A Method for Constructing a Predictive Model for Chemotherapy Toxicity in Hematological Diseases

Abstract

This invention relates to the field of chemotherapy toxicity prediction modeling, specifically to a method for constructing a chemotherapy toxicity prediction model for hematological diseases. The method obtains the dynamic metabolic coefficient of the patient based on changes in absolute neutrophil count, albumin level, and bilirubin level after the current chemotherapy. It then obtains the injury coefficient based on the drug dosage and toxicity level of the current chemotherapy. Furthermore, it combines the levels of toxicity markers detected before the previous chemotherapy session with the drug exposure dose of the current chemotherapy to obtain the cumulative toxicity coefficient of the patient under the current chemotherapy. Finally, it constructs bone marrow state vectors and liver state vectors, adjusts the liver state vector to obtain the adjusted liver state vector of the patient under the current chemotherapy, and combines multimodal data from all patients throughout all historical chemotherapy periods to construct a chemotherapy toxicity prediction model. This invention can establish an accurate chemotherapy toxicity prediction model and reduce prediction bias.

Description

Technical Field

[0001] This invention relates to the field of chemotherapy toxicity prediction modeling, and specifically to a method for constructing a chemotherapy toxicity prediction model for hematological diseases. Background Technology

[0002] Chemotherapy is an indispensable core treatment for hematological diseases, playing a key role in controlling and alleviating the condition of malignant hematological diseases such as leukemia and lymphoma. However, while chemotherapy drugs kill tumor cells, they also produce serious toxicity to normal tissues and organs, leading to adverse reactions such as bone marrow suppression, liver and kidney damage, and gastrointestinal reactions. These adverse reactions not only affect the patient's quality of life and treatment compliance, but may even endanger life due to toxic reactions.

[0003] In related technologies, models are usually trained based on multimodal data such as the drugs and dosages used by multiple patients during chemotherapy, and various blood component indicators, in order to build corresponding toxicity prediction models. However, due to the great heterogeneity of toxic reactions in patients during chemotherapy, traditional models are difficult to accurately depict the accumulation process of chemotherapy toxicity in the body and the complex interactions between organs, resulting in deviations between the prediction results and the actual toxic reactions. As a result, existing methods cannot establish accurate chemotherapy toxicity prediction models. Summary of the Invention

[0004] To address the technical problem that existing methods cannot establish accurate predictive models of chemotherapy toxicity, the present invention aims to provide a method for constructing a predictive model of chemotherapy toxicity in hematological diseases. The specific technical solution adopted is as follows:

[0005] This invention proposes a method for constructing a predictive model for chemotherapy toxicity in hematological diseases, the method comprising:

[0006] Acquire multimodal data from multiple patients during each historical chemotherapy session;

[0007] Any patient is selected as the test patient. Blood components are tested during the current chemotherapy period, and the dynamic metabolic coefficient of the test patient is determined based on the changes in blood components. The cumulative toxicity coefficient of the test patient during the current chemotherapy is obtained based on the content of various toxicity markers of the test patient before the current chemotherapy and the previous chemotherapy, the pharmacological characteristics of the drug, and the dynamic metabolic coefficient of the test patient.

[0008] Construct the bone marrow state vector and liver state vector of the patient under current chemotherapy, perform state transition analysis based on the bone marrow state vector and liver state vector, adjust the liver state vector, and obtain the adjusted liver state vector of the patient under current chemotherapy.

[0009] A chemotherapy toxicity prediction model was constructed by using multimodal data from all patients during all historical chemotherapy sessions as the training set and combining the cumulative toxicity coefficient of the patient under test in the current chemotherapy with the adjusted liver state vector.

[0010] Furthermore, obtaining the dynamic metabolic coefficient of the patient under test includes:

[0011] Record the number of days it takes for the absolute neutrophil count of the patient under test to recover from the minimum value to the preset normal value after the current chemotherapy, as the recovery days of the patient under test after the current chemotherapy;

[0012] The numerator is the difference between the preset normal value and the minimum absolute neutrophil count of the patient under test after the current chemotherapy, and the denominator is the product of the number of recovery days of the patient under test after the current chemotherapy and the minimum absolute neutrophil count. The ratio is used as the daily neutrophil recovery rate of the patient under test under the current chemotherapy.

[0013] Based on the difference between the albumin level and the standard albumin level of the patient under test after current chemotherapy, and the difference between the bilirubin level and the standard bilirubin level, the liver function regulatory factors of the patient under test during current chemotherapy are obtained.

