CN-122025098-A - Medical data processing method, system, storage medium and computer program for predicting acute aortic dissection postoperative extension mechanical ventilation risk

CN122025098ACN 122025098 ACN122025098 ACN 122025098ACN-122025098-A

Abstract

The invention provides a medical data processing method, a system, a storage medium and a computer program for predicting the risk of prolonged mechanical ventilation after acute A-type aortic dissection, and relates to the technical field of medical data processing. The method comprises the steps of firstly cleaning and extracting features of perioperative multi-source heterogeneous data based on a specific physiological index set, and constructing a standardized parameter feature vector. And secondly, discretizing and scoring the numerical characteristics after verification according to a preset threshold value, and generating a comprehensive risk score through an accumulated calculation model. And thirdly, generating a target probability value by using the probability mapping table to look up a table and establishing a risk layering label. Finally, an audit data packet containing the input snapshot and the calculation logic is assembled, encrypted and stored to form a non-tamperable risk assessment record, and a visual risk overview and a attribution list are generated according to the risk assessment record. The invention solves the problems of lack of specificity, nonstandard multi-source data processing and non-traceability of the calculation process in the prior art.

Inventors

ZOU LIANG
MA WENTAO
CHEN CHEN
FU RUI
ZHOU XIAOMING
LIANG YANRU
ZHANG QIN
Zhen Kaiyuan

Assignees

中国医学科学院阜外医院

Dates

Publication Date: 20260512
Application Date: 20260130

Claims (10)

