CN111145905B - Target decision model construction method and device, electronic equipment and storage medium - Google Patents
Target decision model construction method and device, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN111145905B CN111145905B CN201911416966.2A CN201911416966A CN111145905B CN 111145905 B CN111145905 B CN 111145905B CN 201911416966 A CN201911416966 A CN 201911416966A CN 111145905 B CN111145905 B CN 111145905B
- Authority
- CN
- China
- Prior art keywords
- decision
- data
- model
- decision model
- target
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000010276 construction Methods 0.000 title claims abstract description 32
- 238000012549 training Methods 0.000 claims abstract description 70
- 238000000034 method Methods 0.000 claims abstract description 38
- 238000011269 treatment regimen Methods 0.000 claims description 22
- 230000000153 supplemental effect Effects 0.000 claims description 12
- 238000012545 processing Methods 0.000 claims description 9
- 238000003066 decision tree Methods 0.000 claims description 8
- 230000002787 reinforcement Effects 0.000 claims description 6
- 238000013136 deep learning model Methods 0.000 claims description 4
- 238000007477 logistic regression Methods 0.000 claims description 4
- 238000007637 random forest analysis Methods 0.000 claims description 4
- 239000013589 supplement Substances 0.000 claims description 3
- 238000004590 computer program Methods 0.000 claims description 2
- 238000011282 treatment Methods 0.000 description 25
- 238000003745 diagnosis Methods 0.000 description 19
- 238000010586 diagram Methods 0.000 description 15
- 238000002512 chemotherapy Methods 0.000 description 10
- 201000010099 disease Diseases 0.000 description 9
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 9
- 208000024891 symptom Diseases 0.000 description 5
- 230000006870 function Effects 0.000 description 3
- 238000010801 machine learning Methods 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 206010061218 Inflammation Diseases 0.000 description 2
- 206010026673 Malignant Pleural Effusion Diseases 0.000 description 2
- 230000006978 adaptation Effects 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 2
- 230000000295 complement effect Effects 0.000 description 2
- 230000001419 dependent effect Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 210000003238 esophagus Anatomy 0.000 description 2
- 230000004054 inflammatory process Effects 0.000 description 2
- 210000001370 mediastinum Anatomy 0.000 description 2
- 239000013307 optical fiber Substances 0.000 description 2
- 230000036314 physical performance Effects 0.000 description 2
- 230000000644 propagated effect Effects 0.000 description 2
- 210000003437 trachea Anatomy 0.000 description 2
- 206010028980 Neoplasm Diseases 0.000 description 1
- 206010037660 Pyrexia Diseases 0.000 description 1
- 206010064390 Tumour invasion Diseases 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000009400 cancer invasion Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000013499 data model Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 206010013781 dry mouth Diseases 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 229940125753 fibrate Drugs 0.000 description 1
- 230000009545 invasion Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 201000009240 nasopharyngitis Diseases 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
- 238000005728 strengthening Methods 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/70—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for mining of medical data, e.g. analysing previous cases of other patients
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H70/00—ICT specially adapted for the handling or processing of medical references
- G16H70/20—ICT specially adapted for the handling or processing of medical references relating to practices or guidelines
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Public Health (AREA)
- General Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Epidemiology (AREA)
- Biomedical Technology (AREA)
- Primary Health Care (AREA)
- Databases & Information Systems (AREA)
- Pathology (AREA)
- Bioethics (AREA)
- Medical Treatment And Welfare Office Work (AREA)
- Measuring And Recording Apparatus For Diagnosis (AREA)
Abstract
The invention provides a target decision model construction method and device, electronic equipment and storage medium, and relates to the technical field of computers. The method comprises the following steps: acquiring clinical guideline data and constructing a first decision model according to the clinical guideline data; wherein the first decision model comprises a plurality of decision branch paths; acquiring feature data corresponding to a target task and tag data corresponding to the feature data; training a second pre-constructed decision model through the feature data and the label data, and constructing a target decision model according to the trained second decision model and the first decision model. The technical scheme of the embodiment of the invention can improve the interpretation and accuracy of the decision result of the decision model and improve the working efficiency of doctors.
Description
Technical Field
The present invention relates to the field of computer technology, and in particular, to a target decision model construction method, a target decision model construction apparatus, an electronic device, and a computer readable storage medium.
Background
With the rapid development of internet technology, the medical field is increasingly dependent on clinical decisions made by clinical auxiliary decision systems (Clinical Decision Support System, CDSS). The goal of the clinical decision-making aid system is to aid the physician through a decision model in the system.
At present, a decision model in a clinical auxiliary decision system of a related technical scheme is a knowledge-driven model, the interpretation of a corresponding decision result is strong, but the accuracy of a recommended result is low; or a data-based model, the corresponding decision result has high accuracy, but the interpretation of the recommended result is weaker.
It should be noted that the information disclosed in the above background section is only for enhancing understanding of the background of the present disclosure and thus may include information that does not constitute prior art known to those of ordinary skill in the art.
Disclosure of Invention
The embodiment of the invention aims to provide a target decision model construction method, a target decision model construction device, electronic equipment and a computer readable storage medium, so that the problem that the accuracy and the interpretability of a decision result corresponding to a decision model in a related scheme cannot be achieved at least to a certain extent is solved.
Other features and advantages of the invention will be apparent from the following detailed description, or may be learned by the practice of the invention.
According to a first aspect of an embodiment of the present invention, there is provided a target decision model construction method, including: acquiring clinical guideline data and constructing a first decision model according to the clinical guideline data; wherein the first decision model comprises a plurality of decision branch paths; acquiring feature data corresponding to a target task and tag data corresponding to the feature data; training a second pre-constructed decision model through the feature data and the tag data, and constructing a target decision model according to the second decision model and the first decision model after training.
In some example embodiments of the present invention, based on the foregoing solution, training the pre-constructed second decision model by the feature data and the tag data includes: inputting the characteristic data into the first decision model to determine a target decision branch path corresponding to the characteristic data, and determining initial decision data corresponding to the characteristic data according to the target decision branch path; training a second pre-constructed decision model according to the initial decision data and the label data.
In some example embodiments of the present invention, based on the foregoing, the second decision model comprises a subdivision decision model, and the training of the pre-constructed second decision model by the initial decision data and the tag data comprises: when the initial decision data is determined to correspond to a plurality of decision results, acquiring a pre-constructed subdivision decision model; training the subdivision decision model according to a plurality of decision results and the tag data so that the subdivision decision model classifies the initial decision data.
In some example embodiments of the present invention, based on the foregoing, the second decision model includes an enhanced decision model, the training of the pre-constructed second decision model by the initial decision data and the tag data includes: when the initial decision data is determined to be the decision data to be enhanced, acquiring a pre-constructed enhanced decision model; training the reinforced decision model according to the characteristic data and the tag data so that the reinforced decision model determines a decision result corresponding to the decision data to be reinforced; wherein the decision result comprises a plurality of decision classifications.
