Path dynamic adjustment method based on medical decision

Abstract

The invention relates to the field of medical decision, in particular to a path dynamic adjustment method based on medical decision. The method comprises the following steps: acquiring a training set for training a model, and inputting the acquired training set data into the model to enable the model to construct a medical decision path data set; obtaining a verification set for verifying the model, transmitting each piece of verification medical record information to the model, generating a plurality of pieces of output decision information and a plurality of pieces of output result information by the model according to each piece of verification medical record information, performing matching, and determining whether the construction of the model for the medical decision path data set is qualified or not according to the matching degree; when the output result information is unqualified, a corresponding adjustment mode is generated based on an unqualified reason, the reason comprises matching disqualification of the output result information and model disqualification, and the adjustment mode comprises adjustment of a preset matching value and adjustment of a sample size; according to the invention, the applicability of the medical decision data path is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of medical decision-making, and in particular to a path dynamic adjustment method based on medical decision-making. BACKGROUND

[0002] With the rapid development of intelligent medical technology, the path dynamic adjustment method based on medical decision-making provides personalized and precise rehabilitation treatment plans for patients by integrating artificial intelligence and clinical diagnosis and treatment experience. In the data processing stage, the existing method integrates multi-dimensional information such as electronic medical records, diagnosis and treatment decisions, and rehabilitation effects, uses a deep learning model to construct a medical decision path data set, and dynamically optimizes the decision path according to real-time verification results, significantly improving the accuracy and effectiveness of rehabilitation treatment. However, during model training, it is usually necessary to learn based on a large amount of complete case data of rehabilitation patients, which includes training case information, training decision information, and training result information, resulting in a large scale and complex dimension of medical data. Therefore, how to realize the intelligent dynamic adjustment of the medical decision path to improve the scientificity and adaptability of the rehabilitation treatment plan is a technical problem that needs to be solved by those skilled in the art.

[0003] Chinese Patent Publication No. CN120108702A discloses a medical auxiliary decision-making system based on a multi-modal large model, which includes electronic medical records, medical images, and laboratory examination results. The collected multi-modal medical data is preprocessed to obtain preprocessed multi-modal medical data. The data processing module is also used to compare the preprocessed multi-modal medical data with the existing medical knowledge base, verify the rationality of the data and evaluate it, delete the multi-modal medical data that does not meet the requirements, and obtain the evaluated multi-modal medical data. The decision reasoning module solves the problem of limited generalization ability of multi-modal large models in the medical field, provides rich background knowledge for the model through the construction and distillation of a medical knowledge graph, improves the understanding and processing ability of the model for different tasks, and solves the problem of insufficient reliability of multi-modal large models in actual clinical applications through the verification module. It can be seen that the above technical solution has the following problems: it does not consider that the training model cannot be verified in real time in actual application scenarios, which leads to the inability to dynamically adjust the medical decision path, poor applicability of the medical decision path, and poor treatment effect. SUMMARY

[0004] Therefore, the present application provides a path dynamic adjustment method based on medical decision-making to overcome the problem that the training model cannot be verified in real time in actual application scenarios, which leads to the inability to dynamically adjust the medical decision path, poor applicability of the medical decision path, and poor treatment effect in the prior art.

[0005] To achieve the above object, the application provides a path dynamic adjustment method based on medical decision, comprising: obtaining a training set for training a model, wherein the training set comprises training medical record information, training decision information and training result information; inputting the obtained training set data into the model to make the model construct a medical decision path data set; obtaining a verification set for verifying the model, wherein the verification set comprises a plurality of corresponding verification medical record information, verification decision information and verification result information; feeding each verification medical record information into the model, and the model generates a plurality of output decision information and a plurality of output result information according to each verification medical record information, matches each verification result information with the corresponding output result information, and determines whether the model is qualified for the construction of the medical decision path data set according to the matching degree; when it is determined that the model is unqualified for the construction of the medical decision path data set, constructing a corresponding adjustment mode based on the determined reason, wherein the reason includes unqualified matching of the output result information and unqualified model, and the adjustment mode includes adjusting a preset matching value and adjusting a sample size; when it is determined that the model is qualified for the construction of the medical decision path data set, putting the model into use.