[0014] The daily recovery rate of neutrophils was adjusted using the liver function regulating factors of the patient under current chemotherapy to obtain the dynamic metabolic coefficient of the patient.

[0015] Furthermore, the acquisition of liver function regulatory factors in the patient undergoing current chemotherapy includes:

[0016] The albumin level of the patient under test after the current chemotherapy is used as the numerator, the standard albumin level is used as the denominator, and the ratio is used as the relative value of the albumin level of the patient under test after the current chemotherapy.

[0017] The ratio of the bilirubin level of the patient after the current chemotherapy to the numerator and the standard bilirubin level to the denominator is used as the relative value of the bilirubin level of the patient after the current chemotherapy.

[0018] The product of the relative values ​​of albumin content and bilirubin level is used as the liver function regulator of the patient under current chemotherapy.

[0019] Furthermore, the process of adjusting the daily recovery rate of neutrophils using the liver function regulating factors of the patient under current chemotherapy to obtain the dynamic metabolic coefficient of the patient includes:

[0020] The product of the liver function regulating factor and the daily recovery rate of neutrophils is used as the dynamic metabolic coefficient of the patient under test.

[0021] Furthermore, the pharmacological characteristics of the medication include dosage and toxicity level. Based on the dosage and toxicity level of the current chemotherapy administered to the patient, an injury coefficient is obtained for the patient. The method for obtaining the injury coefficient for the patient includes:

[0022] The ratio is used as the relative value of the current chemotherapy dosage for the patient under test, with the standard dosage as the numerator and the standard dosage as the denominator.

[0023] The ratio of the toxicity level of the patient under current chemotherapy to the relative value of the drug dosage of the patient under current chemotherapy to the denominator is used as the damage coefficient of the patient under current chemotherapy.

[0024] Furthermore, the pharmacological characteristics also include the drug exposure dose of the current chemotherapy, and obtaining the cumulative toxicity coefficient of the patient under test in the current chemotherapy includes:

[0025] The Euclidean norm of the levels of all toxicity markers in the patient before the current chemotherapy is used as the organ metabolic load of the patient before the current chemotherapy.

[0026] The cumulative toxicity coefficient of the patient under current chemotherapy is calculated using the formula for calculating the cumulative toxicity coefficient.

[0027]

[0028] in, This represents the cumulative toxicity coefficient of the current chemotherapy for the tested patient; This represents the dynamic metabolic coefficient of the patient being tested; This indicates the organ metabolic load of the patient in the previous chemotherapy session before the current chemotherapy; Indicates the injury coefficient of the patient being tested; This indicates the drug exposure dose of the patient being tested during the current chemotherapy.

[0029] Furthermore, constructing the bone marrow state vector and liver state vector of the patient under current chemotherapy includes:

[0030] The vectors formed by standardizing the absolute neutrophil count, reticulocyte percentage, IL-6 level, and G-CSF level in the blood routine report of the patient under current chemotherapy are used as the bone marrow status vector of the patient under current chemotherapy.

[0031] The vectors formed by standardizing the ALT enzyme level, albumin level, bilirubin level, and coagulation function level in the blood routine report of the patient under current chemotherapy are used as the liver status vector of the patient under current chemotherapy.

[0032] Furthermore, the state transition analysis includes: constructing a state transition matrix for the patient under current chemotherapy based on the bone marrow state vector and the liver state vector. Specifically, the bone marrow state vector and liver state vector of the patient under current chemotherapy are input into the Markov chain mathematical model, and the state transition matrix of the patient under current chemotherapy is output.

[0033] Furthermore, obtaining the adjusted liver state vector of the patient under current chemotherapy includes:

[0034] The adjusted liver state vector of the patient under current chemotherapy is obtained using the formula for calculating the adjusted liver state vector.

[0035]

[0036] in, This represents the adjusted liver state vector of the patient under current chemotherapy. This represents the liver state vector of the patient under current chemotherapy. This represents the status transition matrix of the patient under test during the current chemotherapy. This represents the bone marrow status vector of the patient under current chemotherapy. This represents the inverse of the state transition matrix of the patient under test during the current chemotherapy.

[0037] Furthermore, the construction of the chemotherapy toxicity prediction model includes:

[0038] The multimodal data of each patient during all historical chemotherapy sessions are used as a sample data. All sample data are used as the training set for the gradient boosting decision tree. The gradient boosting decision tree is trained to obtain an initial prediction model. The cumulative toxicity coefficient of the patient under the current chemotherapy and the adjusted liver state vector are added to different layers of the decision tree of the initial prediction model to obtain a chemotherapy toxicity prediction model.

