1 is the second standard deviation parameter; λ2 is the third standard deviation parameter; q≤Q.

[0218] By adjusting the standard deviation influence parameter sequence λ, the second loss function value is minimized and the corresponding optimal solution of the standard deviation influence parameter sequence is obtained. The optimal solution of the standard deviation influence parameter sequence can be analytically expressed as:

[0219]

[0220] in, It is a matrix composed of the mean values ​​of the predicted values ​​of Q subsets; is a vector consisting of the standard deviations of Q subsets; is the expected value mean of the qth subset; is the standard deviation of the qth subset; Q is the number of subsets; q≤Q.

[0221] The adverse reaction standard deviation prediction submodel is constructed according to the calculated standard deviation influence parameter sequence, and the adverse reaction prediction model is constructed based on the standard deviation prediction submodel and the expected value prediction submodel.

[0222] Step S6. Obtaining chemotherapy data of the target case;

[0223] Step S7. Inputting chemotherapy data of the target case into the adverse reaction prediction model to obtain the expected value of adverse reactions of the target case and the corresponding standard deviation;

[0224] Specifically, by inputting the chemotherapy data of the target case into the trained adverse reaction prediction model for prediction, a more accurate adverse reaction expected value and corresponding standard deviation can be obtained.

[0225] Embodiment 2:

[0226] like Figure 2 As shown, this embodiment provides a device for predicting adverse reactions of chemotherapy drugs, the device comprising:

[0227] A first acquisition unit 10 is used to acquire a historical case data set, the historical case data set includes chemotherapy data and adverse reaction evaluation data of multiple historical cases, the chemotherapy data includes chemotherapy drugs, chemotherapy cycles and patient body conditions, and the adverse reaction evaluation data includes bone marrow suppression index, liver toxicity index and kidney toxicity index;

[0228] A first calculation unit 20, configured to calculate a plurality of adverse reaction measured values ​​based on a plurality of adverse reaction assessment data;

[0229] An initialization unit 30 is used to initialize a preset prediction model based on a historical case data set to obtain an initial prediction model;

[0230] A first training unit 40 is used to train the initial prediction model based on the historical case data set, and when the preset first loss function meets the set condition, the training is stopped to obtain the expected value prediction sub-model;

[0231] The second training unit 50 is used to construct a second loss function based on the historical case data set and the expected value prediction sub-model, find the optimal solution, and obtain the standard deviation prediction sub-model;

[0232] A second acquisition unit 60 is used to acquire a chemotherapy data set of a target case;

[0233] The input unit 70 is used to input the chemotherapy data of the target case into the adverse reaction prediction model to obtain the adverse reaction expected value and the corresponding standard deviation of the target case. The adverse reaction prediction model consists of an expected value prediction sub-model and a standard deviation prediction sub-model.

[0234] In a specific implementation disclosed in the present application, the first computing unit 20 includes:

[0235] a third acquisition unit, for acquiring the neutrophil content, hemoglobin content, platelet content, alanine aminotransferase content, aspartate aminotransferase content, bilirubin content, urea nitrogen content and creatinine content of the historical cases before chemotherapy from the evaluation data, as the former first indicator, the former second indicator, the former third indicator, the former fourth indicator, the former fifth indicator, the former sixth indicator, the former seventh indicator and the former eighth indicator, respectively;

[0236] a fourth acquisition unit, for acquiring the neutrophil content, hemoglobin content, platelet content, alanine aminotransferase content, aspartate aminotransferase content, bilirubin content, urea nitrogen content and creatinine content of historical cases after chemotherapy from the evaluation data, as the first post-indicator, the second post-indicator, the third post-indicator, the fourth post-indicator, the fifth post-indicator, the sixth post-indicator, the seventh post-indicator and the eighth post-indicator, respectively;

[0237] a fifth acquisition unit, for acquiring the neutrophil content, hemoglobin content, platelet content, alanine aminotransferase content, aspartate aminotransferase content, bilirubin content, urea nitrogen content and creatinine content in a normal state as the first indicator, the second indicator, the third indicator, the fourth indicator, the fifth indicator, the sixth indicator, the seventh indicator and the eighth indicator, respectively;

[0238] A second calculation unit is used to calculate a bone marrow suppression evaluation value based on the front first index, the front second index, the front third index, the rear first index, the rear second index, the rear third index, the first index, the second index, the third index and the first weight adjustment coefficient;

[0239] A third calculation unit is used to calculate a liver toxicity assessment value based on the front fourth index, the front fifth index, the front sixth index, the rear fourth index, the rear fifth index, the rear sixth index, the fourth index, the fifth index, the sixth index and the second weight adjustment coefficient;

[0240] A fourth calculation unit is used to calculate a renal toxicity assessment value based on the first seventh index, the first eighth index, the last seventh index, the last eighth index, the seventh index, the eighth index and the third weight adjustment coefficient;

[0241] As a unit, it is used to determine the maximum value from the bone marrow suppression assessment value, liver toxicity assessment value and kidney toxicity assessment value as the actual measured value of the adverse reaction.

