Search

Detection of Premature Ventricular Contractions (PVC) is of crucial importance in the cardiology field, not only to improve the health system but also to reduce the workload of experts who analyze electrocardiograms (ECG) manually. PVC is a non-harmful common occurrence represented by extra heartbeats, whose diagnosis is not always easily identifiable, especially when done by long-term manual ECG analysis. In some cases, it may lead to disastrous consequences when associated with other pathologies. This work introduces an approach to identify PVCs using machine learning techniques without feature extraction and cross-validation techniques. In particular, a group of six classifiers has been used: Decision Tree, Random Forest, Long-Short Term Memory (LSTM), Bidirectional LSTM, ResNet-18, MobileNetv2, and ShuffleNet. Two types of experiments have been performed on data extracted from the MIT-BIH Arrhythmia database: (i) the original dataset and (ii) the balanced dataset. MobileNetv2 came in first in both experiments with high performance and promising results for PVCs’ final diagnosis. The final results showed 99.90% of accuracy in the first experiment and 99.00% in the second one, despite no feature detection techniques were used. The approach we used, which was focused on classification without using feature extraction and cross-validation techniques, allowed us to provide excellent performance and obtain better results. Finally, this research defines as first step toward understanding the explanations for deep learning models’ incorrect classifications.

Electrocardiogram (ECG) analysis has been used to identify different heart problems and deep learning is emerging as a common tool to analyse ECGs. Premature ventricular contraction (PVC) is the most common cause of abnormal heartbeats; in most cases this is harmless but under specific conditions, it can lead to a life-threatening cardiac disease. Automated PVC detection in this scenario is a task of significant importance for relieving the heavy workloads of experts in the manual analysis of long-term ECGs. To identify PVCs, this research aims to use the MIT-BIH Arrhythmia Database to classify QRS complexes using five different deep neural networks: Long Short Term Memory, AlexNet, GoogleNet, Inception V3 and ResNet-50. The results showed high efficiency and reliability in the final diagnoses during two separate experiments (one with the entire dataset and the other with a balanced dataset). The ResNet-50 was the first experiment's best classifier (accuracy = 99.8%, F1-score = 99.2%), and the second experiment's best classifier was Inception V3 (accuracy = 98.8%, F1-score=98.8%). Relevant information, in this research, was extrapolated from a study of the confusion matrix to conduct a “failure analysis” to understand where and why the classifiers made incorrect classifications.