ECG Classification Based on Multimodal Feature Extraction and Deep Learning Model

Heart health diagnosis plays a vital role in preventing and managing cardiovascular diseases, which remain a leading cause of death globally. Early and precise detection of heart related problems is vital for effective treatment and enhancing patient outcomes. Traditional methods often rely on single-modal data, which can limit diagnostic accuracy. In this study, we present a comprehensive approach to heart health analysis by extracting multimodal features derived from Electrocardiogram (ECG) signals and training a Deep Learning model. The methodology involves extraction of both time-domain features using Continuous Wavelet Transform: Morlet transform (CWT) and frequency-domain features using Fast Fourier Transform (FFT), enabling a thorough examination of the ECG data. These features were then input into a deep neural network model. The model achieved an impressive accuracy of 99.1% in classifying various heart conditions, highlighting its robustness and effectiveness. Notably, the neural network architecture employed is relatively simple, making it feasible to deploy on mobile or edge devices. This characteristic ensures that the heart health analysis can be performed in real-time without relying on extensive computational resources, thereby making it practical for widespread, on-device use. This approach not only enhances the accessibility of heart health monitoring but also promotes timely and accurate diagnosis in everyday settings.