[0006] Further, the process of determining whether the model is qualified for the construction of the medical decision path data set based on the matching degree comprises: determining whether the model training is qualified based on whether the matching degree is in a preset interval; if the matching degree is in the preset interval, determining that the model training is completed and put into use; if the matching degree is not in the preset interval, determining that the model training is unqualified and the unqualified reason is that the training of the model does not meet the demand and constructing a corresponding adjustment mode according to the decision information matching degree; wherein the matching degree is the ratio of the number of the output result information matched with the corresponding verification result information to the total number of the output result information.

[0007] Further, the process of determining whether the corresponding output result information and the verification result information are matched comprises: obtaining each item of output parameter in each output result information and each item of verification parameter in each corresponding verification result information in sequence; taking the ratio of the corresponding output parameter and the verification parameter as a matching value to obtain a plurality of matching values corresponding to each parameter type; calculating the average value of each matching value to obtain the average matching value of the single output result information and the corresponding verification result information; comparing each average matching value with a preset matching value in sequence, wherein for a single average matching value; If the average matching value is greater than the preset matching value, it is determined that the output result information corresponding to the average matching value matches the corresponding verification result information.

[0008] Further, the process of determining the adjustment mode of the construction of the medical decision path data set according to the decision information matching degree comprises: obtaining each output decision parameter in each output decision information and each verification decision parameter in each corresponding verification decision information, the ratio of the corresponding output decision parameter to the verification decision parameter is recorded as the decision information matching degree, so as to obtain several decision information matching degrees corresponding to each parameter type; comparing each decision information matching degree with a preset decision information matching degree, If the decision information matching degree is greater than the preset decision information matching degree, it is determined that the reason for the unqualified matching of the output result information, and the adjustment mode is determined as adjusting the preset matching value; If the decision information matching degree is less than or equal to the preset decision information matching degree, it is determined that the reason for the unqualified model, and the adjustment mode is determined as adjusting the sample size.

[0009] Further, the adjustment range of the sample size is adjusted based on the difference between the preset matching degree and the actual matching degree, and the adjustment range of the sample size is in a positive correlation with the difference.

[0010] Further, the process of correcting the Jaccard coefficient based on the adjusted sample size comprises: The Jaccard coefficient is increased based on the sample size, and the increase range of the sample size and the Jaccard coefficient is in a negative correlation.

[0011] Further, after the adjustment is completed, the model is updated, and each verification medical record information is re-input. If the model is still unqualified for the construction of the medical decision path data set after re-verification and determination according to the verification result, the operation of adjusting the sample size is repeated, and whether the adjustment critical state is satisfied is determined after each adjustment is completed. The adjustment critical state is that the sample size reaches a critical number or the number of repeated adjustments reaches a preset value.

[0012] Further, for the model in the critical adjustment state, it is determined that the unqualified reason is that the matching of the output result information is unqualified, and the adjustment mode is determined as adjusting the preset matching value. The adjustment mode of the preset matching value is determined based on the number of types of data in the output result information. The decrease range of the preset matching value is in a positive correlation with the number of types.

[0013] Further, the process of correcting the weight coefficient of the kind of parameter in the calculation process of the corresponding average matching value based on the value ratio of the marked parameter in the verification medical record information comprises: The increase range of the weight coefficient is determined based on the value ratio of the marked parameter, and the increase range of the weight coefficient is in a positive correlation with the value ratio; The marked parameter refers to the kind of parameter corresponding to the maximum value greater than the regular interval or the minimum value less than the regular interval in the medical record information, and the value ratio is the ratio of the absolute value of the difference to the preset value.

[0014] Further, in the case that the adjustment of the preset matching value is completed and the construction of the medical decision path data set by the model is still unqualified, it is determined that the reason is that the training model used does not meet the use requirements, and a model replacement notice is directly issued.