[0242] In a specific implementation disclosed in the present application, the initialization unit 30 includes:

[0243] A first determination unit is used to determine a first data subset from the historical case data set, wherein the chemotherapy cycle corresponding to the first data subset is a first preset value, and the corresponding special genotype serial number is a second preset value;

[0244] A simplification unit, used to simplify the prediction model based on the first data subset to obtain a first model, wherein the first model only considers the influence of the dosage of the chemotherapy drug data;

[0245] A second determining unit is used to determine a second data subset from the historical case data set, wherein the special genotype serial number corresponding to the second data subset is a second preset value;

[0246] A first expansion unit is used to expand the first model based on the second data subset to obtain a second model, wherein the second model further considers the influence of the chemotherapy cycle compared with the first model;

[0247] A third determination unit is used to determine a plurality of third data subsets from the historical case data set, wherein the special genotype serial number corresponding to each third data subset is the same value;

[0248] A second expansion unit is used to expand the second model based on multiple third data subsets to obtain a third model, and the third model further considers the additional impact of gene polymorphism compared with the second model;

[0249] A third expansion unit is used to expand the third model based on the historical case data set to obtain a fourth model, and the fourth model also considers the influence of gender compared with the third model;

[0250] The fourth expansion unit is used to expand the fourth model based on the historical case data set to obtain an initial prediction model. Compared with the fourth model, the initial prediction model also considers the influence of age, physical condition, nutritional status and pathological state of the body.

[0251] In a specific implementation disclosed in the present application, the first training unit 40 includes:

[0252] The third training unit is used for the first training operation: inputting the historical case data set into the initial prediction model for the first training, and updating the initial prediction model when the value of the first loss function is the minimum to obtain an adjusted prediction model; the first training is used to characterize the influence of fixed medication dosage, chemotherapy cycle, gender, age, physical condition, nutritional status and pathological state of the body, and optimize the additional influence of gene polymorphism based on the derivation algorithm;

[0253] The fourth training unit is used for the second training operation: inputting the historical case data set into the adjusted prediction model for the second training, and obtaining the updated initial prediction model when the value of the first loss function decreases; the second training is used to characterize the influence of the additional influence of the fixed gene polymorphism, and optimize the influence of the medication dosage, the chemotherapy cycle, the gender, the age, the physical condition, the nutritional condition and the pathological state of the body based on the optimal gradient algorithm;

[0254] A fifth calculation unit, configured to perform a calculation operation: inputting the historical case data set into the updated initial prediction model and calculating a first loss function to obtain a loss function value;

[0255] The repeating unit is used to repeat the first training operation, the second training operation and the calculation operation until the difference between two adjacent loss function values ​​is less than a set threshold, and the updated initial prediction model is used as the expected value prediction sub-model.

[0256] In a specific implementation disclosed in the present application, the second training unit 50 includes:

[0257] A sixth calculation unit, based on the expected value prediction sub-model, calculates a plurality of adverse reaction expected values ​​corresponding to the historical case data set;

[0258] A combining unit, used for combining a plurality of adverse reaction expected values ​​with a plurality of corresponding adverse reaction measured values ​​to obtain adverse reaction sample pairs;

[0259] A division unit, used for dividing the plurality of adverse reaction sample pairs into a plurality of subsets based on the magnitudes of the plurality of adverse reaction expected values, to obtain a plurality of adverse reaction subsets;

[0260] a seventh calculation unit, used to calculate the mean and standard deviation of the adverse reaction expected value in each adverse reaction subset, to obtain multiple adverse reaction means and adverse reaction standard deviations;

[0261] A construction unit, used to construct a second loss function based on multiple adverse reaction means and multiple adverse reaction standard deviations, and obtain a standard deviation influence parameter when the second loss function is minimized;

