CN-122000006-A - Diagnosis assisting decision-making method and device, electronic equipment and storage medium

Abstract

The embodiment of the application provides a diagnosis auxiliary decision-making method, a diagnosis auxiliary decision-making device, electronic equipment and a storage medium, and relates to the technical field of artificial intelligence, wherein the method comprises the steps of acquiring diagnosis data of a target patient, processing the diagnosis data and extracting characteristic information representing the illness state of the target patient; the method comprises the steps of matching the characteristic information with a preset association rule base, acquiring the supplementary information of a target patient under the condition of successful matching, updating the characteristic information according to the supplementary information, inputting the updated characteristic information into a pre-trained triage model, and outputting the supplementary decision information of the target patient, wherein the triage model is trained based on historical triage data. By the method provided by the embodiment of the application, the information difference caused by unclear patient complaints or limited sign information is effectively compensated, more reliable auxiliary decision advice is output, and the misclassification risk caused by insufficient information is reduced from the whole flow.

Inventors

• ZHAO YICONG

Assignees

• 中国移动通信集团有限公司

Dates

Publication Date

20260508

Application Date

20260211

Claims (16)

1. 1. A triage aid decision making method, comprising: Acquiring diagnosis data of a target patient, processing the diagnosis data, and extracting characteristic information representing the illness state of the target patient; based on the characteristic information and a preset association rule base, acquiring the supplementary information of the target patient under the condition of successful matching, and updating the characteristic information according to the supplementary information; And inputting the updated characteristic information into a pre-trained triage model, and outputting auxiliary decision information of the target patient, wherein the triage model is obtained based on historical triage data training.
2. 2. The triage aid decision making method according to claim 1, wherein the visit data includes numerical physiological indicators and text complaint data; The processing the visit data to extract characteristic information representing the condition of the target patient comprises the following steps: performing abnormal value detection and elimination processing on the numerical physiological indexes, and performing missing value filling processing on the data subjected to abnormal value elimination to obtain first characteristic sub-information; performing term consistency standardization processing on the text type main complaint data to obtain second characteristic sub-information; and forming the characteristic information through the first characteristic sub-information and the second characteristic sub-information.
3. 3. The diagnosis-assisting decision-making method according to claim 2, wherein the term consistency normalization processing is performed on the text-type complaint data to obtain second characteristic sub-information, including: carrying out term consistency standardization processing on the text type main complaint data to obtain standardized main complaint data; And inputting the standardized complaint data into a pre-trained medical entity identification model, and outputting the identified symptoms, signs and disease entities as the second characteristic sub-information, wherein the medical entity identification model is constructed based on a two-way long-short-term memory network and a conditional random field framework.
4. 4. The triage aid decision making method according to claim 2, wherein the composing the feature information by the first feature sub-information and the second feature sub-information includes: combining the first characteristic sub-information and the second characteristic sub-information, and performing characteristic engineering processing on the combined data to generate the characteristic information; the feature engineering processing comprises the steps of carrying out segmentation discretization processing on the digital type feature and carrying out word frequency-inverse document frequency weighting and transcoding processing on the text type entity feature.
5. 5. The triage aid decision making method according to claim 1, wherein the process of constructing the preset association rule base includes: acquiring the historical triage data and constructing a characteristic data set based on the historical triage data; scanning the characteristic data set, and extracting at least one high-frequency symptom combination meeting preset frequency conditions to form a frequent item set; Performing multiple scans based on the frequent item set, and iteratively calculating the support and confidence of the at least one high-frequency symptom combination; Based on the support and confidence of the at least one high frequency symptom combination, strong rules associated with the target high risk disease are determined to construct the association rule base.
6. 6. The triage aid decision making method according to claim 1, wherein the matching based on the feature information and a preset association rule base, and obtaining the supplementary information of the target patient if the matching is successful, includes: Comparing the main complaint symptoms in the characteristic information with rule front pieces in the association rule base to generate comparison results, wherein the main complaint symptoms are uncomfortable symptoms actively stated by the target patient, and the rule front pieces are rule condition parts; Triggering a supplementary question process aiming at the target patient under the condition that the comparison result indicates that the comparison is successful, wherein the supplementary question process is to automatically push at least one key main complaint corresponding to the triggered rule for inquiring; And receiving a response result of the at least one key complaint inquiry, and taking the response result as the supplementary information.
7. 7. The triage aid decision making method according to claim 1, wherein the triage model comprises a grading prediction module and a severe prediction module, wherein the aid decision making information comprises triage grade and severe risk prediction results; The step of inputting the updated characteristic information into a pre-trained triage model and outputting the auxiliary decision information of the target patient comprises the following steps: And respectively inputting the updated characteristic information into the grading prediction module and the severe prediction module, and outputting the triage level of the target patient and the severe risk prediction result of the target patient.
8. 8. The triage aid decision making method according to claim 7, wherein, The hierarchical prediction module is configured to employ a multi-model integrated architecture that integrates at least four neural network models, a text convolutional neural network, a two-way long and short term memory network, a shading correction pre-training language model, and a meta controller pre-training language model.
9. 9. The triage aid decision making method according to claim 8, wherein the inputting the updated feature information into the hierarchical prediction module outputs the triage level of the target patient comprises: Based on the updated characteristic information, the real-time load and the data characteristics of the updated characteristic information, the weight of the output result of the integrated neural network model is dynamically adjusted through the hierarchical prediction module so as to jointly determine the triage level of the target patient.
10. 10. The triage aid decision making method of claim 9, wherein the dynamically adjusting weights of the integrated neural network model output results by the hierarchical prediction module based on the updated feature information, real-time load, and data features of the updated feature information comprises: And under the condition that the system throughput requirement is higher than a preset threshold, increasing the weight of the text convolutional neural network model and reducing the weight of the two-way long-short-term memory network model.
11. 11. The triage aid decision making method according to claim 7, wherein the severe prediction module is a gradient lifting decision tree model, and the aid decision making information further comprises an interpretability report corresponding to the severe risk prediction result; The step of inputting the updated characteristic information into a pre-trained triage model, outputting auxiliary decision information of the target patient, and further comprising: inputting the updated characteristic information into the severe risk prediction module, and calculating the contribution degree of each characteristic in the updated characteristic information to the severe risk prediction result; And generating the interpretability report for explaining the influence degree of each feature according to the contribution degree of each feature to the severe risk prediction result.
12. 12. The triage aid decision making method according to claim 5, wherein the determining strong rules associated with the target high risk disease based on the support and confidence of the at least one high frequency symptom combination, after constructing the association rule base, further comprises: and acquiring new triage data according to a preset period, and re-executing strong rule mining based on the new triage data so as to update the association rule base.
13. 13. A triage aid decision making apparatus, comprising: the acquisition module is used for acquiring the diagnosis data of the target patient, processing the diagnosis data and extracting characteristic information representing the illness state of the target patient; the updating module is used for matching the characteristic information with a preset association rule base, acquiring the supplementary information of the target patient under the condition of successful matching, and updating the characteristic information according to the supplementary information; The decision module is used for inputting the updated characteristic information into a pre-trained triage model and outputting auxiliary decision information of the target patient, wherein the triage model is obtained based on historical triage data training.
14. 14. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the triage aid decision making method of any one of claims 1 to 12 when the computer program is executed by the processor.
15. 15. A non-transitory computer readable storage medium having stored thereon a computer program, which when executed by a processor implements the triage aid decision making method of any one of claims 1 to 12.
16. 16. A computer program product comprising a computer program, characterized in that the computer program, when executed by a processor, implements the triage aid decision making method of any one of claims 1 to 12.