[0015] Compared with the prior art, the method has the beneficial effects that the training set of the training model is constructed by obtaining the training medical record information, the training decision information and the training result information, and the medical decision path data set is constructed by using the training data set, the medical decision path set can be constructed based on multiple information to make the construction of the medical decision path set more accurate; meanwhile, the matching between each verification result information and the corresponding output result information is performed, and whether the model construction is qualified is determined according to the matching degree, and the corresponding processing mode is generated based on the unqualified reason when it is unqualified, wherein the processing mode comprises adjusting the preset matching value or adjusting the sample size, so as to realize the real-time verification of the training model, and further improve the applicability of the medical decision data path.

[0016] Further, the present application can quickly determine whether the model training is qualified based on whether the matching value is in the preset interval, so as to improve the efficiency of determining whether the model training is qualified, and further improve the efficiency of self-adaptive adjustment of the model when it is unqualified.

[0017] Further, the present application determines the matching between the output result information and the corresponding verification result information by comparing the average matching value with the preset matching value, so that the rationality and accuracy of the process of determining whether the matching is stronger, thereby further improving the applicability of the medical decision data path.

[0018] Further, the present application determines the adjustment mode of the construction of the medical decision path data set based on the decision information matching degree, and determines the two kinds of adjustment modes based on the comparison result of the decision information matching degree and the preset decision information matching degree, so that the determination of the adjustment mode is more targeted, thereby further improving the applicability of the medical decision data path.

[0019] Further, the application effectively guarantees the accuracy of the adjustment range of the sample size by adjusting the adjustment range of the sample size based on the difference between the preset matching ratio and the actual matching ratio, further realizes the accurate analysis of the path of the medical decision-making, and further improves the applicability of the medical decision-making data path.

[0020] Further, the application further improves the collection standard of the related parameters in the training case information by correcting the Jaccard coefficient based on the adjusted sample size, and further improves the applicability of the medical decision-making data path.

[0021] Further, the application updates the model based on the adjustment, re-verify, and determines whether the model is still unqualified for the construction of the medical decision-making path data set according to the verification result, repeatedly executes the operation of adjusting the sample size, and determines whether the adjustment critical state is met after each adjustment; by setting the adjustment critical state, the unreasonable adjustment caused by excessive adjustment is avoided, the rationality of the adjustment is improved, and the applicability of the medical decision-making data path is further improved.

[0022] Further, the application determines the unqualified reason of the model in the critical adjustment state and correspondingly determines the adjustment preset matching value, and determines the downward adjustment range of the preset matching value according to the number of data categories in the output result information, effectively guarantees the accuracy of the preset matching value, and further improves the applicability of the medical decision-making data path.

[0023] Further, the application corrects the weight coefficient of the parameter of this kind in the calculation process of the corresponding average matching value based on the value ratio of the marked parameter in the verification medical record information, effectively enhances the rationality of the weight coefficient, and further improves the applicability of the medical decision-making data path.

[0024] Further, the application directly issues a model replacement notice based on the result that the model is unqualified after multiple adjustments, avoids the waste of resources caused by invalid adjustment, and further improves the applicability of the medical decision-making data path. BRIEF DESCRIPTION OF DRAWINGS

[0025] Figure 1 is a unit connection diagram of the path dynamic adjustment system based on medical decision-making of the application; Figure 2 is a step flowchart of the path dynamic adjustment method based on medical decision-making of the application; Figure 3 is a flowchart of the application based on the matching degree determination model for determining whether the construction of the medical decision-making path data set is qualified; Figure 4A flowchart of an adjustment mode of the present application for constructing a medical decision path data set according to a matching degree of decision information. DETAILED DESCRIPTION

[0026] The preferred embodiments of the present application will be described below with reference to the accompanying drawings. Those skilled in the art will understand that the embodiments are only used to explain the technical principles of the present application and are not intended to limit the protection scope of the present application.

[0027] It should be noted that, in the description of the present application, the terms indicating the direction or positional relationship such as "upper", "lower", "left", "right", "inner", "outer" and the like are based on the direction or positional relationship shown in the drawings, which is only for the convenience of description and does not indicate or imply that the device or element must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation on the present application.

