文献类型：会议论文

英文题名：An assistant decision-making method for rare diseases based on RNNs model

作者：Li, Qi[1,2];Jiang, Ming[3];Ying, Changtian[2]

机构：[1]Guangxi Acad Math Sci Technol Co Ltd, Nanning, Guangxi, Peoples R China;[2]Shaoxing Univ, Dept Comp Sci & Engn, Shaoxing, Zhejiang, Peoples R China;[3]Guangxi Acad Math Sci Technol Co Ltd, Nanning, Guangxi, Peoples R China

会议论文集：2022 International Conference on Bioinformatics and Biomedicine-BIBM-Annual

会议日期：DEC 06-08, 2022

会议地点：Las Vegas, NV

语种：英文

外文关键词：Rare Disease; Deep Learning; Decision-Making Assistance; Graph Neural Network

外文摘要：Big data and artificial intelligence technologies have great potential for development in the process of disease diagnosis and treatment. However, existing methods tend to focus on common diseases, but rarely involve the prediction of rare diseases. In this paper, we propose a rare disease-assisted decision-making method based on RNNs enhanced graph network. The model inferred rare disease probabilities based on patient complaints. Specifically, firstly, the overall syntax of the text is captured using a BERT-based pre-trained model and the local semantics are modeled using convolutional neural networks to identify patient intent and word slots. Secondly, the RNNs model is designed to capture the multi-hop node information around rare diseases, and a random walk algorithm with weights is used, which is finally transformed into a language model problem for prediction. Finally, in the experiment, the model achieves the current optimal effect in the real disease graph, which proves the effectiveness and advancement of the model.