1. A medical data processing method for predicting the risk of prolonged mechanical ventilation after acute aortic dissection a, comprising: cleaning and extracting the characteristics of multi-source heterogeneous data of a patient in the perioperative period based on a preset specific physiological index set to obtain standardized parameter characteristic vectors with five key dimensions, and extracting the numerical characteristics and corresponding unit attributes of all the key dimensions; Determining a validity check result of each dimension according to the numerical characteristics and the corresponding unit attributes, acquiring a preset classification threshold set, and determining a discretization risk weight according to the classification threshold set and the numerical characteristics; Generating a risk sub-scoring sequence based on the discretization risk weight in a parameter mapping space of each dimension, establishing an accumulation calculation model for the risk sub-scoring sequence, and generating a comprehensive risk score according to an integer summation rule; Constructing a probability mapping table on the value fields of all the comprehensive risk scores, and generating a target probability value in the probability mapping table by adopting a table look-up indexing algorithm to cover all possible scoring results, wherein the key value of the probability mapping table is an integer total score of 0 to 9; Sequentially generating risk layering labels according to the target probability values, assembling an audit data packet containing an input snapshot and calculation logic, and performing time stamp marking and encryption storage on the audit data packet to obtain a risk assessment record; And carrying out clinical decision support according to the risk assessment record to obtain a final risk prediction result.
2. The method for processing medical data for predicting an acute aortic dissection type a postoperative prolonged mechanical ventilation risk according to claim 1, wherein the steps of cleaning and feature extraction of multi-source heterogeneous data of a perioperative patient based on a preset specific physiological index set to obtain a standardized parameter feature vector comprising five key dimensions, extracting numerical features and corresponding unit attributes of all the key dimensions comprise: defining the specific physiological index set to comprise five dimensions of age, body mass index, preoperative white blood cell count, preoperative platelet count and extracorporeal circulation time, and constructing a field mapping protocol for connecting the five dimensions with source fields in the multi-source heterogeneous data; Calling a data grabbing interface, traversing the multi-source heterogeneous data according to the field mapping protocol, and positioning and extracting original data fragments corresponding to each dimension through the unique index path; Performing character segmentation on each original data segment by using a preset regularization analysis rule to separate a pure digital part as the numerical characteristic and a measurement character part as the unit attribute; and arranging all the extracted numerical features according to a preset dimension sequence of the specific physiological index set, generating the standardized parameter feature vector in five dimensions, and storing the corresponding unit attribute association into metadata of the standardized parameter feature vector.
3. The medical data processing method for predicting an acute type-a aortic dissection post-operation prolonged mechanical ventilation risk according to claim 1, wherein determining a validity check result of each dimension according to the numerical feature and the corresponding unit attribute, and obtaining a preset classification threshold set, determining a discretization risk weight according to the classification threshold set and the numerical feature, comprises: calling a preset standard unit library, and comparing the unit attribute with the standard units corresponding to the five key dimensions, wherein if the comparison result is inconsistent, the numerical value characteristics are normalized and converted by using a preset conversion coefficient; then verifying whether the converted numerical characteristics fall within a preset physiological effective range, if so, marking the numerical characteristics as effective input values, otherwise, triggering an exception handling logic to generate a numerical set to be evaluated; The preset two-classification threshold value set is read, and the five key dimensions are divided into a positive correlation group and a negative correlation group according to pathological correlation, wherein the positive correlation group comprises age, body mass index, white blood cell count and extracorporeal circulation time; Based on the dimension of the to-be-evaluated numerical value set belonging to the forward correlation group, when the numerical value is larger than the corresponding classification threshold, giving a preset non-zero risk weight, otherwise giving zero weight; based on the dimensionality belonging to the negative correlation group in the to-be-evaluated numerical value set, when the numerical value is smaller than or equal to the corresponding classification threshold value, giving a preset non-zero risk weight, otherwise giving zero weight; and generating the discretization risk weights corresponding to the five key dimensions one by one based on the scoring result.
4. A medical data processing method for predicting risk of acute aortic dissection type a post-operative extension mechanical ventilation according to claim 1, wherein generating a risk sub-score sequence based on the discretized risk weights in a parameter mapping space of each dimension, and establishing an accumulated computation model for the risk sub-score sequence, generating a comprehensive risk score according to an integer summation rule, comprises: mapping the discretization risk weights corresponding to the five key dimensions into a one-dimensional array space according to a preset arrangement sequence of the five key dimensions to construct a risk sub-scoring sequence, wherein each element position in the risk sub-scoring sequence corresponds to a physical meaning of one key dimension; Constructing an accumulation calculation model, and taking the constructed risk sub-scoring sequence as an input accumulation calculation model; acquiring an output value of the accumulation calculation model, and verifying whether the output value falls into a preset closed interval integer range of 0 to 9; after validation, the output value is instantiated as the composite risk score.
5. The method for medical data processing for predicting acute aortic dissection type a postoperative prolonged mechanical ventilation risk according to claim 1, wherein constructing a probability mapping table on all the value ranges of the comprehensive risk scores and generating target probability values in the probability mapping table by adopting a table look-up indexing algorithm to cover all possible scoring results comprises: Initializing a read-only key value database as the probability mapping table, and preloading ten groups of discrete key value mapping relations in the table; invoking the table lookup index algorithm, taking the comprehensive risk score as a unique retrieval key value, and inputting the unique retrieval key value into a query interface of the probability mapping table, wherein the table lookup index algorithm executes hash direct addressing, and locks a storage address which is completely matched with the retrieval key value in the key domain; And extracting the corresponding experience probability constant from the storage address, assigning the experience probability constant to a temporary variable, and formally instantiating the temporary variable into the target probability value after checking non-null so as to complete nonlinear conversion from integer score to percentage probability.
6. The medical data processing method for predicting an acute type-a aortic dissection post-operation extension mechanical ventilation risk according to claim 1, wherein the sequentially generating risk stratification labels according to the target probability value, assembling an audit data packet including an input snapshot and calculation logic, and performing time-stamping and encryption storage on the audit data packet to obtain a risk assessment record comprises: When the comprehensive risk score falls into a preset low-score section, generating a low-risk text label, and when the comprehensive risk score falls into a preset high-score section, generating a high-risk text label so as to establish the risk layering label; Constructing a container of a structured audit data packet, performing serialization operation, and encapsulating three types of data objects into the container, wherein the three types of data objects comprise the standardized parameter feature vector and an original unit attribute thereof; acquiring the current system time as a time stamp mark, and adding the time stamp mark into a metadata head of the audit data packet; And calling a preset hash algorithm or an asymmetric encryption algorithm to carry out digital signature processing on the audit data packet added with the time stamp, generating a ciphertext data stream with a non-falsifiable attribute, and writing the ciphertext data stream into a read-only database to obtain the risk assessment record.
7. A medical data processing method for predicting risk of acute aortic dissection type a postoperative prolonged mechanical ventilation according to claim 1, wherein the clinical decision support is performed according to the risk assessment record to obtain a final risk prediction result, comprising: invoking a decryption algorithm to analyze the risk assessment record, and restoring the comprehensive risk score, the target probability value and the risk layering label which are packaged in the risk assessment record; Invoking a graphic user interface rendering engine, mapping the comprehensive risk score into a dashboard numerical control, mapping the target probability value into a trend chart control, and mapping the risk layering label into a status indicator control with a preset warning color value so as to generate a visual risk overview on an interactive interface; Extracting the standardized parameter feature vector and the original numerical value thereof from the input snapshot of the risk assessment record, and reversely reconstructing the score contribution degree of each dimension by combining with the calculation logic identification in the audit data packet; dynamically generating a risk attribution list on the interactive interface, visually displaying contribution weights of age, body mass index, white blood cell count, platelet count and extracorporeal circulation time to the comprehensive risk score, and providing an explanatory basis for clinical decision; Executing a hierarchical response strategy according to the risk hierarchical label, and triggering a modal popup window or a highlight alarm on the interactive interface when the label is high risk; And packaging and sending the visual risk overview and the risk attribution list to a main index database of a hospital information system or an electronic medical record system through a standard medical data exchange interface so as to obtain system-level synchronization and lasting archiving of a final risk prediction result.
8. A medical data processing system for predicting risk of prolonged mechanical ventilation after acute aortic dissection a, comprising: The data cleaning and feature extraction module is used for cleaning and feature extraction of multi-source heterogeneous data of a patient in the perioperative period based on a preset specific physiological index set to obtain standardized parameter feature vectors with five key dimensions, and extracting numerical features and corresponding unit attributes of all the key dimensions; The verification and weight determination module is used for determining a validity verification result of each dimension according to the numerical value characteristic and the corresponding unit attribute, acquiring a preset classification threshold set and determining a discretization risk weight according to the classification threshold set and the numerical value characteristic; The scoring sequence generating and accumulating module is used for generating a risk sub-scoring sequence based on the discretization risk weight in the parameter mapping space of each dimension, establishing an accumulating calculation model for the risk sub-scoring sequence and generating a comprehensive risk score according to an integer summation rule; The probability mapping and indexing module is used for constructing a probability mapping table on the value ranges of all the comprehensive risk scores, and generating target probability values in the probability mapping table by adopting a table look-up indexing algorithm so as to cover all possible scoring results, wherein the key value of the probability mapping table is an integer total score from 0 to 9; the risk layering and auditing storage module is used for sequentially generating risk layering labels according to the target probability value, assembling an auditing data packet containing an input snapshot and calculation logic, and performing time stamp marking and encryption storage on the auditing data packet to obtain a risk assessment record; and the decision support and result output module is used for carrying out clinical decision support according to the risk assessment record to obtain a final risk prediction result.
9. A computer-readable storage medium storing computer instructions for causing a computer to execute the medical data processing method according to any one of claims 1 to 7.
10. A computer program for executing the medical data processing method according to any one of claims 1 to 7.