[0262] The unit is obtained, and based on the standard deviation influencing parameter, a standard deviation prediction sub-model is obtained;

[0263] In a specific implementation disclosed in the present application, the second computing unit includes:

[0264] an eighth calculation unit, used to calculate the difference between the first index and the previous first index and the next first index, respectively, to obtain a first difference and a second difference;

[0265] a ninth calculation unit, configured to calculate ratios of the first difference and the second difference to a preset first normalization parameter, respectively, to obtain a first ratio and a second ratio;

[0266] a tenth calculation unit, configured to calculate the product of the first ratio and the first weight adjustment coefficient to obtain a first product;

[0267] an eleventh calculation unit, configured to calculate a difference between the preset fixed value and the first weight adjustment coefficient, and multiply the difference by the second ratio to obtain a second product;

[0268] The twelfth calculation unit is used to calculate the sum of the first product and the second product to obtain a neutrophil evaluation value.

[0269] It should be noted that, regarding the device in the above embodiment, the specific manner in which each module performs the operation has been described in detail in the embodiment of the method, and will not be elaborated here.

[0270] Embodiment 3:

[0271] Corresponding to the above method embodiment, this embodiment also provides a device for predicting adverse reactions of chemotherapy drugs. The device for predicting adverse reactions of chemotherapy drugs described below and the method for predicting adverse reactions of chemotherapy drugs described above can refer to each other.

[0272] Figure 3 FIG. 8 is a block diagram of a device 800 for predicting adverse reactions of chemotherapy drugs according to an exemplary embodiment. Figure 3 As shown, the device 800 for predicting adverse reactions of chemotherapy drugs may include: a processor 801 and a memory 802. The device 800 for predicting adverse reactions of chemotherapy drugs may also include one or more of a multimedia component 803, an I / O interface 804, and a communication component 805.

[0273] The processor 801 is used to control the overall operation of the device 800 for predicting adverse reactions of chemotherapy drugs to complete all or part of the steps in the above-mentioned method for predicting adverse reactions of chemotherapy drugs. The memory 802 is used to store various types of data to support the operation of the device 800 for predicting adverse reactions of chemotherapy drugs, such as instructions for any application or method used to operate on the device 800 for predicting adverse reactions of chemotherapy drugs, and application-related data, such as contact data, sent and received messages, pictures, audio, video, etc. The memory 802 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, disk or optical disk. The multimedia component 803 may include a screen and an audio component. The screen may be, for example, a touch screen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone, which is used to receive external audio signals. The received audio signal may be further stored in the memory 802 or sent via the communication component 805. The audio component also includes at least one speaker for outputting audio signals. The I / O interface 804 provides an interface between the processor 801 and other interface modules, and the above-mentioned other interface modules can be keyboards, mice, buttons, etc. These buttons can be virtual buttons or physical buttons. The communication component 805 is used for wired or wireless communication between the chemotherapy drug adverse reaction prediction device 800 and other devices. Wireless communication, such as Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G or 4G, or a combination of one or more of them, so the corresponding communication component 805 can include: Wi-Fi module, Bluetooth module, NFC module.

[0274] In an exemplary embodiment, the device 800 for predicting adverse reactions to chemotherapy drugs can be implemented by one or more application specific integrated circuits (ASIC), digital signal processors (DSP), digital signal processing devices (DSPD), programmable logic devices (PLD), field programmable gate arrays (FPGA), controllers, microcontrollers, microprocessors or other electronic components to execute the above-mentioned method for predicting adverse reactions to chemotherapy drugs.

[0275] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided, and when the program instructions are executed by a processor, the steps of the above-mentioned method for predicting adverse reactions of chemotherapy drugs are implemented. For example, the computer-readable storage medium can be the above-mentioned memory 802 including program instructions, and the above-mentioned program instructions can be executed by the processor 801 of the device 800 for predicting adverse reactions of chemotherapy drugs to complete the above-mentioned method for predicting adverse reactions of chemotherapy drugs.

[0276] Embodiment 4:

[0277] Corresponding to the above method embodiment, a readable storage medium is also provided in this embodiment. The readable storage medium described below and the method for predicting adverse reactions of chemotherapy drugs described above can refer to each other.

[0278] A readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the method for predicting adverse reactions of chemotherapy drugs in the above method embodiment.

[0279] The readable storage medium may specifically be a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, or other readable storage medium that can store program codes.

