CINV patient risk prediction model construction method

Abstract

The invention relates to the technical field of model construction, in particular to a CINV patient risk prediction model construction method, which comprises the following steps: extracting a plurality of real-time vomiting indications of any patient in a preset period and corresponding real-time vomiting characterization data, and analyzing the correlation between the plurality of real-time vomiting indications and the real-time vomiting characterization data; determining a plurality of real-time significant representation variables corresponding to the real-time vomiting indications based on the real-time correlation analysis result; matching the plurality of real-time vomiting indications with vomiting indications stored in a data resource library, and determining a plurality of matching characterization variables based on a matching result; determining a real-time vomiting main factor based on the plurality of real-time significant characterization variables and the plurality of matched characterization variables; determining a predicted vomiting indication based on the real-time vomiting main factor and a CINV patient risk prediction model; and correcting the CINV patient risk prediction model based on a comparison result of the predicted vomiting indication and the real-time vomiting indication, or adjusting a preset period. The accuracy of the prediction model is effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of model construction, and in particular to a method for constructing a risk prediction model for CINV patients. Background Art

[0002] Chemotherapy-induced nausea and vomiting (CINV) is a common adverse reaction in cancer patients during chemotherapy, which can further lead to adverse consequences such as dehydration and metabolic disorders, often seriously affecting the quality of life of patients and reducing treatment compliance. In the current field of medical technology, risk prediction for patients with chemotherapy-induced nausea and vomiting usually uses statistical analysis methods based on historical data. These methods rely on the patient's past vomiting records, chemotherapy regimens, and general physiological information, and construct prediction models through regression analysis or machine learning algorithms.

[0003] The patent document with Chinese patent application publication number CN118039161A discloses a construction method and system based on an internal medicine patient risk prediction model, the method comprising the following steps: obtaining internal medicine medical information data; constructing an entity relationship data set for the internal medicine medical information data to generate internal medicine medical information practice-relationship data; mapping the internal medicine medical information practice-relationship data to a low-dimensional vector space for entity linking to generate internal medicine medical information link data; performing word sense disambiguation on the internal medicine medical information link data to generate internal medicine medical information disambiguation data; obtaining clinical feature data of internal medicine patients; performing multimodal feature fusion on the clinical feature data of internal medicine patients and the internal medicine medical information disambiguation data to generate multimodal feature data of internal medicine patients.

[0004] In the prior art, when establishing a preset model, most of them are based on historical data and ignore real-time physiological data, resulting in limited accuracy of the prediction results. Summary of the invention

[0005] To this end, the present invention provides a method for constructing a CINV patient risk prediction model, which can solve the problem of inaccurate CINV patient risk prediction model.

[0006] To achieve the above object, the present invention provides a method for constructing a CINV patient risk prediction model, the method comprising:

[0007] Extracting a number of real-time vomiting indications and their corresponding real-time vomiting characterization data of any patient within a preset period, analyzing the correlation between the number of real-time vomiting indications and their corresponding real-time vomiting characterization data, and obtaining a real-time correlation analysis result;

[0008] Determine a number of real-time significant characterization variables corresponding to the real-time vomiting indication based on the real-time correlation analysis result;

[0009] Matching the real-time vomiting indications with the vomiting indications stored in the data resource library, and determining a plurality of matching characterization variables based on the matching results;

[0010] Determine a real-time vomiting main factor based on the plurality of real-time significant characterization variables and the plurality of matching characterization variables;

[0011] Determining predicted vomiting indications based on the real-time vomiting main factors and the CINV patient risk prediction model;

[0012] The CINV patient risk prediction model is modified based on the comparison result between the predicted vomiting indication and the real-time vomiting indication, or the preset period is adjusted.

[0013] Furthermore, the step of extracting a plurality of real-time vomiting indications of any patient within a preset period and their corresponding real-time vomiting characterization data includes:

[0014] Acquire a plurality of the real-time vomiting indications by extracting a plurality of real-time vomiting intervals and their corresponding real-time vomiting durations of the patient during a plurality of chemotherapy processes;

[0015] A number of real-time chemotherapy data and a number of real-time physiological data during the patient's several chemotherapy processes are extracted to determine a number of real-time vomiting characterization data, wherein the number of real-time chemotherapy data include a number of real-time chemotherapy drug types and their corresponding real-time doses, and the number of real-time physiological data include a number of real-time blood pressure change values, a number of real-time heart rate change values, and a number of real-time body temperature change values ​​before and after the chemotherapy process.

