CN-121983305-A - Medical multi-source cross-platform data real-time monitoring and intelligent risk early warning method

CN121983305ACN 121983305 ACN121983305 ACN 121983305ACN-121983305-A

Abstract

The invention relates to the technical field of medical information processing, in particular to a medical multi-source cross-platform data real-time monitoring and intelligent risk early warning method, which comprises the following steps of acquiring vital signs of a patient and converting medical term texts into embedded vectors through a plurality of medical institutions, generating an optimized medical term embedded set by adopting mutual information calculation and screening semantic contribution degree dimensions, realizing cross-platform alignment for different platform embedded vector coordinate axis rotations, executing time synchronization and smoothing processing to obtain a cross-platform joint monitoring data stream, inputting an intelligent risk judging model, calculating risk scores based on mutual information matching to generate real-time risk early warning state information.

Inventors

ZHAO JIAO
YIN CHUNYU
SONG GE
ZHAO XIAOYAN
LI XUEMEI

Assignees

中国人民解放军总医院第一医学中心

Dates

Publication Date: 20260505
Application Date: 20251230

Claims (10)

1. The medical multi-source cross-platform data real-time monitoring and intelligent risk early warning method is characterized by comprising the following steps of: s1, acquiring patient vital signs and medical term texts through a plurality of medical institutions, converting the texts into corresponding medical term embedding vectors, and generating a primary fusion medical data set by combining the patient vital signs; S2, calling the preliminary fusion medical data set, analyzing semantic information contribution degree of the medical term embedded vector by adopting a mutual information calculation method, screening the dimension of which the contribution degree meets a contribution threshold, marking heterogeneous dimensions among different platforms, and generating an optimized medical term embedded set; S3, according to the optimized medical term embedding set, rotating and adjusting coordinate axes of embedding vectors of different platforms, aligning a cross-platform embedding space by calculating a coordinate rotation angle, and outputting an aligned embedding vector data structure; S4, based on the alignment embedded vector data structure and vital signs of a patient, executing time synchronization processing, and performing smoothing processing on vital sign parameter data to obtain a cross-platform joint monitoring data stream; s5, calling the cross-platform joint monitoring data flow to input an intelligent risk discrimination model, comparing vital signs with a preset sign threshold value time by time, calculating a correlation risk score based on the statistical deviation degree, and generating real-time risk early warning state information.
2. The method for real-time monitoring and intelligent risk early warning of medical multi-source cross-platform data according to claim 1, wherein the primary fusion medical data set comprises a data unified identifier, a feature integration domain and a parameter association frame, the optimized medical term embedding set comprises a semantic screening domain, a weight distribution set and a structure labeling layer, the alignment embedding vector data structure comprises a coordinate correspondence table, a space reference frame and an alignment transformation domain, the cross-platform joint monitoring data flow comprises a time sequence continuous band, a parameter fusion channel and a joint observation track, and the real-time risk early warning state information comprises a risk indication level, a threshold comparison domain and an early warning trigger signal.
3. The medical multi-source cross-platform data real-time monitoring and intelligent risk early warning method according to claim 1, wherein the specific steps of S1 are as follows: s101, acquiring vital sign data frames of a patient through a plurality of medical institutions, performing interval discretization calculation on heart rate, blood pressure and body temperature values, performing serialization rearrangement according to vital sign acquisition time stamps, and performing aggregation and normalization calculation on rearranged numerical values according to time indexes to generate a vital sign sequence; S102, invoking the vital sign sequence, acquiring a medical term text through a medical institution, extracting medical term words, carrying out mapping calculation on the extracted character sequence and a medical term embedding vector, and carrying out alignment calculation on a mapping result and the vital sign sequence to obtain a term embedding alignment sequence; S103, based on the term embedded alignment sequence, invoking a normalization index in the vital sign sequence, performing weighted aggregation calculation on vectors of the two sequences at the same time index position, and integrating the aggregated vectors in time sequence to establish a primary fusion medical data set.
4. The medical multi-source cross-platform data real-time monitoring and intelligent risk early warning method according to claim 3, wherein the specific steps of S2 are as follows: S201, based on the medical term embedded vectors in the primary fusion medical data set, performing probability distribution quantization on the dimension and the marked value by adopting a mutual information calculation method aiming at the dimension and the term semantic marked value of each embedded vector, and performing joint distribution quantization and variance measurement to generate a semantic contribution matrix; S202, invoking the semantic contribution matrix, screening the dimension with the contribution value larger than the contribution threshold according to the comparison between the contribution value of each dimension and the contribution threshold, and aggregating the corresponding index to obtain a semantic contribution dimension set; S203, calling medical term embedded vectors in the primary fusion medical data set according to the semantic contribution dimension set, executing consistency marking, marking the value difference of the embedded vectors of different platform sources on the same dimension, binding the platform identification and the dimension index, and finally obtaining the optimized medical term embedded set.