[0039] The present invention has the following beneficial effects:

[0040] This invention collects multimodal data from hematological disease patients during their historical chemotherapy sessions, constructs an initial machine learning model based on a gradient boosting decision tree, and then specifically introduces a toxicity accumulation mechanism and an organ state transfer network. By dynamically updating the patient's toxicity accumulation coefficient features and liver state vector features in the gradient boosting decision tree model, dynamic prediction of chemotherapy toxicity is achieved. Compared with traditional prediction methods that rely on a single clinical indicator, this application solves the problems of limited accuracy and inability to capture long-term toxicity accumulation and organ cascade effects in existing technologies. It can more accurately predict the hematological toxicity risk of hematological disease patients undergoing chemotherapy, providing a basis for personalized clinical treatment. Attached Figure Description

[0041] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0042] Figure 1 The flowchart illustrates a method for constructing a chemotherapy toxicity prediction model for hematological diseases, as provided in one embodiment of the present invention. Detailed Implementation

[0043] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a method for constructing a chemotherapy toxicity prediction model for hematological diseases based on the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0044] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0045] The following describes in detail, with reference to the accompanying drawings, a specific scheme for constructing a chemotherapy toxicity prediction model for hematological diseases provided by the present invention.

[0046] Please see Figure 1 The diagram illustrates a flowchart of a method for constructing a chemotherapy toxicity prediction model for hematological diseases according to an embodiment of the present invention. The method includes:

[0047] Step S1: Obtain multimodal data for multiple patients during each historical chemotherapy session.

[0048] This embodiment of the invention first collects multimodal data from multiple patients during each historical chemotherapy session, wherein the multimodal data includes:

[0049] Metabolic monitoring data: Plasma samples were collected at 0h, 12h, 24h and 48h after each historical chemotherapy session. The concentration of SN-38 (irinotecan active metabolite) was quantified by LC-MS / MS, and urinary 8-OHdG (DNA oxidative damage marker) was dynamically monitored.

[0050] Symptom records during treatment: such as high fever, grade of oral mucosal inflammation, daily blood oxygen saturation, etc.

[0051] Genomic testing data: Peripheral blood samples were collected from patients, and targeted panels were used preferentially to cover core drug metabolism genes such as DPYD, GSTP1, and CYP2D6.

[0052] Imaging data: Collect imaging data of patients before and after each historical chemotherapy session, mainly functional images. Extraction and quantification of imaging features: including liver stiffness value (kPa), spleen length diameter (cm), portal vein flow velocity (cm / s), etc., to assess organ morphological and functional changes and help determine the impact of toxic reactions on organs.

[0053] Toxicity grading: Blood components were tested during each of the patient's previous chemotherapy sessions, and based on the test results, post-chemotherapy blood toxicity was graded using the existing CTCAE v5.0 standard, with toxicity levels ranging from 1 to 4.

[0054] Step S2: Select any patient as the test patient, perform blood component testing on the test patient during the current chemotherapy, and determine the dynamic metabolic coefficient of the test patient based on the changes in blood components; obtain the cumulative toxicity coefficient of the test patient during the current chemotherapy based on the content of various toxicity markers of the test patient before the current chemotherapy and the previous chemotherapy, the pharmacological characteristics of the drug, and the dynamic metabolic coefficient of the test patient.

[0055] Because the toxic reactions experienced by patients during chemotherapy are highly heterogeneous, directly using the collected data for model training would result in a model that cannot accurately depict the cumulative process of chemotherapy toxicity in the body and the complex interactions between organs, leading to a deviation between the predicted results and the actual toxic reactions. Therefore, this embodiment of the invention first analyzes any one patient as the test patient and uses a blood component detection device to perform blood component detection on the test patient during the current chemotherapy period, thereby collecting data such as the content of various toxicity detection markers and drug exposure doses of the test patient for subsequent calculation and analysis. The content of toxicity detection markers includes absolute neutrophil count, albumin content, bilirubin level, etc.