[0016] Furthermore, the step of analyzing the correlation between a plurality of real-time vomiting indications and their corresponding real-time vomiting characterization data includes:

[0017] performing univariate correlation analysis on the real-time types of chemotherapy drugs, the real-time dosages, the real-time blood pressure change values, the real-time heart rate change values, the real-time body temperature change values ​​and the real-time vomiting intervals, respectively, to obtain a plurality of first correlations according to the analysis results;

[0018] Univariate correlation analysis is performed on several of the real-time chemotherapy drug types, several of the real-time dosages, several of the real-time blood pressure change values, several of the real-time heart rate change values, and several of the real-time body temperature change values ​​with several of the real-time vomiting durations, so as to obtain several second correlations based on the analysis results.

[0019] Further, the step of obtaining the first correlation according to the analysis result includes:

[0020] The real-time chemotherapy drug type, the real-time dosage, the real-time blood pressure change value, the real-time heart rate change value and the real-time body temperature change value are used as a number of real-time variables;

[0021] Respectively calculating the real-time correlation coefficients corresponding to any of the real-time variables and a number of the real-time vomiting intervals and the corresponding real-time significance judgment values;

[0022] A plurality of first correlations corresponding to a plurality of the real-time variables are determined based on the real-time correlation coefficient and the real-time significance judgment value.

[0023] Furthermore, the step of determining a plurality of real-time significant characterization variables corresponding to the real-time vomiting indication based on the real-time correlation analysis result includes:

[0024] Respectively extracting first real-time characterization variables and second real-time characterization variables corresponding to the real-time significant judgment values ​​being less than the preset significant judgment values ​​from the first correlations and the second correlations;

[0025] The intersection of a plurality of the first real-time characterization variables and a plurality of the second real-time characterization variables is selected as the real-time significant characterization variable, and is sorted according to the real-time significant correlation coefficients corresponding to the real-time significant characterization variables.

[0026] Furthermore, the step of determining a plurality of matching characterization variables based on the matching result includes:

[0027] Matching the real-time vomiting interval and the real-time vomiting duration corresponding to any of the real-time vomiting indications with the stored vomiting interval and the stored vomiting duration corresponding to the stored vomiting indications in the data resource library;

[0028] The stored characterization data corresponding to the same stored vomiting interval and stored vomiting duration in the matching result are selected as the matching characterization variable.

[0029] Furthermore, the step of determining the real-time vomiting main factor based on the plurality of real-time significant characterization variables and the plurality of matching characterization variables comprises:

[0030] Comparing the plurality of real-time significant characterization variables with the plurality of matching characterization variables to obtain a comparison result;

[0031] Based on the comparison result, the variable intersection of several of the real-time significant characterization variables and several of the matching characterization variables is selected as the real-time vomiting main factor.

[0032] Further, based on the comparison result between the predicted vomiting indication and the real-time vomiting indication, the CINV patient risk prediction model is modified, or the step of adjusting the preset period includes:

[0033] comparing the predicted vomiting indication with the real-time vomiting indication, and calculating a result difference value based on the comparison result;

[0034] The result difference value is compared with a preset difference value range, and the CINV patient risk prediction model is modified according to the comparison result, or the preset period is adjusted.

[0035] Further, the step of modifying the CINV patient risk prediction model according to the comparison result, or adjusting the preset period includes:

[0036] When the time interval difference value is less than or equal to the preset time interval difference range and the duration difference value is less than or equal to the preset duration difference range, the CINV patient risk prediction model is modified;

[0037] When the time interval difference value is greater than the preset time interval difference range or the duration difference value is greater than the preset duration difference range, the preset period is adjusted, several vomiting indications and their corresponding vomiting characterization data are recollected, and a new CINV patient risk prediction model is constructed based on the above data.

[0038] Furthermore, the step of modifying the CINV patient risk prediction model includes:

[0039] Matching the predicted vomiting indication with the vomiting indication stored in the data resource library, and determining the predicted characterization variable based on the matching result;

[0040] The predicted characterization variable is compared with the real-time vomiting main factor, and the corresponding parameter weights in the CINV patient risk prediction model are adjusted based on the comparison result to modify the CINV patient risk prediction model.

