CardiacRT-NN: Real-Time Detection of Cardiovascular Disease Using Self-attention CNN-LSTM for Embedded Systems

Yixin Li, Ning Sui, Anil Gehi, Chengan Guo, Zhishan Guo

Abstract: Deep learning (DL) has emerged as a critical technology in the advancement of non-invasive cardiac monitoring by analyzing electrocardiogram (ECG) data. Although traditional approaches utilizing up to 12-lead ECGs have been effective, their practicality for real-time monitoring during daily activities is constrained. In contrast, single-lead ECG data lacks the comprehensive diagnostic information necessary for precise assessment. To overcome these challenges, this research introduces a novel deep neural network architecture, CardiacRT-NN. This model integrates a self-attention-based convolutional neural network (CNN) with long short-term memory (LSTM) components to accurately classify various heart diseases from the mean of customized ECG patterns. The experiment for real-time diagnosis has been validated on Raspberry Pi 4 to ensure the balance of accuracy and efficiency. This research lays a foundational step towards interpretable AI in healthcare, with the aim of facilitating proactive detection and management of potential cardiac issues, ultimately striving to mitigate the risk of sudden cardiac death.