Perioperative period critical event prediction method based on cross-modal deep learning

Abstract

The invention relates to a perioperative period critical event prediction method based on cross-modal deep learning, and belongs to the field of artificial intelligence and medical application. The method comprises the following steps: step 1, constructing a multi-modal medical monitoring data set; 2, performing bimodal fusion feature learning on patient monitoring data and personalized data; 3, cross-modal collaborative learning feature extraction; 4, constructing a multi-modal critical event (death risk) prediction model; 5, verifying model feedback. The method serves as a critical adverse event prediction and early warning tool, and is an effective method for achieving real-time tracking, early diagnosis and early warning of main critical events after operation.

Description

technical field [0001] The invention belongs to the field of artificial intelligence and medical applications, and relates to a method for predicting perioperative critical events based on cross-modal deep learning. Background technique [0002] At present, various critical adverse events in the perioperative period in China are as high as 12%, leading to a mortality rate of up to 1.1% among hospitalized patients. Under the simple "harbinger score" critical event early warning system, the precursors of critical adverse events are often not caught in time, resulting in severe or late disease once a critical adverse event occurs, which is difficult to treat and the intervention effect is limited. Actively carry out research on tracking, early warning and intervention reminders of critical adverse events, which is helpful for early detection, early warning, diagnosis and intervention reminders, and has important scientific significance and social value. Some domestic and forei...

AI Technical Summary

Problems solved by technology

Some domestic and foreign institutions have carried out cutting-edge explorations on whether intraoperative direct monitoring data has the value of early diagnosis and early warning of critical adverse events after time-series data preprocessing analysis and data mining; although a lot of work has been done on the selection of monitoring indicators and warning thresholds, the system The start-up indicators of early warning intervention are still triggered by isolated thresholds, and have not integrated the clinical data of HIS / PACS / EMR, which began to be popularized at the same time. The necessary dynamic comprehensive judgment depends on the empirical judgment and analysis of individual medical staff

Images