[0041] Compared with the prior art, the beneficial effect of the present invention lies in that, by extracting a number of real-time vomiting indications and their corresponding real-time vomiting characterization data of any patient within a preset period and analyzing the correlation, the intrinsic connection between the patient's vomiting-related conditions within a specific time period can be fully understood, and basic data support can be provided for the subsequent accurate construction of a risk prediction model. By analyzing the correlation, it is helpful to determine the key factors more targetedly, and by screening the real-time significant characterization variables, the variables that are of great significance to vomiting prediction can be screened out, the interference of irrelevant or minor factors can be reduced, and the accuracy and efficiency of the risk prediction model can be improved. Only focusing on the significant characterization variables makes the model more concise and efficient, and reduces the computational complexity. By comparing the real-time vomiting indications with the vomiting indications stored in the data resource library, the real-time vomiting indications can be compared with the vomiting indications stored in the data resource library. Matching is performed and several matching characterization variables are determined. Existing data resources are fully utilized to enrich the variable sources of the model. The commonalities and differences between the current patients and the existing patient groups are discovered through matching, providing a reference for more accurate risk prediction. The determination of the main factors of real-time vomiting helps to identify the core factors affecting the risk of vomiting, providing a key basis for the prediction of the model. The prediction using the main factors improves the accuracy and reliability of the prediction. By comparing the predicted vomiting indications with the real-time vomiting indications, the model can be continuously optimized so that it can be continuously improved with the addition of new data and changes in actual conditions. By correcting the model, its prediction accuracy and adaptability can be improved. The preset cycle can be adjusted to be personalized according to the conditions and needs of different patients, further improving the practicality of the model.

[0042] In particular, by performing univariate correlation analysis separately, the degree of influence of each factor on the vomiting interval can be clearly determined, which helps to determine which factors will speed up or slow down the frequency of vomiting. Similarly, a similar univariate correlation analysis is performed on the real-time vomiting duration to clarify the effect of each factor on the vomiting duration. The correlation analysis results are an important basis for subsequent steps such as determining real-time significant characterization variables, matching characterization variables, and determining real-time vomiting main factors, avoiding the inclusion of irrelevant factors in the model, thereby improving the accuracy and effectiveness of the entire CINV patient risk prediction model.

[0043] In particular, by comparing the predicted vomiting indications with the real-time vomiting indications and calculating the difference value of the results, the prediction error of the model can be quantified, and the model can be corrected or the preset period can be adjusted based on the comparison result of the difference value with the preset difference value range, thereby continuously improving the accuracy of the model. By comparing the results, it is decided whether to correct the model or adjust the preset period, thereby improving the accuracy of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 A schematic diagram of a process for constructing a CINV patient risk prediction model provided by an embodiment of the present invention;

[0045] Figure 2A schematic diagram of a process for extracting real-time vomiting indications and real-time vomiting characterization data in a method for constructing a CINV patient risk prediction model provided by an embodiment of the present invention;

[0046] Figure 3 A schematic diagram of a process for determining real-time significant characterization variables in a method for constructing a CINV patient risk prediction model provided by an embodiment of the present invention;

[0047] Figure 4 A flowchart of step S600 in the method for constructing a CINV patient risk prediction model provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0048] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0049] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the protection scope of the present invention.

[0050] It should be noted that, in the description of the present invention, terms such as "up", "down", "left", "right", "inside" and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the drawings. This is merely for the convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it should not be understood as a limitation on the present invention.

[0051] In addition, it should be noted that in the description of the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected" and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, or it can be the internal communication of two components. For those skilled in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0052] See also Figure 1 As shown, an embodiment of the present invention provides a method for constructing a CINV patient risk prediction model, the method comprising:

[0053] Step S100, extracting a plurality of real-time vomiting indications and their corresponding real-time vomiting characterization data of any patient within a preset period, analyzing the correlation between the plurality of real-time vomiting indications and their corresponding real-time vomiting characterization data, and obtaining a real-time correlation analysis result;

[0054] Step S200, determining a plurality of real-time significant characterization variables corresponding to the real-time vomiting indication based on the real-time correlation analysis result;

[0055] Step S300, matching the real-time vomiting indications with the vomiting indications stored in the data resource library, and determining a plurality of matching representation variables based on the matching results;

[0056] Step S400, determining a real-time vomiting main factor based on a plurality of the real-time significant characterization variables and a plurality of the matching characterization variables;

[0057] Step S500, determining predicted vomiting indications based on the real-time vomiting main factors and the CINV patient risk prediction model;

[0058] Step S600, based on the comparison result between the predicted vomiting indication and the real-time vomiting indication, the CINV patient risk prediction model is modified, or the preset period is adjusted.