[0280] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

[0281] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for predicting adverse reactions of chemotherapy drugs, characterized in that: include: Acquire a historical case data set, wherein the historical case data set includes chemotherapy data and adverse reaction evaluation data of multiple historical cases, wherein the chemotherapy data includes chemotherapy drugs, chemotherapy cycles, and patient body conditions, and the adverse reaction evaluation data includes bone marrow suppression indicators, liver toxicity indicators, and kidney toxicity indicators; Based on multiple adverse reaction assessment data, multiple adverse reaction measured values ​​are calculated; Initializing a preset prediction model based on the historical case data set to obtain an initial prediction model; Based on the historical case data set, the initial prediction model is trained, and when the preset first loss function meets the set condition, the training is stopped to obtain the expected value prediction sub-model; Based on the historical case data set and the expected value prediction sub-model, a second loss function is constructed to obtain an optimal solution and a standard deviation prediction sub-model is obtained; Obtain chemotherapy datasets for target cases; The chemotherapy data of the target case is input into an adverse reaction prediction model to obtain the expected value of adverse reactions of the target case and the corresponding standard deviation. The adverse reaction prediction model is composed of the expected value prediction sub-model and the standard deviation prediction sub-model.

2. The method for predicting adverse reactions of chemotherapeutic drugs according to claim 1, characterized in that Based on multiple adverse reaction evaluation data, multiple adverse reaction measured values ​​are calculated, the bone marrow suppression index includes neutrophil content, hemoglobin content and platelet content, the liver toxicity index includes alanine aminotransferase content, aspartate aminotransferase content and bilirubin content, the kidney toxicity index includes urea nitrogen content and creatinine content, including: Obtaining the neutrophil content, hemoglobin content, platelet content, alanine aminotransferase content, aspartate aminotransferase content, bilirubin content, urea nitrogen content and creatinine content of the historical cases before chemotherapy from the evaluation data, as the first former indicator, the second former indicator, the third former indicator, the fourth former indicator, the fifth former indicator, the sixth former indicator, the seventh former indicator and the eighth former indicator, respectively; Obtaining the neutrophil content, hemoglobin content, platelet content, alanine aminotransferase content, aspartate aminotransferase content, bilirubin content, urea nitrogen content and creatinine content of the historical cases after chemotherapy from the evaluation data, as the first index after treatment, the second index after treatment, the third index after treatment, the fourth index after treatment, the fifth index after treatment, the sixth index after treatment, the seventh index after treatment and the eighth index after treatment, respectively; The neutrophil content, hemoglobin content, platelet content, alanine aminotransferase content, aspartate aminotransferase content, bilirubin content, urea nitrogen content and creatinine content in the normal state are obtained as the first index, the second index, the third index, the fourth index, the fifth index, the sixth index, the seventh index and the eighth index respectively; Based on the first preceding indicator, the second preceding indicator, the third preceding indicator, the first succeeding indicator, the second succeeding indicator, the third succeeding indicator, the first preceding indicator, the second succeeding indicator, the third succeeding indicator, the first preceding indicator, the second preceding indicator, the third succeeding indicator and a first weight adjustment coefficient, a bone marrow suppression assessment value is calculated; Based on the front fourth index, the front fifth index, the front sixth index, the rear fourth index, the rear fifth index, the rear sixth index, the fourth index, the fifth index, the sixth index and the second weight adjustment coefficient, a liver toxicity assessment value is calculated; Based on the first seventh index, the first eighth index, the last seventh index, the last eighth index, the seventh index, the eighth index and the third weight adjustment coefficient, a kidney toxicity assessment value is calculated; A maximum value is determined from the bone marrow suppression assessment value, the liver toxicity assessment value, and the kidney toxicity assessment value as the adverse reaction actual measured value.

3. The method for predicting adverse reactions of chemotherapeutic drugs according to claim 1, characterized in that Based on the historical case data set, the preset prediction model is initialized to obtain an initial prediction model, wherein the influencing factors include special genotype sequence number, gender, age, physical condition, nutritional status and pathological state of the body, including: Determine a first data subset from the historical case data set, wherein the chemotherapy cycle corresponding to the first data subset is a first preset value, and the corresponding special genotype serial number is a second preset value; Simplifying the prediction model based on the first data subset to obtain a first model, wherein the first model only considers the effect of the dosage in the chemotherapy drug data; Determine a second data subset from the historical case data set, wherein the special genotype sequence number corresponding to the second data subset is the second preset value; Expanding the first model based on the second data subset to obtain a second model, wherein the second model further considers the influence of chemotherapy cycles compared to the first model; Determine a plurality of third data subsets from the historical case data set, wherein the special genotype sequence number corresponding to each third data subset is the same value; Expanding the second model based on multiple third data subsets to obtain a third model, wherein the third model further considers the additional impact of gene polymorphism compared to the second model; Expanding the third model based on the historical case data set to obtain a fourth model, wherein the fourth model further considers the influence of gender compared to the third model; The fourth model is expanded based on the historical case data set to obtain the initial prediction model. Compared with the fourth model, the initial prediction model also considers the influence of age, physical condition, nutritional status and pathological state of the body.