5. The method for real-time monitoring and intelligent risk early warning of medical multi-source cross-platform data according to claim 4, wherein the contribution threshold is determined according to statistical features of medical term embedded vectors in a primary fusion medical dataset, data distribution rules and requirements of related tasks; and screening the dimensions with contribution values larger than the contribution threshold value, namely comparing the contribution value of each dimension with the contribution threshold value one by one, and if the contribution value is larger than the threshold value, incorporating the index of the dimension into the semantic contribution dimension set.
6. The method for real-time monitoring and intelligent risk early warning of medical multi-source cross-platform data according to claim 4, wherein the specific steps of S3 are as follows: S301, selecting medical term embedding vectors with consistent cross-platform semantics based on the optimized medical term embedding set, obtaining vector representations in an embedding space, calculating an included angle difference value between vectors, extracting rotation parameters, and generating a rotation parameter array; s302, calling the rotating parameter array and optimizing a medical term embedding set, performing angle correction based on each vector, adjusting a vector numerical sequence according to radian values in the rotating parameter array, and aggregating rotated vectors to obtain a rotating vector matrix; s303, calling the rotation vector matrix, performing dimension reconstruction on vector sequences, integrating adjacent sequences according to a position alignment rule, establishing an aligned vector set structure frame, and performing convergence processing on the vector to obtain an aligned embedded vector data structure.
7. The medical multi-source cross-platform data real-time monitoring and intelligent risk early warning method according to claim 6, wherein the specific steps of S4 are as follows: s401, based on the alignment embedded vector data structure and a time stamp sequence in vital signs of a patient, performing matching judgment operation, obtaining a time interval difference value, comparing the difference value with a preset synchronization threshold value, and if the difference value is lower than the threshold value, marking a synchronization time point to generate a time synchronization index matrix; S402, calling the time synchronization index matrix, traversing the time stamp position, obtaining a corresponding vital sign data frame, calculating adjacent parameter difference values, comparing the adjacent parameter difference values with a preset smooth reference value, and executing weighting operation when the difference values are lower than the reference value, and aggregating a plurality of parameter data to obtain a smooth parameter vector set; s403, performing cross-source aggregation operation on the parameter vectors according to the smooth parameter vector set, performing alignment embedded vector data structure mapping calculation, performing cascade combination on mapped data and original data, and performing data frame splicing to obtain a cross-platform joint monitoring data stream.
8. The method for real-time monitoring and intelligent risk early warning of medical multi-source cross-platform data according to claim 7, wherein the preset synchronization threshold value sets a synchronization limit according to a sampling period of a time stamp sequence, acquisition channel precision, time step span and synchronization link calibration error and according to a distinguishable interval of adjacent marks of the time stamps; the preset smoothing reference value is obtained by carrying out statistical fitting on amplitude distribution of adjacent parameter difference values according to the acquisition noise range, equipment precision, parameter fluctuation interval and variation gradient of vital sign parameters.
9. The method for real-time monitoring and intelligent risk early warning of medical multi-source cross-platform data according to claim 7, wherein the specific steps of S5 are as follows: S501, based on the cross-platform joint monitoring data flow, inputting an intelligent risk discrimination model to compare heart rate, blood pressure and body temperature values in vital signs with corresponding sign threshold values time by time, and carrying out discrimination operation according to the difference amplitude to generate a deviation matrix value; s502, invoking the deviation matrix value, calculating the similarity between a deviation sequence and a corresponding sign threshold sequence based on a distance method, and carrying out normalization processing according to a similarity result to obtain an associated risk parameter set; and S503, according to the associated risk parameter set, performing magnitude relation judgment on the risk parameters and the risk judgment reference value, screening the risk parameters with the magnitude larger than the risk judgment reference value, aggregating parameter indexes and encoding to generate real-time risk early warning state information.
10. The medical multi-source cross-platform data real-time monitoring and intelligent risk early warning method according to claim 9, wherein the sign threshold is obtained by collecting original health data of a target monitoring object and performing statistical analysis, a mean value and a standard deviation are calculated for the collected heart rate data, blood pressure data and body temperature data respectively, and an interval formed by adding and subtracting standard deviation multiples from the mean value is used as a sign threshold range of a corresponding vital sign; and the risk judging reference value is determined by carrying out quantile analysis on the deviation matrix value, extracting samples with the numerical values in the associated risk parameter set in a set percentile interval, and calculating a sample mean value as the risk judging reference value.