[0059] It is understandable that the establishment of a risk prediction model for CINV patients is a prior art and will not be elaborated here.

[0060] Specifically, the embodiment of the present invention extracts a number of real-time vomiting indications and their corresponding real-time vomiting characterization data of any patient within a preset period, and analyzes the correlation, so as to fully understand the intrinsic connection between the patient's vomiting-related conditions within a specific time period, and provide basic data support for the subsequent accurate construction of a risk prediction model. By analyzing the correlation, it is helpful to determine the key factors more targetedly, and by screening the real-time significant characterization variables, the variables that are of great significance to vomiting prediction are screened out, the interference of irrelevant or minor factors is reduced, and the accuracy and efficiency of the risk prediction model are improved. Only focusing on the significant characterization variables makes the model more concise and efficient, and reduces the computational complexity. By matching the real-time vomiting indications with the vomiting indications stored in the data resource library and Determine several matching characterization variables and make full use of existing data resources to enrich the variable sources of the model. Through matching, find the commonalities and differences between the current patients and the existing patient groups, and provide a reference for more accurate risk prediction. The determination of the main factors of real-time vomiting helps to identify the core factors affecting the risk of vomiting and provide a key basis for the prediction of the model. The use of main factors for prediction improves the accuracy and reliability of the prediction. By comparing the predicted vomiting indications with the real-time vomiting indications, the model can be continuously optimized so that it can be continuously improved with the addition of new data and changes in actual conditions. By correcting the model, its prediction accuracy and adaptability can be improved. The preset cycle can be adjusted to be personalized according to the conditions and needs of different patients, further improving the practicality of the model.

[0061] See also Figure 2 As shown, the step of extracting a plurality of real-time vomiting indications of any patient within a preset period and their corresponding real-time vomiting characterization data includes:

[0062] Step S110, obtaining a plurality of real-time vomiting indications by extracting a plurality of real-time vomiting intervals and their corresponding real-time vomiting durations of the patient during a plurality of chemotherapy processes;

[0063] Step S210, determining a number of real-time vomiting characterization data by extracting a number of real-time chemotherapy data and a number of real-time physiological data during a number of chemotherapy processes of the patient, wherein the number of real-time chemotherapy data include a number of real-time chemotherapy drug types and their corresponding real-time doses, and the number of real-time physiological data include a number of real-time blood pressure change values, a number of real-time heart rate change values, and a number of real-time body temperature change values ​​before and after the chemotherapy process.

[0064] It can be understood that the real-time vomiting interval in the embodiment of the present invention is the time interval from the start of chemotherapy to the occurrence of vomiting during chemotherapy, and the real-time vomiting duration is the corresponding time from the start of vomiting to the end of vomiting during chemotherapy.

[0065] Specifically, the embodiments of the present invention can accurately grasp the time pattern of vomiting of patients and judge the frequency trend of vomiting of patients by extracting the real-time vomiting intervals and the corresponding real-time vomiting durations of patients during several chemotherapy processes as real-time vomiting indicators. The real-time vomiting duration reflects the severity of vomiting. The real-time types of chemotherapy drugs and their corresponding real-time doses are extracted to clarify the degree of influence of different drugs on the vomiting reaction of patients. The blood pressure, heart rate and body temperature change values ​​in the real-time physiological data reflect the overall state of the patient's body. Chemotherapy may have an impact on the patient's physiological functions, and there may be a correlation between the changes in these physiological indicators and the vomiting reaction. For example, drastic fluctuations in blood pressure may be related to the body's stress response caused by vomiting, and changes in heart rate may indicate the patient's body's tolerance to chemotherapy. By monitoring these physiological data, a more comprehensive understanding of the patient's physical condition can be obtained, providing an accurate data basis for the establishment and prediction of subsequent models.