In some example embodiments of the present invention, based on the foregoing, the training the enhanced decision model according to the feature data and the tag data includes: marking the label of the characteristic data as a target label in advance, and judging whether leaf node data of the characteristic data in the first decision model is consistent with label data corresponding to the characteristic data; wherein the leaf node data comprises a first treatment regimen and the tag data comprises a second treatment regimen; upon determining that the first treatment regimen is consistent with the second treatment regimen, taking the target label as reinforcement label data; training the reinforced decision model according to the characteristic data, the tag data and the reinforced tag data so that the reinforced decision model carries out reinforced decision on the data to be reinforced and determines the decision classification corresponding to the data to be reinforced.
In some example embodiments of the present invention, based on the foregoing, the second decision model includes a supplemental decision model, the training of the pre-constructed second decision model by the initial decision data and the tag data includes: when the initial decision data corresponding to the feature data is determined to be the feature data according to the target decision branch path, determining that the target decision branch path is in an empty state and acquiring a pre-constructed supplementary decision model; training the supplemental decision model according to the feature data and the tag data to enable the supplemental decision model to classify the feature data.
In some example embodiments of the invention, based on the foregoing, the first decision model comprises a rule decision tree model and the second decision model comprises any one of a decision tree model, a logistic regression model, a random forest model, a bayesian model, a deep learning model.
According to a second aspect of embodiments of the present invention, there is provided a clinical aid decision making method comprising: acquiring input decision influencing factors, and determining initial decision data corresponding to the decision influencing factors according to a first decision model in a pre-trained target decision model; and determining auxiliary decision data corresponding to the initial decision data according to a second decision model in the target decision model so as to assist a target object to determine a decision result corresponding to the decision influence factor according to the auxiliary decision data.
In some example embodiments of the present invention, based on the foregoing solution, the second decision model includes an enhanced decision model, and determining auxiliary decision data corresponding to the initial decision data according to a second decision model in the target decision model includes: when the initial decision data is determined to be the decision data to be enhanced, inputting the decision data to be enhanced into the enhanced decision model to obtain the decision classification of the decision data to be enhanced; and taking the decision classification as auxiliary decision data corresponding to the initial decision data.
According to a third aspect of the embodiment of the present invention, there is provided a target decision model construction apparatus, including: the first decision model construction module is used for acquiring clinical guideline data and constructing a first decision model according to the clinical guideline data; wherein the first decision model comprises a plurality of decision branch paths; the sample data acquisition module is used for acquiring the characteristic data and the label data corresponding to the characteristic data; the target decision model building module is used for training the pre-built second decision model through the characteristic data and the tag data, and building a target decision model according to the second decision model and the first decision model which are completed by training.
According to a fourth aspect of embodiments of the present invention, there is provided a clinical aid decision making apparatus comprising: the decision influencing factor acquisition unit is used for acquiring the input decision influencing factors; the initial decision data determining unit is used for determining initial decision data corresponding to the decision influencing factors according to a first decision model in the pre-trained target decision model; and the auxiliary decision data determining unit is used for determining auxiliary decision data corresponding to the initial decision data according to a second decision model in the target decision model so as to assist a target object to determine a decision result corresponding to the decision influencing factor according to the auxiliary decision data.
According to a fifth aspect of an embodiment of the present invention, there is provided an electronic apparatus including: a processor; and a memory having stored thereon computer readable instructions which when executed by the processor implement the target decision model construction method or the clinical assist decision method of any of the above.
According to a sixth aspect of embodiments of the present invention, there is provided a computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements a target decision model construction method or a clinical auxiliary decision method according to any of the above.
The technical scheme provided by the embodiment of the invention can have the following beneficial effects:
according to the target decision model construction method in the example embodiment of the invention, a first decision model is constructed according to clinical guideline data, feature data is input into the first decision model to determine a target decision branch path, initial decision data corresponding to the feature data is determined according to the target decision branch path, a second pre-constructed decision model is trained through the initial decision data and tag data, and a target decision model is constructed according to the trained second decision model and the first decision model. On the one hand, initial decision data are determined according to the first decision model, a second decision model is trained according to the initial decision data and the label data, and finally a target decision model is built according to the first decision model and the second decision model, so that the interpretability of a decision result obtained by the target decision model can be improved, and more reference data are provided for a target object (such as a doctor); on the other hand, initial decision data is determined by carrying out preliminary decision on the feature data according to the first decision model, and a second decision model is trained according to the initial decision data and the label data so that the second decision model can further classify the initial decision data and determine a decision result, so that the accuracy of the decision result obtained by the target decision model can be improved.
It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the invention as claimed.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and together with the description, serve to explain the principles of the invention. It is evident that the drawings in the following description are only some embodiments of the present invention and that other drawings may be obtained from these drawings without inventive effort for a person of ordinary skill in the art. In the drawings:
FIG. 1 schematically illustrates a flow diagram of a method of target decision model construction according to some embodiments of the invention;
FIG. 2 schematically illustrates a flow diagram of training a subdivision decision model, in accordance with some embodiments of the invention;
FIG. 3 schematically illustrates a flow diagram of training an enhanced decision model according to some embodiments of the invention;
FIG. 4 schematically illustrates a flow diagram of training a supplemental decision model according to some embodiments of the invention;
FIG. 5 schematically illustrates a flow diagram of a clinical aid decision making method according to some embodiments of the invention;
FIG. 6 schematically illustrates a schematic diagram of a target decision model structure according to some embodiments of the invention;
FIG. 7 schematically illustrates a schematic diagram of a target decision model building apparatus according to some embodiments of the invention;
FIG. 8 schematically illustrates a schematic diagram of a clinical aid decision making device according to some embodiments of the invention;
FIG. 9 schematically illustrates a structural diagram of a computer system of an electronic device, in accordance with some embodiments of the present invention;
fig. 10 schematically illustrates a schematic diagram of a computer-readable storage medium according to some embodiments of the invention.
In the drawings, the same or corresponding reference numerals indicate the same or corresponding parts.
Detailed Description
Example embodiments will now be described more fully with reference to the accompanying drawings. However, the exemplary embodiments may be embodied in many forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the concept of the example embodiments to those skilled in the art.
Furthermore, the described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. In the following description, numerous specific details are provided to give a thorough understanding of embodiments of the invention. One skilled in the relevant art will recognize, however, that the invention may be practiced without one or more of the specific details, or with other methods, components, devices, steps, etc. In other instances, well-known methods, devices, implementations, or operations are not shown or described in detail to avoid obscuring aspects of the invention.
Moreover, the drawings are only schematic illustrations and are not necessarily drawn to scale. The block diagrams depicted in the figures are merely functional entities and do not necessarily correspond to physically separate entities. That is, the functional entities may be implemented in software, or in one or more hardware modules or integrated circuits, or in different networks and/or processor devices and/or microcontroller devices.
