A Comprehensive Approach to Intelligent Cardiac Patient Screening Based on Electrocardiography Using Deep Neural Networks

Cardiovascular diseases remain a leading cause of global mortality, making early detection in pre-hospital settings critical. This paper explores the need for developing intelligent screening tools for cardiac patients, leveraging electrocardiography and artificial intelligence to support emergency medical technicians and general practitioners in underserved areas. The objective is to design a mobile application that rapidly analyzes electrocardiogram (ECG) data to classify patients into three categories (normal, suspicious, critical) and, in necessary cases, identify the specific cardiac condition. The study utilized a dataset of 704 ECG samples collected from medical centers in Arak, Markazi Province, Iran. The proposed methodology encompasses three approaches: (1) manual extraction of nine key ECG features combined with a multilayer perceptron (MLP) neural network, (2) ECG image analysis using convolutional neural network (CNN), and (3) processing of raw 10-second ECG signals (5,000 data points) with a hybrid CNN and Bidirectional Long Short-Term Memory (BiLSTM) model. The models achieved accuracies of 95% for the manual feature extraction approach, 97% for the ECG image analysis, and 98% for the raw signal processing. The study’s innovations include the development of a high-accuracy multi-task model optimized for mobile execution and a user-friendly interface tailored for non-specialist users. By incorporating automated preprocessing, noise filtering, and actionable next-step recommendations, this tool enables rapid and accurate screening in emergency settings, reducing mortality and enhancing pre-hospital care.