[0028] In addition, it should also be noted that, in the description of the present application, unless otherwise explicitly specified and limited, the terms "mounting", "connecting", "connection" should be understood in a broad sense, for example, can be fixedly connected, can be detachably connected, or integrally connected; can be mechanically connected, can be electrically connected; can be directly connected, can be indirectly connected through an intermediate medium, or can be the communication inside two elements. Those skilled in the art can understand the specific meaning of the above terms in the present application according to the specific circumstances.

[0029] Please refer to Figure 1 As shown in the figure, it is a structure schematic diagram of the route planning system based on complex airspace conditions according to the embodiment of the present application.

[0030] The system comprises an acquisition unit, a training unit, a verification unit, a matching unit and an adjustment unit.

[0031] The acquisition unit is configured to acquire a training set, wherein the training set comprises training medical record information, training decision information and training result information; The training unit is connected with the acquisition unit and configured to receive the training set, and configured to deliver the training set to a model to make the model construct a medical decision path data set; The verification unit is connected with the training unit and configured to acquire a verification set for verifying the model, wherein the verification set comprises a plurality of corresponding verification medical record information, verification decision information and verification result information; the verification unit is further configured to deliver each of the verification medical record information to the model to make the model generate a plurality of output decision information and a plurality of output result information according to each of the verification medical record information; The matching unit is connected with the verification unit, which is used to match each verification result information and corresponding output result information, and determine whether the construction of the medical decision path data set is qualified according to the matching degree determination model; The adjustment unit is connected with the matching unit, which is used to construct a corresponding adjustment mode based on the determined reason when it is determined that the construction of the medical decision path data set is unqualified, wherein the reason includes the unqualified matching of the output result information and the unqualified model, the adjustment mode includes adjusting the preset matching value and adjusting the sample size, and the model is put into use when it is determined that the construction of the medical decision path data set is qualified.

[0032] Please refer to Figure 2 The figure is a step flow chart of the medical decision path dynamic adjustment method based on the embodiment of the application.

[0033] The steps of the system in the embodiment of the application in the actual operation process include: S1, acquiring a training set for training the model based on the acquiring unit, wherein the training set includes training medical record information, training decision information and training result information; S2, inputting the acquired training set into the model by the training unit to make the model construct a medical decision path data set; S3, acquiring a verification set for verifying the model by the verification unit, wherein the verification set includes a plurality of corresponding verification medical record information, verification decision information and verification result information; and inputting each verification medical record information into the model to make the model generate a plurality of output decision information and a plurality of output result information according to each verification medical record information; S4, the matching unit is used to match each verification result information and corresponding output result information, and determine whether the construction of the medical decision path data set is qualified according to the matching degree determination model; S5, the adjustment unit is used to construct a corresponding adjustment mode based on the determined reason when it is determined that the construction of the medical decision path data set is unqualified, wherein the reason includes the unqualified matching of the output result information and the unqualified model, the adjustment mode includes adjusting the preset matching value and adjusting the sample size, and the model is put into use when it is determined that the construction of the medical decision path data set is qualified.

[0034] The process of determining whether the construction of the medical decision path data set is qualified based on the matching degree determination model in the embodiment of the application includes: determining whether the model training is qualified based on whether the matching degree is in a preset interval; if the matching degree is in the preset interval, determining that the model training is completed and put into use; if the matching degree is not in the preset interval, determining that the model training is unqualified and the unqualified reason is that the training of the model does not meet the demand and the adjustment mode determined according to the decision information matching degree is adjusting the preset matching value or adjusting the sample size; wherein the matching degree is the ratio of the number of output result information matched with the corresponding verification result information to the total number of output result information.

[0035] Specifically, in the embodiment, the matching degree P can be divided into the maximum matching degree P1 and the minimum matching degree P2, and the maximum matching degree P1=1 and the minimum matching degree P2=0.85 are set in the matching value standard. It should be noted that in other embodiments, the values of P1 and P2 can also be determined according to the construction demand of the medical decision path data set. The comparison process based on the matching degree P and P1 and P2 is specifically as follows: If the matching degree P is greater than the minimum matching degree P2 and less than or equal to the maximum matching degree P1, it is determined that the model training is completed and put into use. If the matching degree P is less than or equal to the minimum matching degree P2 or the matching degree P is greater than the maximum matching degree P1, it is determined that the model training is unqualified and the unqualified reason is that the training of the model does not meet the demand and the adjustment mode determined according to the decision information matching degree Q is adjusting the preset matching value or adjusting the sample size.