[0066] Specifically, the step of analyzing the correlation between a plurality of real-time vomiting indications and their corresponding real-time vomiting representation data includes:

[0067] performing univariate correlation analysis on the real-time types of chemotherapy drugs, the real-time dosages, the real-time blood pressure change values, the real-time heart rate change values, the real-time body temperature change values ​​and the real-time vomiting intervals, respectively, to obtain a plurality of first correlations according to the analysis results;

[0068] Univariate correlation analysis is performed on several of the real-time chemotherapy drug types, several of the real-time dosages, several of the real-time blood pressure change values, several of the real-time heart rate change values, and several of the real-time body temperature change values ​​with several of the real-time vomiting durations, so as to obtain several second correlations based on the analysis results.

[0069] Specifically, the embodiments of the present invention can clearly determine the degree of influence of each factor on the vomiting interval by performing univariate correlation analysis separately, which helps to determine which factors will speed up or slow down the frequency of vomiting. Similarly, a similar univariate correlation analysis is performed on the real-time vomiting duration to clarify the effect of each factor on the vomiting duration. The correlation analysis results are an important basis for subsequent steps such as determining real-time significant characterization variables, matching characterization variables, and determining real-time vomiting main factors, avoiding the inclusion of irrelevant factors in the model, thereby improving the accuracy and effectiveness of the entire CINV patient risk prediction model.

[0070] It is understandable that, in the embodiment of the present invention, if a certain type of chemotherapy drug shows a significant negative correlation with the vomiting interval, it means that the use of this drug may cause the patient to vomit more frequently; if a strong positive correlation is found between the real-time dose and the vomiting duration, it means that increasing the drug dose may prolong the patient's vomiting time.

[0071] Specifically, the step of obtaining the first correlation according to the analysis result includes:

[0072] The real-time chemotherapy drug type, the real-time dosage, the real-time blood pressure change value, the real-time heart rate change value and the real-time body temperature change value are used as a number of real-time variables;

[0073] Respectively calculating the real-time correlation coefficients corresponding to any of the real-time variables and a number of the real-time vomiting intervals and the corresponding real-time significance judgment values;

[0074] A plurality of first correlations corresponding to a plurality of the real-time variables are determined based on the real-time correlation coefficient and the real-time significance judgment value.

[0075] It can be understood that the embodiment of the present invention performs the following steps for each real-time variable:

[0076] a. Select a real-time variable (e.g., real-time chemotherapy drug type);

[0077] b. Calculate the correlation coefficient between the real-time variable and the real-time vomiting interval using an appropriate correlation analysis method (such as Pearson correlation coefficient or Spearman rank correlation coefficient, depending on the variable type);

[0078] c. Repeat steps a and b to calculate each real-time variable (real-time dose, real-time blood pressure change value, real-time heart rate change value, real-time body temperature change value) respectively.

[0079] It is understandable that, in the embodiment of the present invention, while or after calculating the correlation coefficient, the following steps are performed:

[0080] a. Conduct hypothesis testing on the correlation between each real-time variable and the real-time vomiting interval to determine the significance of the correlation;

[0081] b. Calculate the significance judgment value p corresponding to each correlation coefficient, which usually involves the following steps: i. Determine the test statistic (such as t-statistic); ii. Calculate the significance judgment value p based on the distribution of the test statistic (such as t-distribution) and the degrees of freedom.

[0082] It can be understood that the example of calculating the significant judgment value p in the embodiment of the present invention is to assume that we have the following data: real-time heart rate change value (times / minute): [5, 3, -2, 1, 4], real-time vomiting interval (hours): [2, 3, 1, 4, 2], calculate the average values ​​of the heart rate change value and the vomiting interval respectively, the average value of the heart rate change value is 2, the average value of the vomiting interval is 2.4, use the Pearson correlation coefficient to calculate the correlation between the heart rate change value and the vomiting interval, the correlation r≈0.657 (the formula of the Pearson correlation coefficient is the prior art and will not be repeated here), calculate the test statistic t based on the t-statistic formula, t≈1.635 (the formula of the t-statistic is the prior art and will not be repeated here), use the t-distribution table or statistical software, when performing a bilateral test, find the t value to be 1.635, and the p value when the degrees of freedom are 3 is approximately 0.187.

[0083] It can be understood that when the p value is less than 0.05, it is considered that the real-time variable has a significant correlation with the real-time vomiting interval.

