Research On Cardiovascular Disease Prediction Model Based On Multimodal Features

With continuous development and progress of society,the living standards have been constantly improving.Consequently,cardiovascular disease has become one of the most serious health issues worldwide.Many scholars are working hard to develop a more accurate,reliable,and effective new technology for the prevention and diagnosis of cardiovascular diseases.Through in-depth research,it has been found that the existing cardiovascular disease prediction models mainly have issues such as poor predictive performance,weak generalization capabilities,and most research is only based on single-modal data modeling,lacking research on multi-modal data.The purpose of this study is to improve the accuracy and reliability of cardiovascular disease prediction,as well as to explore the algorithms and improvement needed to achieve this goal,and to attempt a prediction method for cardiovascular diseases based on multimodal data.The research work of this thesis includes the following two aspects:(1)Aiming at the poor predictive performance and inadequate generation ability of traditional predicition models,this thesis proposed a new cardiovascular disease prediction model called XGBLR-IMBODE,which is constructed by combing the Stacking ensemble method with model hyperparameter optimization based on IMBODE.Experimental results demonstrate that the model exhibits excellent performance on multiple public datasets,indicating its strong generalization ability and the value and effectiveness of cardiovascular disease prediction.(2)Aiming at the shortcomings of traditional attention models,such as the requirement for high computational resources,poor performance on small datasets,the difficulty in obtaining image location information,and the difficulty in capturing longterm depencdecies in the features learned by the convolutional layers of neural networks.This thesis proposed an Efficient-ECGNet model based on the LKA module and lightweight ParC convolution.The model demonstates excellent performance in predicting myocardial infarction based on 12-lead ECG images,while maintaining a low parameter count and high accuracy.Additionaly,a myocardial infarction predciton method based on multimodal features.Experimental results demonstrate that the prediction accuracy is further improved by using multimodal features,thus proving the feasibility and effectiveness of mutlmodal experiments.