[0056] The residual effects of chemotherapy drugs in organs follow a dynamic balance between damage and repair. Taking platinum-based drugs as an example, the half-life of the adducts they form with DNA can reach 14-21 days, causing the bone marrow hematopoietic microenvironment to be continuously exposed to damage stress. Even after drug withdrawal, toxicity continues to accumulate. Human organs such as the liver can affect the body's metabolic capacity and toxicity accumulation. Therefore, in this embodiment of the invention, the dynamic metabolic coefficient of the patient is obtained based on the change in the absolute neutrophil count after the current chemotherapy, as well as the albumin content and bilirubin level after the current chemotherapy. Subsequently, the toxicity accumulation of the patient under the current chemotherapy can be accurately analyzed based on the dynamic metabolic coefficient.

[0057] Preferably, in one embodiment of the present invention, the method for obtaining the dynamic metabolic coefficient of the patient to be tested specifically includes:

[0058] First, the number of days it takes for the absolute neutrophil count of the patient under test to recover from its minimum value to a preset normal value after the current chemotherapy is recorded as the recovery days of the patient under test after the current chemotherapy. In one embodiment of the present invention, the preset normal value is set to... The default normal value can also be set by the implementer according to the specific implementation scenario, and is not limited here.

[0059] The ratio of the difference between the preset normal value and the minimum absolute neutrophil count of the patient under test after the current chemotherapy is used as the numerator, and the recovery days of the patient under test after the current chemotherapy is used as the denominator. The ratio is used as the daily neutrophil recovery rate of the patient under test under the current chemotherapy.

[0060] As an example, in one embodiment of the present invention, the expression for the daily neutrophil recovery rate of the patient under current chemotherapy can be specifically as follows:

[0061]

[0062] in, This indicates the daily neutrophil recovery rate of the patient under test during the current chemotherapy regimen; This indicates the default normal value; This represents the minimum absolute neutrophil count in the patient being tested after the current chemotherapy. This indicates the number of days the patient under test has recovered after the current chemotherapy.

[0063] Among them, the daily neutrophil recovery rate of the patients under test during current chemotherapy The larger the value, the faster the individual bone marrow microenvironment reconstruction capacity and repair rate of the tested patient, and the more significant the attenuation of historical toxic residues.

[0064] Then, since abnormal liver function can inhibit the activity of enzymes that metabolize chemotherapy drugs such as cisplatin, the liver function regulatory factors of the patient under current chemotherapy can be obtained based on the differences between the albumin level and the standard albumin level and the bilirubin level after the current chemotherapy.

[0065] Preferably, in one embodiment of the present invention, the method for obtaining liver function regulatory factors of the patient under current chemotherapy specifically includes:

[0066] The numerator is the albumin level of the patient under test after current chemotherapy, the denominator is the standard albumin level, and the ratio is used as the relative value of the albumin level of the patient under test after current chemotherapy. The standard albumin level refers to the albumin level in a healthy human body, which is a known value.

[0067] The ratio of the patient's current post-chemotherapy bilirubin level to the standard bilirubin level is used as the numerator and the patient's current post-chemotherapy bilirubin level to the denominator. The ratio is used as the relative value of the patient's current post-chemotherapy bilirubin level. Similarly, the standard bilirubin level refers to the level of bilirubin in a healthy human body and is a known value.

[0068] The product of the relative values ​​of albumin content and bilirubin level is then used as the liver function regulatory factor for the patient undergoing current chemotherapy. This liver function regulatory factor comprehensively reflects the functional status of the liver in terms of synthesis, detoxification, and excretion. The greater the deviation of the liver function regulatory factor from the value of 1, the greater the impact of liver function on drug metabolism. When the liver function regulatory factor is less than 1, it indicates impaired liver function, decreased drug metabolism and detoxification capacity, and an increased possibility of toxicity accumulation. When the liver function regulatory factor is greater than 1, it indicates that the liver is under stress or has other abnormalities, and changes in drug toxicity also need to be monitored.

[0069] As an example, in one embodiment of the present invention, the expression for the liver function regulator of the patient under current chemotherapy can be specifically as follows:

[0070]

[0071] in, This indicates the liver function regulatory factors of the patient under current chemotherapy. This indicates the albumin level of the patient being tested after the current chemotherapy. Indicates standard albumin content; This indicates the relative albumin level of the patient being tested during their current chemotherapy regimen; This indicates the bilirubin level of the patient being tested after the current chemotherapy. This indicates the standard bilirubin level; This indicates the relative bilirubin level of the patient under test during the current chemotherapy regimen.