[0084] It can be understood that the calculation process of the second correlation in the embodiment of the present invention is the same as that of the first correlation, and the same technical means are adopted, which will not be repeated here.

[0085] See also Figure 3 As shown, the step of determining several real-time significant characterization variables corresponding to the real-time vomiting indication based on the real-time correlation analysis result includes:

[0086] Step S210, respectively extracting first real-time characterization variables and second real-time characterization variables corresponding to the real-time significant judgment values ​​less than the preset significant judgment values ​​from the first correlations and the second correlations;

[0087] Step S220: selecting the intersection of a plurality of the first real-time characterization variables and a plurality of the second real-time characterization variables as the real-time significant characterization variables, and sorting them according to the real-time significant correlation coefficients corresponding to the real-time significant characterization variables.

[0088] Specifically, according to the prior art, in the embodiments of the present invention, if the p-value is less than or equal to a preset significance level (usually 0.05), the null hypothesis is rejected and the observed effect or difference is considered to be statistically significant. Therefore, the preset significance judgment value in the embodiments of the present invention is set to 0.05.

[0089] It can be understood that the real-time significant correlation coefficient in the embodiment of the present invention is the sum of the real-time correlation coefficient corresponding to the real-time significant characterization variable in calculating the first correlation and the real-time correlation coefficient corresponding to the real-time significant characterization variable in calculating the second correlation.

[0090] It can be understood that the embodiment of the present invention sorts the real-time significant characterization variables according to the real-time significant correlation coefficients corresponding to the real-time significant characterization variables by sorting the real-time significant characterization variables from large to small.

[0091] Specifically, the step of determining a number of matching characterization variables based on the matching results includes:

[0092] Matching the real-time vomiting interval and the real-time vomiting duration corresponding to any of the real-time vomiting indications with the stored vomiting interval and the stored vomiting duration corresponding to the stored vomiting indications in the data resource library;

[0093] The stored characterization data corresponding to the same stored vomiting interval and stored vomiting duration in the matching result are selected as the matching characterization variable.

[0094] It can be understood that the storage resources of the data resource library described in the embodiment of the present invention include storage of vomiting indication data and its corresponding storage representation data, which data can be obtained through clinical trials or research. The construction of the data resource library can be based on multiple data sources, such as the hospital's electronic medical record system, clinical trial database, medical literature database, etc. The construction of a data resource library based on data is a prior art and will not be repeated here.

[0095] It can be understood that the embodiment of the present invention matches the real-time vomiting interval with the stored vomiting interval, and matches the real-time vomiting duration with the stored vomiting duration. When the real-time vomiting interval and the stored vomiting interval are exactly the same, and the real-time vomiting duration and the stored vomiting duration are exactly the same, the matching characterization variable is obtained.

[0096] Specifically, the step of determining the real-time vomiting main factor based on the plurality of real-time significant characterization variables and the plurality of matching characterization variables comprises:

[0097] Comparing the plurality of real-time significant characterization variables with the plurality of matching characterization variables to obtain a comparison result;

[0098] Based on the comparison result, the variable intersection of several of the real-time significant characterization variables and several of the matching characterization variables is selected as the real-time vomiting main factor.

[0099] See also Figure 4 As shown, the step of modifying the CINV patient risk prediction model based on the comparison result between the predicted vomiting indication and the real-time vomiting indication, or adjusting the preset period includes:

[0100] Step S610, comparing the predicted vomiting indication with the real-time vomiting indication, and calculating a result difference value based on the comparison result;

[0101] Step S620, comparing the result difference value with a preset difference value range, and modifying the CINV patient risk prediction model according to the comparison result, or adjusting the preset period.

[0102] Specifically, the result difference value in the embodiment of the present invention includes a time interval difference value and a duration difference value, wherein the time interval difference value = |predicted vomiting time interval - real-time vomiting time interval|, and the duration difference value = |predicted vomiting duration - real-time vomiting duration|;

[0103] The preset difference value range includes a preset time interval difference range and a preset duration difference range, the preset time interval difference value range is [0.5 hours, 2 hours], and the preset duration difference value range is [5 minutes, 30 minutes].

[0104] It is understandable that the above range is set based on clinical experience and actual needs. Differences within this range can be considered acceptable errors, while those outside the range need to be treated accordingly. This ensures the practicality of the model, avoids frequent adjustments to the model due to overly strict requirements, and improves work efficiency.