4. The method for predicting adverse reactions of chemotherapeutic drugs according to claim 3, characterized in that Based on the historical case data set, the initial prediction model is trained. When the preset first loss function meets the set conditions, the training is stopped to obtain the expected value prediction sub-model, including: A first training operation: inputting the historical case data set into the initial prediction model for a first training, updating the initial prediction model when the value of the first loss function is the minimum to obtain an adjusted prediction model, wherein the first training is used to characterize the influence of a fixed medication dose, the influence of chemotherapy cycles, the influence of gender, the influence of age, the influence of physical fitness, the influence of nutritional status, and the influence of pathological status of the body, and optimizing the additional influence of gene polymorphism based on a derivation algorithm; The second training operation: inputting the historical case data set into the adjusted prediction model for the second training, and obtaining an updated initial prediction model when the value of the first loss function decreases; the second training is used to characterize the influence of the additional influence of the fixed gene polymorphism, and optimize the influence of the medication dosage, the chemotherapy cycle, the gender, the age, the physical condition, the nutritional status and the pathological state of the body based on the optimal gradient algorithm; Calculation operation: inputting the historical case data set into the updated initial prediction model and calculating the first loss function to obtain a loss function value; Repeat the first training operation, the second training operation and the calculation operation until the difference between two adjacent loss function values ​​is less than a set threshold, and use the updated initial prediction model as the expected value prediction sub-model.

5. A device for predicting adverse reactions of chemotherapy drugs, characterized in that: include: A first acquisition unit is used to acquire a historical case data set, wherein the historical case data set includes chemotherapy data and adverse reaction evaluation data of multiple historical cases, wherein the chemotherapy data includes chemotherapy drugs, chemotherapy cycles and patient body conditions, and the adverse reaction evaluation data includes bone marrow suppression indicators, liver toxicity indicators and kidney toxicity indicators; A first calculation unit, configured to calculate a plurality of adverse reaction measured values ​​based on a plurality of adverse reaction assessment data; An initialization unit, used to initialize a preset prediction model based on the historical case data set to obtain an initial prediction model; A first training unit is used to train the initial prediction model based on the historical case data set, and when a preset first loss function meets a set condition, stop the training to obtain an expected value prediction sub-model; A second training unit is used to construct a second loss function based on the historical case data set and the expected value prediction sub-model, find the optimal solution, and obtain a standard deviation prediction sub-model; A second acquisition unit is used to acquire a chemotherapy data set of a target case; An input unit is used to input the chemotherapy data of the target case into an adverse reaction prediction model to obtain the expected value of the adverse reaction of the target case and the corresponding standard deviation. The adverse reaction prediction model is composed of the expected value prediction sub-model and the standard deviation prediction sub-model.