[0072] Furthermore, by utilizing the liver function regulatory factors of the patients undergoing current chemotherapy, the daily recovery rate of neutrophils is adjusted to obtain the dynamic metabolic coefficient of the patients. This allows for a comprehensive and dynamic consideration of the impact of individual organ function differences on drug metabolism and toxicity clearance, making the subsequent calculation and analysis of toxicity accumulation more accurate. This, in turn, enables a more accurate assessment of the toxicity accumulation of chemotherapy drugs in the patient's body, providing a more reliable basis for subsequent toxicity risk prediction modeling.

[0073] Preferably, in one embodiment of the present invention, the method for obtaining the dynamic metabolic coefficient of the patient to be tested further includes:

[0074] The product of liver function regulatory factors and daily neutrophil recovery rate was used as the dynamic metabolic coefficient of the patient under test.

[0075] As an example, in one embodiment of the present invention, the expression for the dynamic metabolic coefficient of the patient to be tested can be specifically as follows:

[0076]

[0077] in, This represents the dynamic metabolic coefficient of the patient being tested; This indicates the liver function regulatory factors of the patient under current chemotherapy. This indicates the daily neutrophil recovery rate of the patient under test during the current chemotherapy regimen.

[0078] The pharmacological characteristics of the drug include dosage, toxicity level, and current chemotherapy drug exposure.

[0079] Considering that chemotherapy drugs can also cause certain damage to the human body, this embodiment of the invention analyzes the dosage and toxicity level of the current chemotherapy in the test patient to obtain the damage coefficient of the test patient. Subsequently, the damage coefficient and dynamic metabolic coefficient of the test patient can be combined to accurately analyze the toxicity accumulation of the test patient.

[0080] Preferably, in one embodiment of the present invention, the method for obtaining the injury coefficient of the patient to be tested specifically includes:

[0081] Using the current chemotherapy dosage of the patient as the numerator and the standard dosage as the denominator, the ratio is taken as the relative value of the current chemotherapy dosage for the patient. Under the same dosage, the more severe the toxic reaction, the stronger the damage efficacy produced by the unit dosage. Therefore, the toxicity level of the patient under the current chemotherapy can be used as the numerator, and the relative value of the current chemotherapy dosage for the patient under the current chemotherapy can be used as the denominator. The ratio is taken as the damage coefficient for the patient. The standard dosage is a known value. For example, the standard dosage threshold of cisplatin is 100 mg / m².

[0082] As an example, in one embodiment of the present invention, the expression for the injury coefficient of the patient to be tested can be specifically as follows:

[0083]

[0084] in, Indicates the injury coefficient of the patient being tested; This indicates the toxicity level of the patient under test in the current chemotherapy regimen; This indicates the dosage of chemotherapy currently being administered to the patient being tested; Indicates the standard dosage; This indicates the relative dosage of the current chemotherapy for the patient being tested.

[0085] Then, the cumulative toxicity of the patient under test in the current chemotherapy is analyzed. The higher the levels of various toxicity markers in the patient before the current chemotherapy and the drug exposure dose of the current chemotherapy, the more severe the cumulative toxicity of the chemotherapy drugs in the patient. Therefore, the cumulative toxicity coefficient of the patient under test in the current chemotherapy can be obtained based on the levels of various toxicity markers in the patient before the current chemotherapy, the drug exposure dose of the current chemotherapy, and the dynamic metabolic coefficient and damage coefficient of the patient obtained above.

[0086] Preferably, in one embodiment of the present invention, the method for obtaining the cumulative toxicity coefficient of the patient under current chemotherapy specifically includes:

[0087] The Euclidean norm of the levels of all toxicity markers detected in the patient before the current chemotherapy was used as the organ metabolic load of the patient before the current chemotherapy.

[0088] The cumulative toxicity coefficient of the patient under current chemotherapy is calculated using the formula for calculating the cumulative toxicity coefficient. The formula for calculating the cumulative toxicity coefficient is as follows:

[0089]

[0090] in, This represents the cumulative toxicity coefficient of the current chemotherapy for the tested patient; This represents the dynamic metabolic coefficient of the patient being tested; This indicates the organ metabolic load of the patient in the previous chemotherapy session before the current chemotherapy; Indicates the injury coefficient of the patient being tested; This indicates the drug exposure dose of the patient being tested during the current chemotherapy.

[0091] in, This represents the residual toxicity from the previous chemotherapy under the influence of the patient's metabolic function. It is the product of the patient's damage coefficient and the current chemotherapy drug exposure dose, representing immediate damage, while the sum of the two represents the cumulative toxicity after current chemotherapy.