[0036] Specifically, the process of determining whether the corresponding output result information and the verification result information match in the embodiment of the application includes: sequentially obtaining each output parameter in each output result information and each verification parameter in each corresponding verification result information; the ratio of the corresponding output parameter and the verification parameter is recorded as a matching value, so as to obtain a plurality of matching values corresponding to each parameter type; the average value of each matching value is calculated to obtain the average matching value of a single output result information and the corresponding verification result information; each average matching value is compared with a preset matching value in sequence, wherein for a single average matching value; if the average matching value is greater than the preset matching value, it is determined that the output result information corresponding to the average matching value and the corresponding verification result information match.

[0037] Specifically, in the embodiment, the preset matching value L0=0.922, and the comparison process based on the average matching value L and the preset matching value L0 is specifically as follows: If the average matching value L is greater than the preset matching value L0, the output result information corresponding to the average matching value and the corresponding verification result information match. If the average matching value L is less than or equal to the preset matching value L0, it indicates that the matching degree of the output result information and the corresponding verification result information is low, and the model training is determined to be unqualified and the unqualified reason is that the training of the model does not meet the requirements.

[0038] Specifically, the process of determining the adjustment mode of the construction of the medical decision path data set according to the decision information matching degree includes: obtaining each output decision parameter in each output decision information and each verification decision parameter in each corresponding verification decision information, taking the ratio of the corresponding output decision parameter and the verification decision parameter as the decision information matching degree, to obtain several decision information matching degrees corresponding to each parameter type; comparing each decision information matching degree with a preset decision information matching degree, if the decision information matching degree is greater than the preset decision information matching degree, it is determined that the unqualified reason is the matching of the output result information, and the adjustment mode is determined as adjusting the preset matching value; if the decision information matching degree is less than or equal to the preset decision information matching degree, it is determined that the unqualified reason is the model, and the adjustment mode is determined as adjusting the sample size.

[0039] Specifically, in this embodiment, the preset decision information matching degree R0=0.91, and the comparison process between the decision information matching degree and the preset decision information matching degree is specifically as follows: If the decision information matching degree R is greater than the preset decision information matching degree R0, it indicates that the decision information matching degree is high, and it is determined that the unqualified reason is the matching of the output result information, and the adjustment mode is determined as adjusting the preset matching value. If the decision information matching degree R is less than or equal to the preset decision information matching degree R0, it indicates that the sample size of the model is not enough to cause the training of the model to not meet the requirements, and it is determined that the unqualified reason is the model, and the adjustment mode is determined as adjusting the sample size.

[0040] Specifically, the embodiment of the present application adjusts the adjustment range of the sample size based on the difference between the preset matching degree and the actual matching degree, and the adjustment range of the sample size and the difference are in a positive correlation.

[0041] Specifically, in this embodiment, the difference P21 between the preset matching degree P11 and the actual matching value P is 0.2 times the original sample size N, wherein the preset matching degree is the middle value of the preset interval.

[0042] Specifically, the process of correcting the Jaccard coefficient based on the adjusted sample size in the embodiment of the present application includes: increasing the Jaccard coefficient based on the sample size, and the increase range of the sample size and the Jaccard coefficient is in a positive correlation.

[0043] Specifically, in the embodiment, whether a single keyword is a keyword used by a patient to characterize a disease is determined by a Jaccard coefficient, which is the co-occurrence ratio of the keyword and the field patents and literatures, and the larger the Jaccard coefficient is, the stronger the field relevance is.