The inventors have found that clinical decision-making aid systems can be divided into knowledge-driven models and data-driven models. In a technical scheme, a clinical auxiliary decision-making system mainly comprises a knowledge driven model, wherein the knowledge driven model can be based on medical books and clinical guidelines, logic of medical diagnosis and treatment is realized in a mode of a conditional statement structure IF-THEN, according to input of symptom data and according to deterministic logic execution, a recommended decision-making result is finally determined, and the system is also called an expert system. However, in a simple knowledge-driven model, although the interpretation of the decision result is strong, the accuracy of the predicted decision result is low when the decision node is not 1, i.e., 0, and the decision result cannot be recommended in many cases due to incomplete coverage of the knowledge rules. In another technical scheme, the clinical auxiliary decision-making system mainly comprises a data-driven model, wherein the data-driven model can be a decision-making model constructed based on a retrieval or machine learning algorithm based on data, and the decision-making model gives out a final recommended decision-making result according to the input of symptom data. However, a simple data-driven model has higher accuracy of the obtained decision result, but is severely dependent on data quality, and the model based on machine learning, especially deep learning, has weaker interpretability, and the model has higher complexity and accuracy and weaker interpretability; sometimes, the generalization ability is weak, that is, the model training is good, but the effect is poor when the model training is changed to another data set for testing.
Based on this, in the present exemplary embodiment, there is provided first a target decision model construction method that can be applied to a terminal device or a server, and the present exemplary embodiment is not particularly limited thereto. In the following, the method is taken as an example of the execution of the method by a server, and fig. 1 schematically shows a flow chart of a method for constructing a target decision model according to some embodiments of the present invention. Referring to fig. 1, the target decision model construction method may include the steps of:
step S110, acquiring clinical guideline data and constructing a first decision model according to the clinical guideline data; wherein the first decision model comprises a plurality of decision branch paths;
step S120, obtaining characteristic data and label data corresponding to the characteristic data;
and step S130, training a pre-constructed second decision model through the initial decision data and the label data, and constructing a target decision model according to the trained second decision model and the first decision model.
According to the target decision model construction method in the present exemplary embodiment, on one hand, initial decision data is determined according to the first decision model, and a second decision model is trained according to the initial decision data and the label data, and finally, a target decision model is constructed according to the first decision model and the second decision model, so that the interpretability of a decision result obtained by the target decision model can be improved, and more reference data can be provided for a target object (such as a doctor); on the other hand, initial decision data is determined by carrying out preliminary decision on the feature data according to the first decision model, and a second decision model is trained according to the initial decision data and the label data so that the second decision model can further classify the initial decision data and determine a decision result, so that the accuracy of the decision result obtained by the target decision model can be improved.
Next, a method of constructing a target decision model in the present exemplary embodiment will be further described.
In step S110, clinical guideline data is acquired, and a first decision model is constructed from the clinical guideline data; wherein the first decision model comprises a plurality of decision branch paths.
In an exemplary embodiment of the present invention, the clinical guideline data may refer to reference data or a treatment scheme for diagnosing and treating a disease in the medical field, for example, the clinical guideline data may be a medical book for diagnosing and treating a disease, may be a clinical guideline, or may be other reference data or a treatment scheme for diagnosing and treating a disease, which is not particularly limited in this exemplary embodiment.
The first decision model may refer to a branch model constructed according to clinical guideline data, where the first decision model may include a plurality of decision branch paths, where each decision branch path corresponds to diagnosis and treatment logic of a target disease in the clinical guideline data, for example, the first decision model may be a model implemented by using a conditional structure IF-THEN in the clinical guideline data, or may be a rule decision tree model obtained by training according to the clinical guideline data, and of course, the first decision model may also be another model capable of implementing diagnosis logic in the clinical guideline data, which is not limited in this example embodiment.
In step S120, feature data corresponding to the target task and tag data corresponding to the feature data are acquired.
In one example embodiment of the invention, the feature data may be corresponding symptom data in real medical record data stored in a database of patient presence, the feature data being used to train the target decision model. The label data may be a corresponding treatment plan in real medical record data stored in the database, the patient occurring, the label data corresponding to the aforementioned feature data, the label data being used with the feature data to train the target decision model. The target task may refer to real medical record data in the clinical guideline data.
In step S130, training a second pre-constructed decision model according to the initial decision data and the label data, and constructing a target decision model according to the second decision model and the first decision model after training.
In an exemplary embodiment of the present invention, the second decision model may refer to a model that makes a further decision on the initial decision data obtained by the first decision model, and the second decision model may be a data model based on a machine learning algorithm, for example, the second decision model may be a deep learning model, a logistic regression model, a decision tree model, a random forest model, or a bayesian model, which is not limited in this exemplary embodiment. The target decision model may refer to a model for assisting a target object in making a diagnostic decision based on a combination of the first decision model and the second decision model.
Specifically, the feature data is input into a first decision model to determine a target decision branch path corresponding to the feature data, and initial decision data corresponding to the feature data is determined according to the target decision branch path. The target decision branch path may refer to diagnosis and treatment logic corresponding to feature data in clinical guideline data, and the initial decision data may refer to a diagnosis and treatment scheme or a diagnosis result corresponding to the target decision branch path. The target object can be enabled to be clear of the judgment logic corresponding to the initial decision data through the target decision branch path, and the interpretability of the initial decision data is improved.
In the present exemplary embodiment, the second decision model may include one or more of a subdivision decision model, an enhancement decision model, and a supplement decision model, and of course, the second decision model may also be a decision model with other functions, which is not limited thereto. For example, where only the initial decision data comprising a plurality of decision results need to be refined, the second decision model may be a subdivision decision model; when not only the initial decision data including a plurality of decision results needs to be finely classified, but also the initial decision data including a plurality of decision factors needs to be explicitly classified, the second decision model may be either a subdivision decision model or an enhancement decision model, and the type of the second decision model may depend on the initial decision data corresponding to the target decision branch path, which is not particularly limited in this example embodiment.
Training a second decision model according to initial decision data obtained by the first decision model and label data of feature data corresponding to the initial decision data, and splicing the first decision model and the trained second decision model to obtain a built target decision model.
FIG. 2 schematically illustrates a flow diagram for training a subdivision decision model, in accordance with some embodiments of the invention.
Referring to fig. 2, in step S210, when it is determined that the initial decision data corresponds to a plurality of decision results, a pre-constructed subdivision decision model is obtained.
In an exemplary embodiment of the present invention, the subdivision decision model may refer to a model that further classifies a plurality of decision results (of course, a plurality of treatment schemes may also be used, and this exemplary embodiment is not limited thereto), for example, it is assumed that the feature data may be fever, dry mouth, and the like, and the feature data may be input into the first decision model to obtain the initial decision data, for example, the initial decision data may be a treatment scheme corresponding to a cold and a common cold or a treatment scheme corresponding to inflammation and inflammation, and the like, where the initial decision data has two decision results, but the decision results at this time are classified according to two old categories, resulting in a more ambiguous decision result, and the classification is not accurate enough, and at this time, the initial decision data may be further determined by training the subdivision decision model in the second decision model, which is merely illustrative, and should not cause any special limitation in this exemplary embodiment.
And step S220, training the subdivision decision model according to a plurality of decision results and the tag data so that the subdivision decision model classifies the initial decision data.
In an exemplary embodiment of the present invention, when it is detected that the initial decision data includes a plurality of decision results or treatment schemes, the decision result obtained by the first decision model is considered to be ambiguous and cannot be directly used as a final decision result, so that the plurality of decision results or treatment schemes included in the initial decision data need to be further classified to determine the most accurate decision result corresponding to the feature data. Specifically, a subdivision decision model may be trained according to the initial decision data, the feature data, and the tag data, and the initial decision data including a plurality of decision results may be further classified according to the trained subdivision decision model, so as to determine a most accurate decision result corresponding to the feature data.