[0092] It should be noted that the first chemotherapy for the patient to be tested does not have a previous chemotherapy. In this case, the Euclidean norm of the levels of all toxicity markers of the patient to be tested before the first chemotherapy can be used as the organ metabolic load of the patient to be tested during the first chemotherapy, and the organ metabolic load of the first chemotherapy can be used as the cumulative toxicity coefficient of the patient to be tested during the current chemotherapy.

[0093] Step S3: Construct the bone marrow state vector and liver state vector of the patient under current chemotherapy. Perform state transition analysis based on the bone marrow state vector and liver state vector, adjust the liver state vector, and obtain the adjusted liver state vector of the patient under current chemotherapy.

[0094] Specifically, the bone marrow state vector and liver state vector of the patient under current chemotherapy are constructed. Based on the bone marrow state vector and liver state vector, the state transition matrix of the patient under current chemotherapy is constructed. The liver state vector is adjusted by combining the bone marrow state vector and the state transition matrix to obtain the adjusted liver state vector of the patient under current chemotherapy.

[0095] Bone marrow suppression inhibits liver CYP enzyme activity through inflammatory factors (IL-6), and intestinal metabolites exacerbate the liver burden through portal vein circulation, forming an organ cascade effect. Therefore, this embodiment of the invention needs to establish a dynamic coupling relationship between organ biomarkers and realize state transfer through vector space mapping. First, construct the bone marrow state vector and liver state vector of the patient under current chemotherapy. Subsequently, a state transition matrix on the regulation of liver state by bone marrow state can be constructed based on the bone marrow state vector and liver state vector.

[0096] Preferably, in one embodiment of the present invention, the method for obtaining the bone marrow state vector and liver state vector of the patient under current chemotherapy specifically includes:

[0097] The vectors formed by standardizing the absolute neutrophil count, reticulocyte percentage, IL-6 level, and G-CSF level in the blood routine report of the patient under current chemotherapy are used as the bone marrow status vector of the patient under current chemotherapy. The bone marrow status vector is in the form of (absolute neutrophil count, reticulocyte percentage, IL-6 level, G-CSF level).

[0098] The vectors formed by standardizing the ALT enzyme level, albumin level, bilirubin level, and coagulation function level (INR) of the patient under current chemotherapy are used as the liver status vector of the patient under current chemotherapy. The liver status vector is in the form of (ALT enzyme level, albumin level, bilirubin level, coagulation function level).

[0099] The pathological logic from bone marrow to liver is as follows: On the one hand, bone marrow suppression leads to an increase in IL-6, which then inhibits the activity of liver CYP3A4 enzyme, resulting in a decrease in drug metabolism capacity. On the other hand, an abnormal increase in G-CSF activates liver Kupffer cells, thereby exacerbating the inflammatory response. The pathological logic from liver to bone marrow is as follows: On the one hand, an increase in coagulation function leads to coagulation dysfunction, which in turn makes the bone marrow microenvironment prone to bleeding. On the other hand, bilirubin accumulation inhibits the proliferation of hematopoietic stem cells, thereby delaying the recovery of coagulation function. Based on the above pathological logic, the state transition matrix of the patient under current chemotherapy can be constructed according to the bone marrow state vector and the liver state vector.

[0100] Preferably, in one embodiment of the present invention, the method for obtaining the state transition matrix of the patient under current chemotherapy specifically includes:

[0101] The bone marrow state vector and liver state vector of the patient under current chemotherapy are input into a Markov chain mathematical model, and the output is the state transition matrix of the patient under current chemotherapy. For the state transition matrix, the row index corresponds to the dimension of the liver state vector, the column index corresponds to the dimension of the bone marrow state vector, and the matrix elements are... The first element representing the bone marrow state vector The dimensional of the liver state vector The regulatory intensity of the dimension, with positive and negative signs corresponding to activation and inhibition effects, respectively. The Markov chain mathematical model is a well-known technique in the field and will not be elaborated here.

[0102] Then, by combining the bone marrow state vector and the state transition matrix, the liver state vector is adjusted to obtain the adjusted liver state vector of the patient under the current chemotherapy. Subsequently, by combining the adjusted liver state vector of the patient under the current chemotherapy and the cumulative toxicity coefficient, an accurate chemotherapy toxicity prediction model can be constructed.