6. The device for predicting adverse reactions of chemotherapy drugs according to claim 5, characterized in that: The first computing unit comprises: a third acquisition unit, for acquiring, from the evaluation data, the neutrophil content, hemoglobin content, platelet content, alanine aminotransferase content, aspartate aminotransferase content, bilirubin content, urea nitrogen content, and creatinine content of the historical case before chemotherapy, as a first former indicator, a second former indicator, a third former indicator, a fourth former indicator, a fifth former indicator, a sixth former indicator, a seventh former indicator, and an eighth former indicator, respectively; a fourth acquisition unit, for acquiring, from the evaluation data, the neutrophil content, hemoglobin content, platelet content, alanine aminotransferase content, aspartate aminotransferase content, bilirubin content, urea nitrogen content, and creatinine content of the historical case after chemotherapy, as the first after-indicator, the second after-indicator, the third after-indicator, the fourth after-indicator, the fifth after-indicator, the sixth after-indicator, the seventh after-indicator, and the eighth after-indicator, respectively; a fifth acquisition unit, for acquiring the neutrophil content, hemoglobin content, platelet content, alanine aminotransferase content, aspartate aminotransferase content, bilirubin content, urea nitrogen content and creatinine content in a normal state as the first indicator, the second indicator, the third indicator, the fourth indicator, the fifth indicator, the sixth indicator, the seventh indicator and the eighth indicator, respectively; A second calculation unit is used to calculate a bone marrow suppression assessment value based on the front first indicator, the front second indicator, the front third indicator, the rear first indicator, the rear second indicator, the rear third indicator, the first indicator, the second indicator, the third indicator and a first weight adjustment coefficient; A third calculation unit is used to calculate a liver toxicity assessment value based on the front fourth index, the front fifth index, the front sixth index, the rear fourth index, the rear fifth index, the rear sixth index, the fourth index, the fifth index, the sixth index and the second weight adjustment coefficient; a fourth calculation unit, configured to calculate a renal toxicity assessment value based on the first seventh index, the first eighth index, the last seventh index, the last eighth index, the seventh index, the eighth index and a third weight adjustment coefficient; As a unit, it is used to determine the maximum value from the bone marrow suppression assessment value, the liver toxicity assessment value and the kidney toxicity assessment value as the adverse reaction actual measurement value.

7. The device for predicting adverse reactions of chemotherapy drugs according to claim 5, characterized in that: The initialization unit comprises: A first determining unit is used to determine a first data subset from the historical case data set, wherein the chemotherapy cycle corresponding to the first data subset is a first preset value, and the corresponding special genotype serial number is a second preset value; A simplification unit, configured to simplify the prediction model based on the first data subset to obtain a first model, wherein the first model only considers the effect of the dosage of the chemotherapy drug data; A second determining unit is used to determine a second data subset from the historical case data set, wherein the special genotype sequence number corresponding to the second data subset is the second preset value; A first expansion unit is used to expand the first model based on the second data subset to obtain a second model, wherein the second model further considers the influence of chemotherapy cycles compared with the first model; A third determining unit is used to determine a plurality of third data subsets from the historical case data set, wherein the special genotype serial number corresponding to each third data subset is the same value; A second expansion unit is used to expand the second model based on multiple third data subsets to obtain a third model, wherein the third model further considers the additional impact of gene polymorphism compared with the second model; A third expansion unit, configured to expand the third model based on the historical case data set to obtain a fourth model, wherein the fourth model further considers the influence of gender compared to the third model; The fourth expansion unit is used to expand the fourth model based on the historical case data set to obtain the initial prediction model. Compared with the fourth model, the initial prediction model also considers the influence of age, physical condition, nutritional status and pathological state of the body.

8. The device for predicting adverse reactions of chemotherapy drugs according to claim 7, characterized in that: The first training unit comprises: A third training unit is used for a first training operation: inputting the historical case data set into the initial prediction model for a first training, updating the initial prediction model when the value of the first loss function is the minimum, and obtaining an adjusted prediction model, wherein the first training is used to characterize the influence of a fixed medication dose, the influence of a chemotherapy cycle, the influence of gender, the influence of age, the influence of physical condition, the influence of nutritional condition, and the influence of a pathological state of the body, and optimizing the additional influence of gene polymorphism based on a derivation algorithm; The fourth training unit is used for a second training operation: inputting the historical case data set into the adjusted prediction model for a second training, and obtaining an updated initial prediction model when the value of the first loss function decreases; the second training is used to characterize the influence of the dosage of medication, the chemotherapy cycle, the gender, the age, the physical condition, the nutritional condition and the pathological state of the body under the condition of the additional influence of the fixed gene polymorphism, and optimizing the influence based on the optimal gradient algorithm; A fifth calculation unit, configured to perform a calculation operation: inputting the historical case data set into the updated initial prediction model and calculating the first loss function to obtain a loss function value; A repeating unit is used to repeat the first training operation, the second training operation and the calculation operation until the difference between two adjacent loss function values ​​is less than a set threshold, and the updated initial prediction model is used as the expected value prediction sub-model.

9. A device for predicting adverse reactions of chemotherapy drugs, characterized in that: include: Memory for storing computer programs; A processor is used to implement the steps of the method for predicting adverse reactions of chemotherapy drugs as claimed in any one of claims 1 to 4 when executing the computer program.

10. A readable storage medium, characterized in that: The readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the method for predicting adverse reactions of chemotherapy drugs as claimed in any one of claims 1 to 4.

Images