Arrhythmia Detection from ECG Traces Images Using Transfer Learning Approach

The increasing prevalence of cardiovascular diseases necessitates efficient and accurate arrhythmia detection methods. Transfer learning offers a powerful technique by leveraging pre-trained neural networks on large image datasets for related tasks. This study proposes an approach for arrhythmia detection from electrocardiogram (ECG) traces utilizing a transfer learning methodology. In this research, pre-trained convolutional neural network architecture is adapted for arrhythmia detection. The Electrocardiogram trace image dataset can be easily obtained using smartphones, making it readily available even in low-resource countries. Hence in this study dataset of labeled traces having five different classes of arrhythmia patterns is used. Data augmentation techniques are applied to enhance the model's ability to generalize to diverse electrocardiogram patterns. The proposed transfer learning model demonstrates good results in terms of accuracy, precision during evaluation on a separate test set with respective convolutional neural network model. The study provides comparative insights into the effectiveness of transfer learning in medical image analysis tasks, particularly in the context of arrhythmia detection.