[0105] Specifically, the embodiments of the present invention can quantify the prediction error of the model by comparing the predicted vomiting indications with the real-time vomiting indications and calculating the difference value of the results, and correct the model or adjust the preset period according to the comparison result of the difference value with the preset difference value range, thereby continuously improving the accuracy of the model, and decide whether to correct the model or adjust the preset period based on the comparison results to improve the accuracy of the model.

[0106] Specifically, the step of modifying the CINV patient risk prediction model according to the comparison result, or adjusting the preset period includes:

[0107] When the time interval difference value is less than or equal to the preset time interval difference range and the duration difference value is less than or equal to the preset duration difference range, the CINV patient risk prediction model is modified;

[0108] When the time interval difference value is greater than the preset time interval difference range or the duration difference value is greater than the preset duration difference range, the preset period is adjusted, several vomiting indications and their corresponding vomiting characterization data are recollected, and a new CINV patient risk prediction model is constructed based on the above data.

[0109] It can be understood that the embodiment of the present invention adjusts the preset period to increase the preset period to obtain more data.

[0110] Specifically, the steps of modifying the CINV patient risk prediction model include:

[0111] Matching the predicted vomiting indication with the vomiting indication stored in the data resource library, and determining the predicted characterization variable based on the matching result;

[0112] The predicted characterization variable is compared with the real-time vomiting main factor, and the corresponding parameter weights in the CINV patient risk prediction model are adjusted based on the comparison result to modify the CINV patient risk prediction model.

[0113] It can be understood that the embodiment of the present invention adjusts the corresponding parameter weights in the CINV patient risk prediction model based on the comparison results, including determining which parameter weights in the model are related to inconsistent predictive characterization variables; adjusting these parameter weights to reduce the model prediction error, wherein the adjustment process may include: increasing the weights of predictive characterization variables that are consistent with the real-time vomiting main factors; reducing or lowering the weights of predictive characterization variables that are inconsistent with the real-time vomiting main factors; if necessary, introducing new parameters or features, or removing unimportant parameters or features; using optimization algorithms (such as gradient descent, genetic algorithms, etc.) to find the optimal combination of parameter weights; and verifying the adjusted model to ensure that the adjusted model has better predictive performance on new data.

[0114] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

[0115] 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.

Claims

1. A method for constructing a risk prediction model for CINV patients, characterized in that: include: Extracting a number of real-time vomiting indications and their corresponding real-time vomiting characterization data of any patient within a preset period, analyzing the correlation between the number of real-time vomiting indications and their corresponding real-time vomiting characterization data, and obtaining a real-time correlation analysis result; Determine a number of real-time significant characterization variables corresponding to the real-time vomiting indication based on the real-time correlation analysis result; Matching the real-time vomiting indications with the vomiting indications stored in the data resource library, and determining a plurality of matching characterization variables based on the matching results; Determine a real-time vomiting main factor based on the plurality of real-time significant characterization variables and the plurality of matching characterization variables; Determining predicted vomiting indications based on the real-time vomiting main factors and the CINV patient risk prediction model; The CINV patient risk prediction model is modified based on the comparison result between the predicted vomiting indication and the real-time vomiting indication, or the preset period is adjusted.

2. The method for constructing a CINV patient risk prediction model according to claim 1, characterized in that: The step of extracting a plurality of real-time vomiting indications of any patient within a preset period and the corresponding real-time vomiting characterization data comprises: Acquire a plurality of the real-time vomiting indications by extracting a plurality of real-time vomiting intervals and their corresponding real-time vomiting durations of the patient during a plurality of chemotherapy processes; A number of real-time chemotherapy data and a number of real-time physiological data during the patient's several chemotherapy processes are extracted to determine a number of real-time vomiting characterization data, wherein the number of real-time chemotherapy data include a number of real-time chemotherapy drug types and their corresponding real-time doses, and the number of real-time physiological data include a number of real-time blood pressure change values, a number of real-time heart rate change values, and a number of real-time body temperature change values ​​before and after the chemotherapy process.