Description

Diagnosis assisting decision-making method and device, electronic equipment and storage medium Technical Field The embodiment of the application relates to the technical field of artificial intelligence, in particular to a triage auxiliary decision-making method, a triage auxiliary decision-making device, electronic equipment and a storage medium. Background The emergency department is used as a core front line for the treatment of critical patients in hospitals, and the accuracy and the efficiency of the triage work are directly related to the life safety of the patients and the reasonable distribution of medical resources. At present, triage work is generally based on personal experience and subjective judgment of triage nurses, and when facing to fuzzy complaints of patients, limited physical sign information or complex hidden illness, level error score is easy to generate, so that critical patients delay treatment or non-emergency patients occupy valuable rescuing resources. In recent years, although research attempts are made to introduce a machine learning model to assist diagnosis decision, the prior art schemes are limited to passive classification of existing static data sets, and prediction accuracy, robustness and interpretability of the existing static data sets are improved when dealing with data characteristics and computational load of dynamic changes of emergency scenes, so that the actual requirements of clinical assistance decision on high reliability are difficult to meet. Therefore, a solution is needed to solve the above-mentioned problems. Disclosure of Invention The embodiment of the application provides a diagnosis assisting decision-making method, a diagnosis assisting decision-making device, electronic equipment and a storage medium, which are used for solving the defects in the prior art. The embodiment of the application provides a triage aid decision making method, which comprises the following steps: Acquiring diagnosis data of a target patient, processing the diagnosis data, and extracting characteristic information representing the illness state of the target patient; based on the characteristic information and a preset association rule base, acquiring the supplementary information of the target patient under the condition of successful matching, and updating the characteristic information according to the supplementary information; And inputting the updated characteristic information into a pre-trained triage model, and outputting auxiliary decision information of the target patient, wherein the triage model is obtained based on historical triage data training. According to the diagnosis assisting decision-making method provided by the embodiment of the application, the diagnosis data comprise numerical physiological indexes and text type complaint data; The processing the visit data to extract characteristic information representing the condition of the target patient comprises the following steps: performing abnormal value detection and elimination processing on the numerical physiological indexes, and performing missing value filling processing on the data subjected to abnormal value elimination to obtain first characteristic sub-information; performing term consistency standardization processing on the text type main complaint data to obtain second characteristic sub-information; and forming the characteristic information through the first characteristic sub-information and the second characteristic sub-information. According to the diagnosis assisting decision-making method provided by the embodiment of the application, the term consistency standardization processing is carried out on the text type main complaint data to obtain second characteristic sub-information, and the method comprises the following steps: carrying out term consistency standardization processing on the text type main complaint data to obtain standardized main complaint data; And inputting the standardized complaint data into a pre-trained medical entity identification model, and outputting the identified symptoms, signs and disease entities as the second characteristic sub-information, wherein the medical entity identification model is constructed based on a two-way long-short-term memory network and a conditional random field framework. According to the diagnosis assisting decision-making method provided by the embodiment of the application, the feature information is formed by the first feature sub-information and the second feature sub-information, and the method comprises the following steps: combining the first characteristic sub-information and the second characteristic sub-information, and performing characteristic engineering processing on the combined data to generate the characteristic information; the feature engineering processing comprises the steps of carrying out segmentation discretization processing on the digital type feature and carrying out word frequency-inverse document freque