Calculation Model and System Implementation of Similarity Measurement of Multiple Electronic Medical Records Based on Deep Learning and Multi-Modal Abstract Extraction

With the wide application of Electronic Medical Records (EMR) in the medical field, how to effectively calculate the similarity between different medical records has become a key problem in medical data analysis.Traditional medical record similarity calculation methods based on rules or statistics have some problems, such as insufficient feature extraction and limited semantic understanding, which are difficult to meet the needs of complex medical data.This paper proposes a calculation model of medical record similarity measure based on deep learning and multi-modal abstract extraction, combining natural language processing (NLP) and Computer Vision (CV) technology to extract deep semantic features from multi-modal data such as text and image.In the text part, BERT (Bidirectional Encoder Representations from Transformers) is used for feature extraction, and Pointer Network is used for abstract extraction.In the image part, convolutional neural network (CNN) such as ResNet is used for feature extraction, and attention mechanism is used to optimize summary generation.Transformer is used for cross-modal feature fusion and medical record similarity calculation based on Siamese network.The experimental results show that the proposed method is superior to the traditional method in accuracy, F1-score and AUC, which proves its validity in the calculation of the similarity of electronic medical records.Finally, we implemented a complete medical record similarity computing system based on Django and PyTorch, which provided technical support for medical big data analysis and intelligent diagnosis support.