Interpretable dynamic disease severity prediction method

Abstract

The invention discloses an interpretable dynamic disease severity prediction method. The method comprises the following steps: extracting SOFA scores, patient states and drug use information; according to the drug use information, linking to a UMLS standard term, and constructing a drug-related knowledge graph; embedding and dimensionality reduction are carried out on the drug-related knowledge graph to obtain embedding of a drug entity; determining the category i according to the SOFA change value at the current moment, and multiplying the patient state and the embedding of the drug entity by the ith row of the corresponding weight matrix; inputting the SOFA score, the patient state subjected to weight processing and the embedded and spliced time sequence data of the drug entity into a TCN prediction model, outputting a predicted SOFA score trend, and training and updating a weight matrix; and explaining the predicted SOFA score trend. According to the method, the change trend of the SOFA score can be more sensitive, the SOFA score can be predicted more accurately, and the prediction result can be explained.

Description

technical field [0001] The invention relates to the technical field of disease severity prediction, in particular to an interpretable dynamic disease severity prediction method. Background technique [0002] With the continuous development of database technology, hospitals have gradually collected and stored a large number of electronic medical records. How to carry out knowledge mining on this massive real data has gradually attracted the attention of researchers. Knowledge discovery and machine learning methods can be used both to discover new patterns in patient data and for classification and predictive purposes such as outcome or risk assessment. For intensive care unit (ICU) nurses, real-time disease severity is an important concern and key to saving patients' lives. If we can learn rich electronic medical record information and provide strong support for ICU clinical decision-making, it will be a great contribution to clinical practice. Since its development in the ...

AI Technical Summary

Problems solved by technology

At present, there are many prediction models of disease severity based on data mining of electronic medical records. However, these methods are not sensitive enough to the change trend of SOFA score, the prediction accuracy is not enough, and the prediction results are not explained enough.

Images