For example, a target decision branch path (which may also be considered as a diagnostic rule or diagnostic logic in a medical book or clinical guideline) corresponding to the feature data is determined from the first decision model, and initial decision data (which may also be considered as a plurality of treatment regimens recommended in a medical book or clinical guideline) corresponding to the target decision branch path is not indicated which treatment regimen is better. At this time, relevant features in the feature data are selected as model features, the real treatment scheme corresponding to the feature data is taken as Label data (Label), and a subdivision decision model is constructed based on the model features and the data set constructed by the Label data to subdivide the treatment scheme corresponding to the feature data entering the target decision branch path, which is, of course, only schematically illustrated herein, and should not be construed as being limiting in any way.
FIG. 3 schematically illustrates a flow diagram of training an enhanced decision model according to some embodiments of the invention.
Referring to fig. 3, in step S310, when it is determined that the initial decision data is to-be-enhanced decision data, a pre-constructed enhanced decision model is obtained.
In an exemplary embodiment of the present invention, the enhanced decision model may refer to a model that accurately classifies initial decision data including a plurality of decision factors obtained by the first decision model, for example, it is assumed that the initial decision data may be chemotherapy for a patient, but whether the chemotherapy for the patient needs to be performed comprehensively considering the decision factors such as age, physical condition, combined basic disease, etc., so that a "decision point" enhanced decision model needs to be added to this target decision branch path for judging "physical ability comprehensive assessment" of the patient corresponding to the feature data, when the enhanced decision model judges "physical ability comprehensive assessment" to be good, the chemotherapy for the patient corresponding to the feature data is performed, when the enhanced decision model judges "physical ability comprehensive assessment" to be bad, the chemotherapy for the patient corresponding to the feature data is not performed, and diagnostic logic of other decision branch paths is performed, which is only illustrative and should not cause any special limitation of the exemplary embodiment.
Step S320, training the enhanced decision model according to the feature data and the tag data to enable the enhanced decision model to determine a decision result corresponding to the decision data to be enhanced; wherein the decision result comprises a plurality of decision classifications.
In an exemplary embodiment of the present invention, the decision data to be enhanced may refer to initial decision data in the case that the first decision model judges that there is a high risk factor or that the T-stage (international TNM clinical stage) is T4 (may refer to primary tumor invasion and mediastinum, heart, great vessels, trachea, esophagus, vertebral body, carina; or malignant pleural effusion; for example, according to clinical guideline data, whether a patient in clinical stage II is to be subjected to chemotherapy needs to comprehensively consider factors that may affect the risk of chemotherapy such as age, physical condition, combined underlying disease, etc., so further enhancement classification is required for the decision data to be enhanced to determine a decision result. When the initial decision data is detected to be the decision data to be enhanced, the decision result obtained by the first decision model is considered to be enhanced decision, if the decision result is further decided, whether the physical energy comprehensive condition of the patient is suitable for chemotherapy is judged, so that the enhanced decision is needed to be performed on the decision data to be enhanced, the most accurate decision result corresponding to the characteristic data is determined, the accuracy of the decision result is ensured through the enhanced decision model, and the life safety of the patient is ensured.
Further, the label of the pre-labeled feature data is a target label, for example, when the physical performance integrated condition of the patient in the initial decision data needs to be determined, the label of the pre-labeled feature data may be a good physical performance condition (i.e., a target label).
Then judging whether the leaf node data of the feature data in the first decision model is consistent with the label data corresponding to the feature data; wherein the leaf node data comprises a first treatment plan and the label data comprises a second treatment plan, e.g. assuming 'physical condition' as 'good' (target label), then continuing to execute the rule tree in the first decision model to the leaf node, if the treatment plan of the leaf node is consistent with the treatment plan ultimately used by the patient, then assuming that 'target label' is correct, otherwise executing other assumptions (e.g. assuming 'physical condition' as 'bad') until the sample label is clear.
Specifically, when the first treatment scheme is determined to be consistent with the second treatment scheme, the assumed target label is considered to be correct, and the target label is taken as the reinforcement label data; when the first treatment scheme is inconsistent with the second treatment scheme, the assumed target label such as ' physical ability status is considered to be wrong, the ' physical ability ' label can be continuously assumed to be ' bad ', then whether leaf node data of the characteristic data in the first decision model is consistent with label data corresponding to the characteristic data is continuously judged, when the first treatment scheme is consistent with the second treatment scheme, the assumed target label such as ' physical ability status is determined to be bad ' is correct, and the target label is taken as strengthening label data.
And finally training the reinforced decision model according to the characteristic data, the tag data and the reinforced tag data so that the reinforced decision model carries out reinforced decision on the data to be reinforced and determines the decision classification corresponding to the data to be reinforced.
Fig. 4 schematically illustrates a flow diagram for training a supplemental decision model according to some embodiments of the invention.
Referring to fig. 4, in step S410, when it is determined that initial decision data corresponding to the feature data is feature data according to the target decision branch path, it is determined that the target decision branch path is in an empty state, and a pre-constructed complementary decision model is obtained.
In an exemplary embodiment of the present invention, when the initial decision data corresponding to the feature data is determined to be the feature data according to the target decision branch path, the first decision model is considered to not include the decision branch path corresponding to the feature data at this time, that is, the target decision branch path corresponding to the feature data is in an empty state. The empty state of the target decision branch path may indicate that the feature data is not covered in the clinical guideline data, and thus a decision branch path corresponding to the feature data needs to be added.
And step S420, training the supplementary decision model according to the characteristic data and the label data so as to enable the supplementary decision model to classify the characteristic data.
In an exemplary embodiment of the present invention, the supplementary decision model may refer to a model for making a decision on feature data not covered by the decision branch path, and may be considered as a decision branch path corresponding to the feature data through the supplementary decision model.
In the present exemplary embodiment, a clinical assistance decision making method is also provided, which may be applied to a terminal device or a server, and the present exemplary embodiment is not particularly limited thereto. In the following, the method is described by taking a server as an example, and fig. 5 schematically shows a flow chart of a clinical assistant decision making method according to some embodiments of the invention.
Referring to fig. 5, in step S510, an inputted decision influencing factor is acquired.
In an exemplary embodiment of the present invention, the decision influencing factors may refer to all relevant data that needs to be subjected to decision analysis to determine the decision result, for example, the decision influencing factors may be symptom data (or may also be basic information data, historical diagnosis data, and other data that may influence the decision result) of the patient input by the target object through the terminal device, or may also be text data on medical record data of the patient, where, of course, the decision influencing factors may also be symptom images of the patient acquired by the image acquisition device, and this embodiment is not limited specifically.
In step S510, initial decision data corresponding to the decision influencing factor is determined according to a first decision model in the pre-trained target decision model.
In one example embodiment of the present invention, the pre-trained target decision model may refer to a target decision model constructed by the aforementioned target decision model construction method, and the first decision model may refer to a model constituting the target decision model. The initial decision data may refer to inputting the decision influencing factors into the target decision model, so that a first decision model in the target decision model determines a target decision branch path corresponding to the decision influencing factors, and then determines initial decision data corresponding to the decision influencing factors according to the target decision branch path.