[0103] Preferably, in one embodiment of the present invention, the method for obtaining the adjusted liver state vector of the patient under current chemotherapy specifically includes:

[0104] The adjusted liver state vector of the patient under current chemotherapy is obtained using the formula for calculating the adjusted liver state vector. The formula for calculating the adjusted liver state vector is as follows:

[0105]

[0106] in, This represents the adjusted liver state vector of the patient under current chemotherapy. This represents the liver state vector of the patient under current chemotherapy. This represents the status transition matrix of the patient under test during the current chemotherapy. This represents the bone marrow status vector of the patient under current chemotherapy. This represents the inverse of the state transition matrix of the patient under test during the current chemotherapy.

[0107] in, This represents the dynamic impact of bone marrow status on the liver. This represents the feedback influence of the liver on the bone marrow, and the liver state vector. In addition to the dynamic impact of bone marrow status on the liver And subtract the characteristics of bone marrow being affected by feedback from the liver. This is to avoid the continuous interaction and impact between organs.

[0108] Step S4: Use multimodal data from all patients during all historical chemotherapy sessions as the training set, and combine the cumulative toxicity coefficient of the patient under test in the current chemotherapy and the adjusted liver state vector to construct a chemotherapy toxicity prediction model.

[0109] After obtaining the cumulative toxicity coefficient and adjusted liver state vector of the patient under test in the current chemotherapy, the model can be trained using multimodal data of all patients during all historical chemotherapy sessions. This solves the problems of limited accuracy of existing prediction models and inability to capture long-term toxicity accumulation and organ cascade effects, resulting in a more accurate chemotherapy toxicity prediction model.

[0110] Preferably, in one embodiment of the present invention, the method for constructing a chemotherapy toxicity prediction model specifically includes:

[0111] The multimodal data of each patient during all historical chemotherapy sessions are used as a sample data. All sample data are used as the training set for the gradient boosting decision tree. The gradient boosting decision tree is trained to obtain an initial prediction model. The cumulative toxicity coefficient of the patient under the current chemotherapy and the adjusted liver state vector are added to different layers of the decision tree of the initial prediction model to obtain a chemotherapy toxicity prediction model. The gradient boosting decision tree is a well-known technique in the art and will not be described in detail here.

[0112] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0113] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A method for constructing a predictive model for chemotherapy toxicity in hematological diseases, characterized in that, The method includes: Acquire multimodal data from multiple patients during each historical chemotherapy session. The multimodal data includes metabolic testing data, symptom records during treatment, genomic testing data, imaging data, and toxicity grading. Any patient is selected as the test patient. Blood components are measured during the current chemotherapy period, and the dynamic metabolic coefficient of the test patient is determined based on the changes in blood components. The cumulative toxicity coefficient of the test patient during the current chemotherapy is obtained based on the levels of various toxicity markers in the test patient before the current chemotherapy and the previous chemotherapy, the pharmacological characteristics of the drug, and the dynamic metabolic coefficient of the test patient. The pharmacological characteristics include drug dosage, drug toxicity level, and drug exposure dose. Construct the bone marrow state vector and liver state vector of the patient under current chemotherapy, and perform state transition analysis based on the bone marrow state vector and liver state vector. The state transition analysis includes: constructing the state transition matrix of the patient under current chemotherapy based on the bone marrow state vector and the liver state vector. Specifically, the bone marrow state vector and liver state vector of the patient under current chemotherapy are input into the Markov chain mathematical model, and the state transition matrix of the patient under current chemotherapy is output. The liver state vector is adjusted by combining the bone marrow state vector and the state transition matrix to obtain the adjusted liver state vector of the patient under current chemotherapy. A chemotherapy toxicity prediction model was constructed by using multimodal data from all patients during all historical chemotherapy sessions as the training set and combining the cumulative toxicity coefficient of the patient under test in the current chemotherapy with the adjusted liver state vector. Methods for determining the bone marrow state vector and liver state vector include: The vectors formed by standardizing the absolute neutrophil count, reticulocyte percentage, IL-6 level, and G-CSF level in the blood routine report of the patient under current chemotherapy are used as the bone marrow status vector of the patient under current chemotherapy. The vectors formed by standardizing the ALT enzyme level, albumin level, bilirubin level, and coagulation function level in the blood routine report of the patient under current chemotherapy are used as the liver status vector of the patient under current chemotherapy. The process of obtaining the adjusted liver state vector of the patient under current chemotherapy includes: The adjusted liver state vector of the patient under current chemotherapy is obtained using the formula for calculating the adjusted liver state vector. in, This represents the adjusted liver state vector of the patient under current chemotherapy. This represents the liver state vector of the patient under current chemotherapy. This represents the status transition matrix of the patient under test during the current chemotherapy. This represents the bone marrow status vector of the patient under current chemotherapy. This represents the inverse of the state transition matrix of the patient under test during the current chemotherapy. The constructed chemotherapy toxicity prediction model includes: The multimodal data of each patient during all historical chemotherapy sessions are used as a sample data. All sample data are used as the training set for the gradient boosting decision tree. The gradient boosting decision tree is trained to obtain an initial prediction model. The cumulative toxicity coefficient of the patient under the current chemotherapy and the adjusted liver state vector are added to different layers of the decision tree of the initial prediction model to obtain a chemotherapy toxicity prediction model.