Description

Medical data processing method, system, storage medium and computer program for predicting acute aortic dissection postoperative extension mechanical ventilation risk Technical Field The invention relates to the technical field of medical data processing, in particular to a medical data processing method, a system, a storage medium and a computer program for predicting the risk of prolonged mechanical ventilation after acute A-aortic dissection. Background Acute aortic dissection type a (ATAAD) is a very high risk of cardiovascular acute severe disease and patients often need mechanical ventilation treatment in the intensive care unit after surgical repair. Prolonged Mechanical Ventilation (PMV) not only significantly increases medical resource consumption, but is also closely related to postoperative infections and high mortality. At present, clinical evaluation of ATAAD postoperative PMV risk mainly depends on subjective experience judgment of a clinician, or is directly predicted by using general intensive care scoring systems such as APACHE II (acute physiology and chronic health score), SOFA (sequential organ failure score) and the like. However, the above prior art has significant technical drawbacks in practical applications. Firstly, the general scoring system comprises a large number of nonspecific physiological parameters (such as arterial blood gas analysis, complex biochemical indexes and the like), the data dimension is high, the acquisition is delayed, the evaluation is difficult to complete immediately after the operation, the models are not optimized for a special pathological mechanism of ATAAD deep low-temperature stop circulation of the operation, and the prediction accuracy is not sufficient. Secondly, in the engineering deployment level, the existing scheme lacks standardized data cleaning, unit normalization and outlier verification mechanisms in the face of multi-source heterogeneous data from the HIS, the LIS and the anesthesia system, so that the data quality is poor, and the model operation robustness is poor. Finally, the existing calculation model mostly adopts a complex regression formula or a 'black box' algorithm, lacks logic for converting continuous variables into clinically readable discretization weights, and lacks audit trails in the calculation process, so that evaluation results cannot be traced and verified, and compliance requirements of medical data processing are difficult to meet. Disclosure of Invention In order to overcome the defects of the prior art, the invention aims to provide a medical data processing method, a system, a storage medium and a computer program for predicting the risk of prolonged mechanical ventilation after acute A aortic dissection operation, and solves the problems that evaluation indexes lack of specificity, multi-source heterogeneous data processing lack of a standardized verification mechanism and a calculation process are not traceable in the prior art. In order to achieve the above object, the present invention provides the following solutions: a medical data processing method for predicting risk of prolonged mechanical ventilation after acute aortic dissection a, comprising: cleaning and extracting the characteristics of multi-source heterogeneous data of a patient in the perioperative period based on a preset specific physiological index set to obtain standardized parameter characteristic vectors with five key dimensions, and extracting the numerical characteristics and corresponding unit attributes of all the key dimensions; Determining a validity check result of each dimension according to the numerical characteristics and the corresponding unit attributes, acquiring a preset classification threshold set, and determining a discretization risk weight according to the classification threshold set and the numerical characteristics; Generating a risk sub-scoring sequence based on the discretization risk weight in a parameter mapping space of each dimension, establishing an accumulation calculation model for the risk sub-scoring sequence, and generating a comprehensive risk score according to an integer summation rule; Constructing a probability mapping table on the value fields of all the comprehensive risk scores, and generating a target probability value in the probability mapping table by adopting a table look-up indexing algorithm to cover all possible scoring results, wherein the key value of the probability mapping table is an integer total score of 0 to 9; Sequentially generating risk layering labels according to the target probability values, assembling an audit data packet containing an input snapshot and calculation logic, and performing time stamp marking and encryption storage on the audit data packet to obtain a risk assessment record; And carrying out clinical decision support according to the risk assessment record to obtain a final risk prediction result. A medical data processing system for predicting risk of prolonge