In step S510, auxiliary decision data corresponding to the initial decision data is determined according to a second decision model in the target decision model, so as to assist the target object to determine a decision result corresponding to the decision influencing factor according to the auxiliary decision data.
In an exemplary embodiment of the present invention, the auxiliary decision data may refer to a decision result obtained by further deciding by a second decision model in the target decision model according to the initial decision data obtained by the first decision model. The target object may refer to a diagnosis result or a treatment plan corresponding to the decision-making influencing factor according to the auxiliary decision-making data, for example, the target object may be a doctor for diagnosing the patient corresponding to the decision-making influencing factor, or may be a medical robot serving the patient, or may be another object capable of determining a diagnosis result or a diagnosis plan corresponding to the decision-making influencing factor according to the auxiliary decision-making data, which is not particularly limited in this example embodiment. And the target decision model outputs the determined auxiliary decision data to the terminal equipment associated with the target object so that the target object determines a decision result corresponding to the decision influencing factor according to the auxiliary decision data.
Specifically, when the initial decision data is determined to be the decision data to be enhanced, the decision data to be enhanced is input into an enhanced decision model to obtain the decision classification of the decision data to be enhanced; and taking the decision classification as auxiliary decision data corresponding to the initial decision data. The decision data to be enhanced may refer to initial decision data under the condition that the first decision model judges that there is a high risk factor or that the T stage (international TNM clinical stage) is T4 (may refer to invasion of primary tumor and mediastinum, heart, great vessels, trachea, esophagus, centrum, carina; or malignant pleural effusion), for example, according to clinical guideline data, whether a patient in clinical stage II is to be subjected to chemotherapy needs to comprehensively consider factors that may affect the risk of chemotherapy, such as age, physical condition, combined basic disease, etc., so that further enhancement classification is needed for the decision data to be enhanced to determine decision results. When the initial decision data is detected to be the decision data to be enhanced, the enhancement decision model in the second decision model is considered to be required to carry out enhancement decision on the obtained decision result, and if the decision is further required, whether the physical ability comprehensive condition of the patient is suitable for chemotherapy is judged, so that the enhancement decision classification is required to be carried out on the decision data to be enhanced. For example, for decision influencing factors classified as high risk factors or as T4 belonging to T-score, factors possibly causing medical risk such as age, physical condition, combined underlying diseases and the like need to be comprehensively considered, and an enhanced decision model is required as a "decision point" to perform decision classification on the decision influencing factors to determine whether physical condition of a patient is good or bad for further diagnosis decision.
FIG. 6 schematically illustrates a schematic diagram of a target decision model structure according to some embodiments of the invention.
Referring to fig. 6, the target decision model 600 may include a first decision model 601 and a second decision model, where the second decision model may include a supplemental decision model 606, an enhanced decision model 611, and a refined decision model 609. The first decision model 601 may include a plurality of decision branch paths, which may include, for example, a decision branch path 602, a decision branch path 603, and a decision branch path 604.
When the feature data flows to the decision branch path 602 (i.e. the feature data can be considered not to be covered by the decision branch path of the first decision model 601), the decision result of the feature data needs to be determined, but the decision result of the feature data is not determined enough by the first decision model 601, the branch path is considered to have a new feature 605 (i.e. the feature corresponding to the feature data), so that a complementary decision model 606 is trained to replace the decision branch path 602 in the empty state to perform decision classification on the feature data 605 flowing to the branch.
For the feature data flowing to the decision branch path 604, the first decision model 601 makes a first decision on the feature data according to the diagnosis logic 607 "whether the feature data is a high risk factor", and determines a diagnosis result 608 (or diagnosis scheme) corresponding to the feature data when the "no high risk factor" is determined, but the diagnosis result 608 includes a plurality of diagnosis schemes, and the diagnosis result is ambiguous, so that the training refinement decision model 609 is required to make a further classification decision on the diagnosis result 608 to determine the most correct decision result corresponding to the feature data flowing to the decision branch path 604.
When the decision branch path 603 is diverted and the diagnosis logic 607 determines that there is a "high risk factor", the decision point 610 is required to determine the "physical condition" according to the decision factors such as the age, physical condition, basic disease, etc. of the patient, so as to determine the decision result under different decision factors. Specifically, the enhanced decision model 611 may be trained to make final decisions on the feature data that flows to the decision point based on decision factors to determine decision results.
It should be noted that although the steps of the method of the present invention are depicted in the drawings in a particular order, this does not require or imply that the steps must be performed in that particular order or that all of the illustrated steps be performed in order to achieve desirable results. Additionally or alternatively, certain steps may be omitted, multiple steps combined into one step to perform, and/or one step decomposed into multiple steps to perform, etc.
In addition, in the present exemplary embodiment, a target decision model construction apparatus is also provided. Referring to fig. 7, the target decision model construction apparatus 700 includes: the first decision model construction module 710 is configured to obtain clinical guideline data and construct a first decision model according to the clinical guideline data; wherein the first decision model comprises a plurality of decision branch paths; the feature data obtaining module 720 is configured to obtain feature data and tag data corresponding to the feature data; the target decision model construction module 730 is configured to train the pre-constructed second decision model according to the initial decision data and the tag data, and construct a target decision model according to the trained second decision model and the first decision model.
In an exemplary embodiment of the present invention, based on the foregoing solution, the target decision model building module 730 further includes a second decision model training unit configured to: inputting the characteristic data into the first decision model to determine a target decision branch path corresponding to the characteristic data, and determining initial decision data corresponding to the characteristic data according to the target decision branch path; training a second pre-constructed decision model according to the initial decision data and the label data.
In an exemplary embodiment of the present invention, based on the foregoing scheme, the target decision model construction module 730 further includes a subdivision decision model training unit configured to: when the initial decision data is determined to correspond to a plurality of decision results, acquiring a pre-constructed subdivision decision model; training the subdivision decision model according to a plurality of decision results and the tag data so that the subdivision decision model classifies the initial decision data.
In an exemplary embodiment of the present invention, based on the foregoing solution, the objective decision model construction module 730 further includes an enhanced decision model training unit configured to: when the initial decision data is determined to be the decision data to be enhanced, acquiring a pre-constructed enhanced decision model; training the reinforced decision model according to the characteristic data and the tag data so that the reinforced decision model determines a decision result corresponding to the decision data to be reinforced; wherein the decision result comprises a plurality of decision classifications.
In an exemplary embodiment of the invention, based on the foregoing, the enhanced decision model training unit is further configured to: marking the label of the characteristic data as a target label in advance, and judging whether leaf node data of the characteristic data in the first decision model is consistent with label data corresponding to the characteristic data; wherein the leaf node data comprises a first treatment regimen and the tag data comprises a second treatment regimen; upon determining that the first treatment regimen is consistent with the second treatment regimen, taking the target label as reinforcement label data; training the reinforced decision model according to the characteristic data, the tag data and the reinforced tag data so that the reinforced decision model carries out reinforced decision on the data to be reinforced and determines the decision classification corresponding to the data to be reinforced.