3. The method for constructing a CINV patient risk prediction model according to claim 2, characterized in that: The step of analyzing the correlation between a plurality of real-time vomiting indications and their corresponding real-time vomiting characterization data comprises: performing univariate correlation analysis on the real-time types of chemotherapy drugs, the real-time dosages, the real-time blood pressure change values, the real-time heart rate change values, the real-time body temperature change values ​​and the real-time vomiting intervals, respectively, to obtain a plurality of first correlations according to the analysis results; Univariate correlation analysis is performed on several of the real-time chemotherapy drug types, several of the real-time dosages, several of the real-time blood pressure change values, several of the real-time heart rate change values, and several of the real-time body temperature change values ​​with several of the real-time vomiting durations, so as to obtain several second correlations based on the analysis results.

4. The method for constructing a CINV patient risk prediction model according to claim 3, characterized in that: The steps of obtaining the first correlation according to the analysis results include: The real-time chemotherapy drug type, the real-time dosage, the real-time blood pressure change value, the real-time heart rate change value and the real-time body temperature change value are used as a number of real-time variables; Respectively calculating the real-time correlation coefficients corresponding to any of the real-time variables and a number of the real-time vomiting intervals and the corresponding real-time significance judgment values; A plurality of first correlations corresponding to a plurality of the real-time variables are determined based on the real-time correlation coefficient and the real-time significance judgment value.

5. The method for constructing a CINV patient risk prediction model according to claim 4, characterized in that: The step of determining a plurality of real-time significant characterization variables corresponding to the real-time vomiting indication based on the real-time correlation analysis result comprises: Respectively extracting first real-time characterization variables and second real-time characterization variables corresponding to the real-time significant judgment values ​​being less than the preset significant judgment values ​​from the first correlations and the second correlations; The intersection of a plurality of the first real-time characterization variables and a plurality of the second real-time characterization variables is selected as the real-time significant characterization variable, and is sorted according to the real-time significant correlation coefficients corresponding to the real-time significant characterization variables.

6. The method for constructing a CINV patient risk prediction model according to claim 5, characterized in that: The step of determining a plurality of matching characterization variables based on the matching results comprises: Matching the real-time vomiting interval and the real-time vomiting duration corresponding to any of the real-time vomiting indications with the stored vomiting interval and the stored vomiting duration corresponding to the stored vomiting indications in the data resource library; The stored characterization data corresponding to the same stored vomiting interval and stored vomiting duration in the matching result are selected as the matching characterization variable.

7. The method for constructing a CINV patient risk prediction model according to claim 6, characterized in that: The step of determining the real-time vomiting main factor based on the plurality of real-time significant characterization variables and the plurality of matching characterization variables comprises: Comparing the plurality of real-time significant characterization variables with the plurality of matching characterization variables to obtain a comparison result; Based on the comparison result, the variable intersection of several of the real-time significant characterization variables and several of the matching characterization variables is selected as the real-time vomiting main factor.

8. The method for constructing a CINV patient risk prediction model according to claim 7, characterized in that: The step of modifying the CINV patient risk prediction model based on the comparison result between the predicted vomiting indication and the real-time vomiting indication, or adjusting the preset period includes: comparing the predicted vomiting indication with the real-time vomiting indication, and calculating a result difference value based on the comparison result; The result difference value is compared with a preset difference value range, and the CINV patient risk prediction model is modified according to the comparison result, or the preset period is adjusted.

9. The method for constructing a CINV patient risk prediction model according to claim 8, characterized in that: The step of modifying the CINV patient risk prediction model according to the comparison result, or adjusting the preset period includes: When the time interval difference value is less than or equal to the preset time interval difference range and the duration difference value is less than or equal to the preset duration difference range, the CINV patient risk prediction model is modified; When the time interval difference value is greater than the preset time interval difference range or the duration difference value is greater than the preset duration difference range, the preset period is adjusted, several vomiting indications and their corresponding vomiting characterization data are recollected, and a new CINV patient risk prediction model is constructed based on the above data.

10. The method for constructing a CINV patient risk prediction model according to claim 9, characterized in that: The steps of modifying the CINV patient risk prediction model include: Matching the predicted vomiting indication with the vomiting indication stored in the data resource library, and determining the predicted characterization variable based on the matching result; The predicted characterization variable is compared with the real-time vomiting main factor, and the corresponding parameter weight in the CINV patient risk prediction model is adjusted based on the comparison result to modify the CINV patient risk prediction model.

Images