[0044] Specifically, after the adjustment is completed, the model is updated, and each of the verification medical record information is re-input. If the re-verification and the verification result determine that the model is still unqualified for the construction of the medical decision path data set, the operation of adjusting the sample size is repeatedly performed, and after each adjustment is completed, it is determined whether the adjustment critical state is satisfied. The adjustment critical state is that the sample size reaches a critical number or the number of repeated adjustments reaches a preset value.

[0045] Specifically, in the embodiment of the application, the sample size Ni after the i-th adjustment is greater than or equal to a preset critical number N0, the preset critical number N0 is 1.8 times the sample size, or the number of repeated adjustments i is equal to a preset number of repeated adjustments J, and the preset number of repeated adjustments J = 4 times, then it is determined that the adjustment critical state is reached.

[0046] Specifically, for the model in the critical adjustment state, the application embodiment determines that the unqualified reason is that the matching of the output result information is unqualified, and determines that the adjustment mode is adjusting the preset matching value. The adjustment mode of determining the preset matching degree based on the number of categories of data in the output result information; The decrease range of the preset matching degree is in a positive correlation with the number of categories.

[0047] Specifically, in the embodiment of the application, the preset number of categories n0 = 3, and the comparison process based on the number of categories n and the preset number of categories n0 is specifically as follows: If the number of categories n is less than or equal to the preset number of categories n0, the original preset matching degree is not adjusted; If the number of categories n is greater than the preset number of categories n0, the preset matching degree is adjusted to 0.92 times the original preset matching degree; The initial interval range is 0.85-1 and the corresponding actual range is 0.15, the range after interval expansion is 0.1575, at this time 0.07875 is divided into two parts, and the divided range is 0.77125-0.92125.

[0048] Specifically, in the embodiment of the application, the process of correcting the weight coefficient of the category parameter in the calculation process of the corresponding average matching value based on the value ratio of the label parameter in the verification medical record information includes: The increase range of the weight coefficient is determined based on the value ratio of the label parameter, and the increase range of the weight coefficient is in a positive correlation with the value ratio; The marking parameter refers to a parameter type corresponding to a maximum value greater than a regular interval or a minimum value less than the regular interval in medical record information, and the value ratio is a ratio of an absolute value of a difference value to a preset value.

[0049] Specifically, in the case that the preset matching value is adjusted and the construction of the medical decision path data set for the model is still unqualified, it is determined that the training model used does not meet the use requirements, and a model replacement notification is directly issued.

[0050] In order to make the purpose and advantages of the present application more clear and obvious, the present application will be further described below in combination with examples; it should be understood that the specific examples described herein are only used to explain the present application, and do not limit the present application.

[0051] Example 1 In this embodiment, the model is used for the construction of a medical decision path for a diabetes disorder. For the diabetes disorder, the parameters used to represent the disorder and the numerical threshold values of each parameter include: Parameter A fasting blood glucose: 3.9-6.1 mmol / L; Parameter B HbA1c: 4%-6%; Parameter C fasting insulin: 5-20 mIU / L; 1684 cases of patient medical record information after recovery were obtained and a training set was constructed based on the above information. At this time, the training set includes a corresponding number of medical record information, treatment decision information for each medical record information, and physical examination information after treatment containing each parameter; the medical record information is recorded as training medical record information, each treatment decision information is recorded as training decision information, and each physical examination information is recorded as training result information; the training set is sent to the model to make the model construct a medical decision path data set for the diabetes disorder; The medical records information of 308 patients who have recovered is obtained, and a verification set is constructed based on the above information. At this time, the verification set includes a corresponding amount of medical record information, treatment decision information for each medical record information, and physical examination information after treatment containing each parameter. The medical record information is recorded as verification medical record information, each treatment decision information is recorded as verification decision information, and each physical examination information is recorded as verification result information. Each verification medical record information is input into the model, and the model generates a number of output decision information and a number of output result information. Each verification result information and the corresponding output result information are matched. For a single medical record, the average matching value corresponding thereto is the average value of the ratio of each output parameter to the corresponding verification parameter. According to the matching result, the number of matched output result information screened out is 247. At this time, the matching degree = the number of matched output result information / the total number of output result information. The matching degree P obtained in this embodiment is 0.8. At this time, it is determined that the model training is unqualified and the unqualified reason is that the training of the model does not meet the requirements and the adjustment mode is determined according to the decision information matching degree Q. The number of output decision information generated by the model is matched with the verification decision information. In this embodiment, for a single medical record in the verification set, the medical record with the same output decision information and verification decision information is recorded as a matched medical record, and the decision information matching degree is determined according to the ratio of the number of matched medical records to the total number of medical records. In this embodiment, the number of medical records with the same output decision information and verification decision information is 285, and the decision information matching degree R obtained is 0.925. At this time, R belongs to the case of being greater than R0=0.91. At this time, the reason for the unqualified matching of the output result information is determined, and the adjustment mode is to adjust the preset matching value. In this embodiment, the number of categories is 4, the preset interval adjustment is 0.77125-0.92125, and the preset matching value is adjusted. After the re-detection, the matching degree is within the range corresponding to the interval adjustment. At this time, the model training is qualified.