2. The method for constructing a predictive model for chemotherapy toxicity in hematological diseases according to claim 1, characterized in that, The dynamic metabolic coefficients of the patients to be tested are obtained as follows: Record the number of days it takes for the absolute neutrophil count of the patient under test to recover from the minimum value to the preset normal value after the current chemotherapy, as the recovery days of the patient under test after the current chemotherapy; The numerator is the difference between the preset normal value and the minimum absolute neutrophil count of the patient under test after the current chemotherapy, and the denominator is the number of recovery days of the patient under test after the current chemotherapy. The ratio is used as the daily recovery rate of neutrophils of the patient under test during the current chemotherapy. Based on the difference between the albumin level and the standard albumin level of the patient under test after current chemotherapy, and the difference between the bilirubin level and the standard bilirubin level, the liver function regulatory factors of the patient under test during current chemotherapy are obtained. The daily recovery rate of neutrophils was adjusted using the liver function regulating factors of the patient under current chemotherapy to obtain the dynamic metabolic coefficient of the patient.

3. The method for constructing a predictive model for chemotherapy toxicity in hematological diseases according to claim 2, characterized in that, The liver function regulators obtained from the patient under current chemotherapy include: The albumin level of the patient under test after the current chemotherapy is used as the numerator, the standard albumin level is used as the denominator, and the ratio is used as the relative value of the albumin level of the patient under test after the current chemotherapy. The ratio of the bilirubin level of the patient after the current chemotherapy to the numerator and the standard bilirubin level to the denominator is used as the relative value of the bilirubin level of the patient after the current chemotherapy. The product of the relative values ​​of albumin content and bilirubin level is used as the liver function regulator of the patient under current chemotherapy.

4. The method for constructing a predictive model for chemotherapy toxicity in hematological diseases according to claim 2, characterized in that, The process of adjusting the daily recovery rate of neutrophils using the liver function regulating factors of the patient under current chemotherapy to obtain the dynamic metabolic coefficient of the patient includes: The product of the liver function regulating factor and the daily recovery rate of neutrophils is used as the dynamic metabolic coefficient of the patient under test.

5. The method for constructing a predictive model for chemotherapy toxicity in hematological diseases according to claim 1, characterized in that, The pharmacological characteristics of the medication include dosage and toxicity level. Based on the dosage and toxicity level of the current chemotherapy regimen, the injury coefficient of the patient is obtained. The injury coefficient of the patient is obtained by: The ratio is used as the relative value of the current chemotherapy dosage for the patient under test, with the standard dosage as the numerator and the standard dosage as the denominator. The ratio of the toxicity level of the patient under current chemotherapy to the relative value of the drug dosage of the patient under current chemotherapy to the denominator is used as the damage coefficient of the patient under current chemotherapy.

6. The method for constructing a predictive model for chemotherapy toxicity in hematological diseases according to claim 5, characterized in that, The pharmacological characteristics also include the drug exposure dose of the current chemotherapy, and obtaining the cumulative toxicity coefficient of the patient under test for the current chemotherapy includes: The Euclidean norm of the levels of all toxicity markers in the patient before the current chemotherapy is used as the organ metabolic load of the patient before the current chemotherapy. The cumulative toxicity coefficient of the patient under current chemotherapy is calculated using the formula for calculating the cumulative toxicity coefficient. in, This represents the cumulative toxicity coefficient of the current chemotherapy for the tested patient; This represents the dynamic metabolic coefficient of the patient being tested; This indicates the organ metabolic load of the patient in the previous chemotherapy session before the current chemotherapy; Indicates the injury coefficient of the patient being tested; This indicates the drug exposure dose of the patient being tested during the current chemotherapy.

Images