Description

Medical multi-source cross-platform data real-time monitoring and intelligent risk early warning method Technical Field The invention relates to the technical field of medical information processing, in particular to a medical multi-source cross-platform data real-time monitoring and intelligent risk early warning method. Background The technical field of medical information processing relates to the acquisition, processing, storage, analysis and application of medical data. The technical field covers various technical means and equipment, and is used for improving the efficiency and the accuracy of the medical industry, in particular to the aspects of disease diagnosis, patient monitoring, treatment scheme design and the like. The core matters comprise standardized processing of medical data, cross-platform data sharing and integration, a real-time monitoring technology, an intelligent risk early warning system and safety protection of medical information. With the continuous development of information technology, the field of medical information processing focuses not only on basic management of medical data, but also on improvement of accuracy and instantaneity of a clinical decision support system through intelligent analysis. The traditional medical multi-source cross-platform data real-time monitoring and intelligent risk early warning method is used for carrying out real-time monitoring and analysis by collecting diversified data from different medical equipment, sensors and systems and carrying out intelligent risk early warning based on the data. The method relies on a data integration technology, can effectively gather data from different sources, and utilizes big data analysis and machine learning technology to monitor and predict potential risks in real time for the state of a patient. The traditional method mainly adopts a rule-based monitoring system and a simple threshold alarming mechanism, and lacks intelligent analysis and adaptability early warning for complex disease modes. The prior art relies on a rule-based monitoring system and a simple threshold alarm mechanism, and is difficult to process complex disease modes and individual differences, so that early warning lacks intelligent adaptability. The data integration capability is limited, and potential risks in multi-source data, especially effective fusion and analysis of cross-platform data, cannot be fully exploited. The traditional method cannot fully utilize the deep semantics of medical terms, so that the data matching and integration effects among different platforms are poor, the real-time performance and accuracy of a monitoring system are insufficient, personalized clinical decisions cannot be effectively supported, misjudgment or delay is easy to occur, and the timeliness and accuracy of an early warning system are affected. Disclosure of Invention In order to solve the problems that the prior art relies on a rule-based monitoring system and a simple threshold alarm mechanism, complex disease modes and individual differences are difficult to process, and intelligent adaptability of early warning is lacking. The data integration capability is limited, and potential risks in multi-source data, especially effective fusion and analysis of cross-platform data, cannot be fully exploited. The traditional method cannot fully utilize deep semantics of medical terms, so that data matching and integration effects among different platforms are poor, real-time performance and accuracy of a monitoring system are insufficient, personalized clinical decisions cannot be effectively supported, misjudgment or delay is easy to occur, and timeliness and accuracy of an early warning system are affected. In order to achieve the purpose, the invention adopts a medical multi-source cross-platform data real-time monitoring and intelligent risk early warning method, which comprises the following steps: s1, acquiring patient vital signs and medical term texts through a plurality of medical institutions, converting the texts into corresponding medical term embedding vectors, and generating a primary fusion medical data set by combining the patient vital signs; S2, calling the preliminary fusion medical data set, analyzing semantic information contribution degree of the medical term embedded vector by adopting a mutual information calculation method, screening the dimension of which the contribution degree meets a contribution threshold, marking heterogeneous dimensions among different platforms, and generating an optimized medical term embedded set; S3, according to the optimized medical term embedding set, rotating and adjusting coordinate axes of embedding vectors of different platforms, aligning a cross-platform embedding space by calculating a coordinate rotation angle, and outputting an aligned embedding vector data structure; S4, based on the alignment embedded vector data structure and vital signs of a patient, executing time synchronization proce