In an exemplary embodiment of the present invention, based on the foregoing solution, the target decision model construction module 730 further includes a supplemental decision model training unit configured to: when the initial decision data corresponding to the feature data is determined to be the feature data according to the target decision branch path, determining that the target decision branch path is in an empty state and acquiring a pre-constructed supplementary decision model; training the supplemental decision model according to the feature data and the tag data to enable the supplemental decision model to classify the feature data.
In an exemplary embodiment of the present invention, based on the foregoing, the first decision model includes a rule decision tree model, and the second decision model includes any one of a decision tree model, a logistic regression model, a random forest model, a fibrate model, and a deep learning model.
In this example embodiment, a clinical aid decision making apparatus is also provided. Referring to fig. 8, the objective decision model construction apparatus 800 includes: the decision-making influencing factor obtaining unit 810 is configured to obtain an input decision-making influencing factor; the initial decision data determining unit 820 is configured to determine initial decision data corresponding to the decision influencing factor according to a first decision model in the pre-trained target decision model; the auxiliary decision data determining unit 830 is configured to determine auxiliary decision data corresponding to the initial decision data according to a second decision model in the target decision model, so as to assist a target object to determine a decision result corresponding to the decision influencing factor according to the auxiliary decision data.
In an exemplary embodiment of the present invention, based on the foregoing scheme, the auxiliary decision data determining unit 830 is further configured to: when the initial decision data is determined to be the decision data to be enhanced, inputting the decision data to be enhanced into the enhanced decision model to obtain the decision classification of the decision data to be enhanced; and taking the decision classification as auxiliary decision data corresponding to the initial decision data.
The specific details of each module of the above-mentioned target decision model building device or clinical auxiliary decision device are already described in detail in the corresponding target decision model building method or clinical auxiliary decision method, so that the details are not repeated here.
It should be noted that although several modules or units of the target decision model building apparatus are mentioned in the above detailed description, this partitioning is not mandatory. Indeed, the features and functions of two or more modules or units described above may be embodied in one module or unit in accordance with embodiments of the invention. Conversely, the features and functions of one module or unit described above may be further divided into a plurality of modules or units to be embodied.
In addition, in the exemplary embodiment of the disclosure, an electronic device capable of implementing the above-mentioned target decision model construction method is also provided.
Those skilled in the art will appreciate that the various aspects of the invention may be implemented as a system, method, or program product. Accordingly, aspects of the invention may be embodied in the following forms, namely: an entirely hardware embodiment, an entirely software embodiment (including firmware, micro-code, etc.) or an embodiment combining hardware and software aspects may be referred to herein as a "circuit," module "or" system.
An electronic device 900 according to such an embodiment of the invention is described below with reference to fig. 9. The electronic device 900 shown in fig. 9 is merely an example, and should not be construed as limiting the functionality and scope of use of embodiments of the present invention.
As shown in fig. 9, the electronic device 900 is embodied in the form of a general purpose computing device. Components of electronic device 900 may include, but are not limited to: the at least one processing unit 910, the at least one storage unit 920, a bus 930 connecting the different system components (including the storage unit 920 and the processing unit 910), and a display unit 940.
Wherein the storage unit stores program code that is executable by the processing unit 910 such that the processing unit 910 performs steps according to various exemplary embodiments of the present invention described in the above-described "exemplary methods" section of the present specification. For example, the processing unit 910 may perform step S110 shown in fig. 1, acquire clinical guideline data, and construct a first decision model from the clinical guideline data; wherein the first decision model comprises a plurality of decision branch paths; step S120, obtaining characteristic data and label data corresponding to the characteristic data; and step S130, training a pre-constructed second decision model through the initial decision data and the label data, and constructing a target decision model according to the trained second decision model and the first decision model.
The storage unit 920 may include readable media in the form of volatile storage units, such as Random Access Memory (RAM) 921 and/or cache memory 922, and may further include Read Only Memory (ROM) 923.
The storage unit 920 may also include a program/utility 924 having a set (at least one) of program modules 925, such program modules 925 including, but not limited to: an operating system, one or more application programs, other program modules, and program data, each or some combination of which may include an implementation of a network environment.
The bus 930 may be one or more of several types of bus structures including a memory unit bus or memory unit controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local bus using any of a variety of bus architectures.
The electronic device 900 may also communicate with one or more external devices 970 (e.g., keyboard, pointing device, bluetooth device, etc.), one or more devices that enable a user to interact with the electronic device 900, and/or any device (e.g., router, modem, etc.) that enables the electronic device 900 to communicate with one or more other computing devices. Such communication may occur through an input/output (I/O) interface 950. Also, electronic device 900 may communicate with one or more networks such as a Local Area Network (LAN), a Wide Area Network (WAN), and/or a public network, such as the Internet, through network adapter 960. As shown, the network adapter 960 communicates with other modules of the electronic device 900 over the bus 930. It should be appreciated that although not shown, other hardware and/or software modules may be used in connection with electronic device 900, including, but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, data backup storage systems, and the like.
From the above description of embodiments, those skilled in the art will readily appreciate that the example embodiments described herein may be implemented in software, or in combination with the necessary hardware. Thus, the technical solution according to the embodiments of the present disclosure may be embodied in the form of a software product, which may be stored in a non-volatile storage medium (may be a CD-ROM, a U-disk, a mobile hard disk, etc.) or on a network, and includes several instructions to cause a computing device (may be a personal computer, a server, a terminal device, or a network device, etc.) to perform the method according to the embodiments of the present disclosure.
In an exemplary embodiment of the present disclosure, a computer-readable storage medium having stored thereon a program product capable of implementing the method described above in the present specification is also provided. In some possible embodiments, the various aspects of the invention may also be implemented in the form of a program product comprising program code for causing a terminal device to carry out the steps according to the various exemplary embodiments of the invention as described in the "exemplary methods" section of this specification, when said program product is run on the terminal device.
Referring to fig. 10, a program product 1000 for implementing the above-described objective decision model construction method or clinical aid decision making method, which may employ a portable compact disc read only memory (CD-ROM) and include program code, and which may be run on a terminal device, such as a personal computer, is described according to an embodiment of the present invention. However, the program product of the present invention is not limited thereto, and in this document, a readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.
The program product may employ any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium can be, for example, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or a combination of any of the foregoing. More specific examples (a non-exhaustive list) of the readable storage medium would include the following: an electrical connection having one or more wires, a portable disk, a hard disk, random Access Memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.
The computer readable signal medium may include a data signal propagated in baseband or as part of a carrier wave with readable program code embodied therein. Such a propagated data signal may take any of a variety of forms, including, but not limited to, electro-magnetic, optical, or any suitable combination of the foregoing. A readable signal medium may also be any readable medium that is not a readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device.
Program code embodied on a readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.
Program code for carrying out operations of the present invention may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, C++ or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computing device, partly on the user's device, as a stand-alone software package, partly on the user's computing device, partly on a remote computing device, or entirely on the remote computing device or server. In the case of remote computing devices, the remote computing device may be connected to the user computing device through any kind of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or may be connected to an external computing device (e.g., connected via the Internet using an Internet service provider).