[0052] Embodiment 2 In this embodiment, the model is used for the construction of the medical decision path for the hypertension disease. For the hypertension disease, the parameters used to represent the disease and the numerical threshold of each parameter include: Parameter 1 resting blood pressure: 80-120 mmHg; Parameter 2 arterial stiffness: 5-10 m / s; Parameter 3 plasma norepinephrine: 200-400 pg / mL; Parameter i left ventricular ejection fraction: 55%-70%.

[0053] Get 2014 cases of patients who have recovered medical information and build a training set based on the above information, at this time, the embodiment of the training medical information, training decision information and training results information is the same as the embodiment 1, not described here; the training set is delivered to the model to make the model build a medical decision path data set for hypertension; Get 460 cases of patients who have recovered medical information, and build a validation set based on the above information, the way of building validation medical information, validation decision information and validation result information in this embodiment is the same as the embodiment 1, not described here; the validation medical information is delivered to the model, the model generates several output decision information and several output result information, each validation result information and the corresponding each output result information are matched, for a single medical record, the corresponding average matching value is the average value of each output parameter and the corresponding validation parameter, according to the above matching result, the number of matched output result information screened out is 414 cases, at this time, the matching degree = the number of matched output result information / the total number of output result information, the matching degree P obtained in this embodiment is 0.9, at this time, it is determined that the model training is completed and put into use.

[0054] So far, the technical scheme of the present application has been described in combination with the preferred embodiments shown in the drawings, but those skilled in the art can easily understand that the protection scope of the present application is obviously not limited to these specific embodiments. Those skilled in the art can make equivalent changes or replacements to related technical features without departing from the principles of the present application, and the technical scheme after the changes or replacements will fall within the protection scope of the present application.

[0055] The above is only the preferred embodiment of the present application, and is not used to limit the present application; for those skilled in the art, the present application can have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for dynamic adjustment of pathways based on medical decision-making, characterized in that, include: Obtain a training set for training the model, wherein the training set includes training medical record information, training decision information, and training result information; The acquired training set data is input into the model to enable the model to construct a medical decision path dataset; Obtain a validation set for validating the model, wherein the validation set includes several corresponding validation medical record information, validation decision information, and validation result information; Each of the aforementioned verification medical record information is fed into the model. The model generates several output decision information and several output result information based on each verification medical record information. Each verification result information and its corresponding output result information are matched. The model is then used to determine whether its construction of the medical decision path dataset is qualified based on the matching degree. When the model is deemed unqualified for constructing a medical decision-making path dataset, a corresponding adjustment method is constructed based on the determined reasons. The reasons include unqualified matching of output result information and unqualified model. The adjustment methods include adjusting the preset matching value and adjusting the sample size. The model will be put into use once it is determined that the model is adequately constructed for the medical decision-making path dataset.

2. The method for dynamic path adjustment based on medical decision-making according to claim 1, characterized in that, The process of determining whether the medical decision-making path dataset is qualified based on the matching degree judgment model includes: The success of model training is determined based on whether the matching degree is within a preset range. If the matching degree is within the preset range, the model is considered to have completed training and can be put into use. If the matching degree is not within the preset range, the model training is deemed unqualified and the reason for the unqualification is that the training of the model does not meet the requirements. The corresponding adjustment method is constructed based on the matching degree of the decision information. The matching degree is the ratio of the number of output results that match the corresponding verification result information to the total number of output results.

