Dual-stream framework for image-based heart infarction detection using convolutional neural networks

Heart infarction has become one of the major causes of global death in recent decades. As the aging society intensifies, many elderly people living alone are facing life-threatening situations brought on by sudden heart attacks or infarctions. Researchers have been investigating vision-based, economical, and effective ways to monitor health status in real time in recent years. Motivated by this, this work presents a dual-stream network framework that takes into account both facial features and general whole image features and employs the feature interaction strategies in the framework structure. Furthermore, to cope with the lack of relevant datasets, we propose a novel image dataset for the detection of chest pain and falls which are the significant vital signs of myocardial infarction. With an accuracy of 88.98 % , our approach validates the viability of identifying cases of myocardial infarction in practical situations. Numerous experiments demonstrate our proposed framework’s effectiveness and its extensibility to be deployed on mobile smart platforms.