Furthermore, the above-described drawings are only schematic illustrations of processes included in the method according to the exemplary embodiment of the present invention, and are not intended to be limiting. It will be readily appreciated that the processes shown in the above figures do not indicate or limit the temporal order of these processes. In addition, it is also readily understood that these processes may be performed synchronously or asynchronously, for example, among a plurality of modules.
From the above description of embodiments, those skilled in the art will readily appreciate that the example embodiments described herein may be implemented in software, or in combination with the necessary hardware. Thus, the technical solution according to the embodiments of the present disclosure may be embodied in the form of a software product, which may be stored in a non-volatile storage medium (may be a CD-ROM, a U-disk, a mobile hard disk, etc.) or on a network, and includes several instructions to cause a computing device (may be a personal computer, a server, a touch terminal, or a network device, etc.) to perform the method according to the embodiments of the present disclosure.
Other embodiments of the disclosure will be apparent to those skilled in the art from consideration of the specification and practice of the disclosure disclosed herein. This application is intended to cover any adaptations, uses, or adaptations of the disclosure following, in general, the principles of the disclosure and including such departures from the present disclosure as come within known or customary practice within the art to which the disclosure pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the disclosure being indicated by the following claims.
It is to be understood that the present disclosure is not limited to the precise arrangements and instrumentalities shown in the drawings, and that various modifications and changes may be effected without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.
Claims (9)
1. The method for constructing the target decision model is characterized by comprising the following steps of:
acquiring clinical guideline data and constructing a first decision model according to the clinical guideline data; wherein the first decision model comprises a plurality of decision branch paths;
acquiring feature data corresponding to a target task and tag data corresponding to the feature data;
training a second pre-constructed decision model through the feature data and the tag data, and performing splicing processing on the second decision model and the first decision model after training to obtain a target decision model, wherein the second decision model comprises one or more of a subdivision decision model, an enhancement decision model and a supplement decision model;
the training of the pre-constructed second decision model by the feature data and the label data comprises: inputting the characteristic data into the first decision model to determine a target decision branch path corresponding to the characteristic data, and determining initial decision data corresponding to the characteristic data according to the target decision branch path; training a second pre-constructed decision model according to the initial decision data and the label data;
Wherein training the pre-constructed second decision model comprises:
when the initial decision data is determined to be the decision data to be enhanced, acquiring a pre-constructed enhanced decision model; training the reinforced decision model according to the characteristic data and the tag data so that the reinforced decision model determines a decision result corresponding to the decision data to be reinforced; wherein the decision result comprises a plurality of decision classifications;
the training of the enhanced decision model includes: marking a label of the characteristic data as a target label in advance, and judging whether leaf node data of the characteristic data in the first decision model is consistent with label data corresponding to the characteristic data; wherein the leaf node data comprises a first treatment regimen and the tag data comprises a second treatment regimen; upon determining that the first treatment regimen is consistent with the second treatment regimen, taking the target label as reinforcement label data; training the reinforced decision model according to the characteristic data, the tag data and the reinforced tag data so that the reinforced decision model carries out reinforced decision on the data to be reinforced and determines the decision classification corresponding to the data to be reinforced.
2. The method for constructing a target decision model according to claim 1, wherein training the second pre-constructed decision model according to the initial decision data and the tag data further comprises:
when the initial decision data is determined to correspond to a plurality of decision results, acquiring a pre-constructed subdivision decision model;
training the subdivision decision model according to a plurality of decision results and the tag data so that the subdivision decision model classifies the initial decision data.
3. The method for constructing a target decision model according to claim 1, wherein training the second pre-constructed decision model according to the initial decision data and the tag data further comprises:
when the initial decision data corresponding to the feature data is determined to be the feature data according to the target decision branch path, determining that the target decision branch path is in an empty state and acquiring a pre-constructed supplementary decision model;
training the supplemental decision model according to the feature data and the tag data to enable the supplemental decision model to classify the feature data.
4. A method of constructing a target decision model according to any one of claims 1 to 3, wherein the first decision model comprises a rule decision tree model and the second decision model comprises any one of a decision tree model, a logistic regression model, a random forest model, a bayesian model, and a deep learning model.
5. A method of clinical aid decision making, comprising:
acquiring input decision influencing factors;
determining initial decision data corresponding to the decision influencing factors according to a first decision model in a pre-trained target decision model;
determining auxiliary decision data corresponding to the initial decision data according to a second decision model in the target decision model so as to assist a target object to determine a decision result corresponding to the decision influence factor according to the auxiliary decision data, wherein the second decision model comprises one or more of a subdivision decision model, an enhancement decision model and a supplement decision model;
the target decision model is obtained by splicing the first decision model and the trained second decision model, wherein characteristic data are input into the first decision model to determine a target decision branch path corresponding to the characteristic data, and the initial decision data corresponding to the characteristic data are determined according to the target decision branch path; training the second pre-constructed decision model according to the initial decision data and the label data;
wherein training the second decision model comprises: when the initial decision data is determined to be the decision data to be enhanced, acquiring a pre-constructed enhanced decision model; training the reinforced decision model according to the characteristic data and the tag data so that the reinforced decision model determines a decision result corresponding to the decision data to be reinforced; wherein the decision result comprises a plurality of decision classifications;
The training of the enhanced decision model includes: marking a label of the characteristic data as a target label in advance, and judging whether leaf node data of the characteristic data in the first decision model is consistent with label data corresponding to the characteristic data; wherein the leaf node data comprises a first treatment regimen and the tag data comprises a second treatment regimen; upon determining that the first treatment regimen is consistent with the second treatment regimen, taking the target label as reinforcement label data; training the reinforced decision model according to the characteristic data, the tag data and the reinforced tag data so that the reinforced decision model carries out reinforced decision on the data to be reinforced and determines the decision classification corresponding to the data to be reinforced.
6. The method of claim 5, wherein the second decision model comprises an enhanced decision model, wherein determining auxiliary decision data corresponding to the initial decision data from a second decision model of the target decision models comprises:
when the initial decision data is determined to be the decision data to be enhanced, inputting the decision data to be enhanced into the enhanced decision model to obtain the decision classification of the decision data to be enhanced;
And taking the decision classification as auxiliary decision data corresponding to the initial decision data.
7. A target decision model building apparatus, comprising:
the first decision model construction module is used for acquiring clinical guideline data and constructing a first decision model according to the clinical guideline data; wherein the first decision model comprises a plurality of decision branch paths;
the characteristic data acquisition module is used for acquiring characteristic data and tag data corresponding to the characteristic data;
the target decision model construction module is used for training a second pre-constructed decision model through the characteristic data and the tag data, and performing splicing processing on the second decision model and the first decision model after training to obtain a target decision model; wherein the second decision model comprises one or more of a subdivision decision model, an enhanced decision model and a supplemental decision model;
the target decision model building module includes a second decision model training unit configured to: inputting the characteristic data into the first decision model to determine a target decision branch path corresponding to the characteristic data, and determining initial decision data corresponding to the characteristic data according to the target decision branch path; training a second pre-constructed decision model according to the initial decision data and the label data;
Wherein training the pre-constructed second decision model comprises: when the initial decision data is determined to be the decision data to be enhanced, acquiring a pre-constructed enhanced decision model; training the reinforced decision model according to the characteristic data and the tag data so that the reinforced decision model determines a decision result corresponding to the decision data to be reinforced; wherein the decision result comprises a plurality of decision classifications;
the training of the enhanced decision model includes: marking a label of the characteristic data as a target label in advance, and judging whether leaf node data of the characteristic data in the first decision model is consistent with label data corresponding to the characteristic data; wherein the leaf node data comprises a first treatment regimen and the tag data comprises a second treatment regimen; upon determining that the first treatment regimen is consistent with the second treatment regimen, taking the target label as reinforcement label data; training the reinforced decision model according to the characteristic data, the tag data and the reinforced tag data so that the reinforced decision model carries out reinforced decision on the data to be reinforced and determines the decision classification corresponding to the data to be reinforced.
