A method and system for real-time prediction of intensive care state based on mask dynamic composition

By constructing a dynamic relationship graph and an adaptive fusion gate in the intensive care unit, the problem of poor prediction performance of sparse data is solved, and multi-scale feature extraction and efficient prediction of patient status are achieved.

CN121839192BActive Publication Date: 2026-06-26JIMEI UNIV
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
JIMEI UNIV
Filing Date
2026-03-13
Publication Date
2026-06-26

AI Technical Summary

Technical Problem

Existing technologies cannot effectively utilize clinical monitoring behavioral information, resulting in poor predictive performance of sparse data, especially in intensive care settings where accurate prediction of future trajectories is difficult to achieve.

Method used

A mask-based dynamic graphing method is adopted to construct a dynamic relationship graph by monitoring the identifier matrix, extract global temporal dependency features and local pattern features, and integrate them using an adaptive weighted fusion gate to directly output the future prediction sequence.

Benefits of technology

It significantly improves the accuracy of intensive care status prediction under sparse data, avoids the risk of overfitting, and achieves multi-scale feature characterization and efficient prediction of patient status.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121839192B_ABST
    Figure CN121839192B_ABST
Patent Text Reader

Abstract

The application discloses a kind of intensive care state real-time prediction method and system based on mask dynamic composition.It directly faces the challenge of extreme sparsity of clinical time series data, and creatively converts the monitoring missing pattern into valuable prediction features. By constructing a monitoring pattern dynamic relationship diagram, the clinical collaborative monitoring strength between indicators is encoded; combined with time series attention mechanism and multi-level local feature extraction, time series dependence is modeled comprehensively; finally, through adaptive multi-channel fusion, the future clinical trajectory of multiple indicators of patients is accurately and end-to-end predicted. Compared with traditional filling or special architecture method, the present application can more effectively mine the knowledge hidden in clinical monitoring behavior, and can achieve superior prediction performance on the intensive care state data set with extremely sparse data, providing a new technical path for the development of intelligent clinical early warning system.
Need to check novelty before this filing date? Find Prior Art