Multi-label classification method for electronic medical records based on symptom extraction and feature representation

Abstract

The present invention provides a multi-label classification method for electronic medical records based on symptom extraction and feature representation. Considering the impact of diseases and symptoms and the relationship between symptoms on the multi-classification of disease labels in electronic medical records, two different symptom representations are used. Method: Use TF‑IDF to construct symptom vectors and word2vec to learn symptom vectors. Two kinds of symptom vector sequences extracted from electronic medical records are used as input sequences of the two bidirectional LSTM models respectively, and two bidirectional LSTM models are trained; for electronic medical records with unknown disease labels, the symptoms extracted from them correspond to two The symptom vector constitutes two kinds of symptom vector sequences, which are respectively input into two trained bidirectional LSTM models to obtain two probability vectors; the two probability vectors are weighted and combined to obtain the final classification vector. This method has good classification effect and applicability.

Description

technical field [0001] The invention belongs to the field of medical big data analysis, in particular to a multi-label classification method for electronic medical records based on symptom extraction and feature representation. Background technique [0002] The multi-label classification of electronic medical records (EMR) is an important task in the field of medical applications. Its purpose is to automatically generate disease labels for electronic medical records based on information such as symptoms, test indicators, drugs, and text in electronic medical records. , not only can save the cost of large-scale electronic medical record management and maintenance, but also provide convenience for medical knowledge mining and application. The multi-label classification based on electronic medical records can also be used in auxiliary diagnosis systems and hospital guidance systems, which can greatly improve doctors' work efficiency and shorten patients' visit time. The multi-...

AI Technical Summary

Problems solved by technology

The multi-label classification of electronic medical records relies on the features extracted from the medical record text. Currently, there are methods based on the entire text information, but there is a large amount of redundant information in the full text data that affects the classification effect; there are also inspection and detection indicators based on records in the text, Index information such as clinical data, medical codes, and drugs, but due to the lack of relevant information in some electronic medical records, these methods cannot be used

Images