8. An electronic device, comprising:
a processor; and
a memory having stored thereon computer readable instructions which, when executed by the processor, implement the target decision model construction method of any one of claims 1 to 4 or the clinical aid decision method of any one of claims 5 to 6.
9. A computer readable storage medium having stored thereon a computer program which when executed by a processor implements the target decision model construction method of any one of claims 1 to 4 or the clinical auxiliary decision method of any one of claims 5 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911416966.2A CN111145905B (en) | 2019-12-31 | 2019-12-31 | Target decision model construction method and device, electronic equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911416966.2A CN111145905B (en) | 2019-12-31 | 2019-12-31 | Target decision model construction method and device, electronic equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111145905A CN111145905A (en) | 2020-05-12 |
| CN111145905B true CN111145905B (en) | 2024-04-05 |
Family
ID=70522873
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911416966.2A Active CN111145905B (en) | 2019-12-31 | 2019-12-31 | Target decision model construction method and device, electronic equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111145905B (en) |
Families Citing this family (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111582489B (en) * | 2020-05-14 | 2023-07-14 | 上海深至信息科技有限公司 | Distributed deployment decision-making system and method of ultrasonic artificial intelligent model |
| CN111816300A (en) * | 2020-06-30 | 2020-10-23 | 平安科技（深圳）有限公司 | Data classification method, device, equipment and medium based on deep reinforcement learning |
| CN111967581B (en) * | 2020-08-06 | 2023-10-31 | 平安科技（深圳）有限公司 | Method, device, computer equipment and storage medium for interpreting grouping model |
| CN112365975B (en) * | 2020-11-12 | 2021-07-23 | 医渡云（北京）技术有限公司 | Medical aid decision model interpretation method and device, storage medium and electronic equipment |
| CN112600282B (en) * | 2021-01-06 | 2021-06-25 | 亨铨实业(深圳)有限公司 | Intelligent solar luggage charging and discharging control method and system |
| CN114745289A (en) * | 2022-04-19 | 2022-07-12 | 中国联合网络通信集团有限公司 | Method, device, storage medium and equipment for predicting network performance data |
| CN114818913A (en) * | 2022-04-22 | 2022-07-29 | 北京百度网讯科技有限公司 | Decision generation method and device |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109378065A (en) * | 2018-10-30 | 2019-02-22 | 医渡云（北京）技术有限公司 | Medical data processing method and processing device, storage medium, electronic equipment |
| CN109801705A (en) * | 2018-12-12 | 2019-05-24 | 平安科技（深圳）有限公司 | Treat recommended method, system, device and storage medium |
| CN110111887A (en) * | 2019-05-15 | 2019-08-09 | 清华大学 | Clinical aid decision-making method and device |
| CN110289095A (en) * | 2019-06-28 | 2019-09-27 | 青岛百洋智能科技股份有限公司 | A kind of fracture of neck of femur clinic intelligence aided decision method and system |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7233931B2 (en) * | 2003-12-26 | 2007-06-19 | Lee Shih-Jong J | Feature regulation for hierarchical decision learning |
| JP5149900B2 (en) * | 2006-09-20 | 2013-02-20 | コーニンクレッカ フィリップス エレクトロニクス エヌ ヴィ | Molecular diagnostic support system |
-
2019
- 2019-12-31 CN CN201911416966.2A patent/CN111145905B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109378065A (en) * | 2018-10-30 | 2019-02-22 | 医渡云（北京）技术有限公司 | Medical data processing method and processing device, storage medium, electronic equipment |
| CN109801705A (en) * | 2018-12-12 | 2019-05-24 | 平安科技（深圳）有限公司 | Treat recommended method, system, device and storage medium |
| CN110111887A (en) * | 2019-05-15 | 2019-08-09 | 清华大学 | Clinical aid decision-making method and device |
| CN110289095A (en) * | 2019-06-28 | 2019-09-27 | 青岛百洋智能科技股份有限公司 | A kind of fracture of neck of femur clinic intelligence aided decision method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111145905A (en) | 2020-05-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111145905B (en) | Target decision model construction method and device, electronic equipment and storage medium | |
| Sullivan | Understanding from machine learning models | |
| Dias et al. | Artificial intelligence in clinical and genomic diagnostics | |
| CN113990495B (en) | Disease diagnosis prediction system based on graph neural network | |
| US10902588B2 (en) | Anatomical segmentation identifying modes and viewpoints with deep learning across modalities | |
| US11688518B2 (en) | Deep neural network based identification of realistic synthetic images generated using a generative adversarial network | |
| Lee et al. | Machine learning in relation to emergency medicine clinical and operational scenarios: an overview | |
| CN109378065A (en) | Medical data processing method and processing device, storage medium, electronic equipment | |
| US10984024B2 (en) | Automatic processing of ambiguously labeled data | |
| D’Antonoli et al. | Large language models in radiology: fundamentals, applications, ethical considerations, risks, and future directions | |
| CN109448858A (en) | Colorectal cancer medical data processing method and processing device, storage medium, electronic equipment | |
| WO2024120504A1 (en) | Data processing method and related device | |
| Akinci D'Antonoli et al. | Large language models in radiology: fundamentals, applications, ethical considerations, risks, and future directions | |
| Wu et al. | Structured information extraction of pathology reports with attention-based graph convolutional network | |
| Melnyk et al. | Generative artificial intelligence terminology: a primer for clinicians and medical researchers | |
| CN111403028B (en) | Medical text classification method and device, storage medium and electronic equipment | |
| Alur et al. | Auditing for human expertise | |
| Abut et al. | Deep Neural Networks and Applications in Medical Research | |
| Zhan | DL 101: Basic introduction to deep learning with its application in biomedical related fields | |
| CN114647737A (en) | Medical rule completion method and device | |
| Li et al. | Sure: screening unlabeled samples for reliable negative samples based on reinforcement learning | |
| Rguibi et al. | CXAI: Explaining Convolutional Neural Networks for Medical Imaging Diagnostic. Electronics 2022, 11, 1775 | |
| Shaik et al. | ML Applications in Healthcare | |
| Klontzas et al. | Large language models in radiology: fundamentals, applications, ethical considerations, risks, and future directions | |
| Jalaboi | Explainable Neural Networks for Skin Lesion Diagnosis |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |