Research On The Construction Of Epilepsy Medical Knowledge Graph Based On Electronic Medical Records

Since 2015,the second Monday of February every year is World Epilepsy Day.In China,according to the latest epidemiology,epilepsy has become the second most common neurological disease after headache.With the promotion of hospital information systems and the continuous advancement of biological high-throughput technologies,the combination of artificial intelligence and medicine has become the general trend.How to effectively organize and use the electronic medical records of epilepsy patients,quickly and accurately obtain effective information and discover new knowledge from these massive text materials,will greatly promote the progress of medical research and achieve major breakthroughs.This research combines epilepsy electronic medical records with knowledge graphs,researches on named entity recognition and relationship extraction during the construction of knowledge graphs,and implements the construction and visualization of epilepsy medical knowledge graphs based on Neo4 j.The main work is as follows:1.The CNN-Bi LSTM-CRF model is proposed to identify named entities for the electronic medical records of epilepsy patients.In this study,the structure and language characteristics of the epilepsy electronic medical record are analyzed in detail,and named entity recognition is used as a kind of sequence labeling to construct a epilepsy electronic medical record corpus.On the basis of the BiLSTM-CRF benchmark model,CNN is introduced to extract local features of the vectorized sentence matrix of the text,to capture the features between multiple consecutive words,share the weights in the same type of features,and learn the abstract in the training process.Spatial characteristics.Through comparative experiments,the results show that the model proposed in this study is superior to other benchmark models in named entity recognition.2.The BiGRU-ATT model is proposed to extract the medical entities in the electronic medical records of epilepsy patients.This study analyzes the medical entity relationships in the epilepsy electronic medical record,and on the basis of drawing on existing research,defines seven medical relationships among them and builds a relationship extraction corpus.Based on the BiGRU model,the self-attention mechanism is introduced to focus on the key semantic information in the classification task.Through the calculation of each word and each word in the sequence,the potential connections between the words are tapped to improve the classification ability of the model.Comparative experimental results show that the model proposed in this study can achieve better results in the relationship extraction task.3.Realize the construction and visual display of epilepsy medical knowledge graph.In this study,the 5 medical entities and 7 medical relationships identified were used to generate structured files,which were imported into the Neo4 j graph database.The powerful storage,retrieval and processing capabilities of the graph database were used to realize the data visualization of the knowledge graph.This study conducted a comprehensive analysis of the electronic medical records of epilepsy patients,and used this as a data source to perform medical entity identification and relationship extraction to achieve visualization based on Neo4 j epilepsy medical knowledge graph.