基于加权贝叶斯网络的心脏破裂风险预测模型  

Risk Prediction Model of Cardiac Rupture Based on Weighted Bayesian Network

作　　者：刘初阳 杨湘 陈艳红[2] LIU Chuyang;YANG Xiang;CHEN Yanhong(Wuhan University of Science and Technology,Wuhan 430065;Wuhan Asia Heart Hospital,Wuhan 430022)

机构地区：[1]武汉科技大学,武汉430065 [2]武汉亚洲心脏病医院,武汉430022

出　　处：《计算机与数字工程》2025年第3期684-691,共8页Computer & Digital Engineering

基　　金：国家自然科学基金项目(编号:U1836118)资助。

摘　　要：心肌梗死后的心脏破裂作为急性心梗中最为致命的并发症,对其的预测和干预尤为重要。由于心脏破裂属于高致死率的罕见疾病,其数据集存在不平衡性和数据缺失性,这使深度模型很难达到较高的准确率,同时模型的结果还需要是可解释的。为了解决以上问题,论文提出了一种基于注意力机制的加权贝叶斯网络模型。该模型通过将医学知识和算法相结合的方式构建出更加准确的网络结构,其次将注意力权重融入贝叶斯网络,可以对显著性指标给予更多关注,增强了模型的准确性和可解释性。最后在真实数据上,对急性心肌梗死后的心脏破裂风险进行评估,实验结果证明,该模型的准确性和可解释性更加优秀,其F1分数和AUC值可以分别达到0.7718和0.7987。As the most fatal complication of acute myocardial infarction,the prediction and intervention of cardiac rupture after myocardial infarction is particularly important.Because heart rupture is a rare disease with high mortality,its data set there is imbalance and data deletion,which makes it difficult for the deep learning model to achieve high accuracy,and the results of the model need to be interpretable.In order to solve the above problems,this paper proposes a weighted Bayesian network model based on attention mechanism.The model builds a more accurate network structure by combining medical knowledge and algorithms.Secondly,by integrating attention weight into Bayesian network,more attention can be paid to significance indicators,and the accuracy and interpretability of the model are enhanced.Finally,on the real data,the risk of cardiac rupture after acute myocardial infarction is evaluated.The experimental results show that the accuracy and interpretability of the model are better,and its F1 score and AUC value can reach 0.7718 and 0.7987 respectively.

关 键 词：疾病分类 电子健康记录 注意力机制 贝叶斯网络 心脏破裂 

分 类 号：O141.4[理学—数学]