3. The method for dynamic path adjustment based on medical decision-making according to claim 2, characterized in that, The process of determining whether the corresponding output result information matches the verification result information includes: Sequentially obtain each output parameter from each output result information and each verification parameter from each corresponding verification result information; The ratio of the corresponding output parameter to the verification parameter is recorded as the matching value, so as to obtain several matching values ​​corresponding to each parameter type. Calculate the average of each matching value to obtain the average matching value between a single output result and the corresponding verification result. Each of the aforementioned average matching values ​​is compared sequentially with a preset matching value, wherein, for a single average matching value; If the average matching value is greater than the preset matching value, then the output result information and the corresponding verification result information corresponding to the average matching value are determined to match.

4. The method for dynamic path adjustment based on medical decision-making according to claim 3, characterized in that, The process of determining the adjustment method for constructing the medical decision-making path dataset based on the matching degree of decision information includes: Obtain the output decision parameters from each of the output decision information and the verification decision parameters from the corresponding verification decision information. The ratio of the corresponding output decision parameters to the verification decision parameters is recorded as the decision information matching degree, so as to obtain several decision information matching degrees corresponding to each parameter type. The matching degree of each decision information is compared with the preset matching degree of decision information. If the matching degree of the decision information is greater than the preset matching degree of the decision information, the reason is determined to be that the matching of the output result information is unqualified, and the adjustment method is determined to be adjusting the preset matching value; If the matching degree of decision information is less than or equal to the preset matching degree of decision information, the reason is determined to be that the model is unqualified, and the adjustment method is determined to be adjusting the sample size.

5. The path dynamic adjustment method based on medical decision-making according to claim 4, characterized in that, The adjustment range of the sample size is adjusted based on the difference between the preset matching degree and the actual matching degree, and the adjustment range of the sample size is positively correlated with the difference.

6. The method for dynamic path adjustment based on medical decision-making according to claim 5, characterized in that, The process of correcting the Jaccard coefficient based on the adjusted sample size includes: The Jaccard coefficient is increased based on the sample size, and the increase in sample size is negatively correlated with the increase in the Jaccard coefficient.

7. The method for dynamic path adjustment based on medical decision-making according to claim 5, characterized in that, After adjustment, update the model and re-enter the verification medical record information. If the model is still not qualified for the medical decision path dataset after re-verification and the verification results indicate that the model is not qualified, repeat the operation of adjusting the sample size and determine whether the adjustment critical state is met after each adjustment. The adjustment critical state is when the sample size reaches a critical number or the number of repeated adjustments reaches a preset value.

8. The method for dynamic path adjustment based on medical decision-making according to claim 7, characterized in that, For the model that is in a critical adjustment state, the reason for the failure is determined to be that the matching of the output result information is unqualified, and the adjustment method is determined to be to adjust the preset matching value. The adjustment method for the preset matching value is determined based on the number of data types in the output result information; The decrease in the preset matching value is positively correlated with the number of categories.

9. The method for dynamic path adjustment based on medical decision-making according to claim 7, characterized in that, The process of correcting the weighting coefficient of the corresponding type of parameter in the calculation of the average matching value based on the numerical ratio of the marked parameters in the verified medical record information includes: The increase in the weight coefficient is determined based on the numerical ratio of the marked parameters, and the increase in the weight coefficient is positively correlated with the numerical ratio. The marking parameters refer to the types of parameters in the medical record information whose values ​​are greater than the maximum value of the normal range or less than the minimum value of the normal range. The value ratio is the ratio of the absolute value of the difference to the preset value.

10. The method for dynamic path adjustment based on medical decision-making according to claim 9, characterized in that, If the model fails to construct a suitable dataset for the medical decision-making path even after adjusting the preset matching values, the reason is determined to be that the training model used does not meet the usage requirements, and a model replacement notification is issued directly.

Images