Ensemble model for ECG Classification

Electrocardiography (ECG) is a relatively popular diagnostic technique that does not involve any kind of bodily invasion, and algorithms are increasingly supporting its interpretation. There has been a shortage of relevant datasets for training in this area of work and there will be lack of clearly defined evaluation procedure, which has slowed progress in the field of automated analysis of electrocardiograms (ECGs). These factors have made it difficult to compare the effectiveness of various algorithms. PTB-XL is a clinical 12-lead electrocardiogram dataset that was just released and is easily available. Here, we illustrate a range of results for this dataset., which covers a classification job based on a variety of ECG statements. This is done so that we may get around the issues that have been brought up. Among the deep learning-based time series classification approaches that were researched and examined, convolutional neural networks, and especially vgg16 and alexnet-based architectures, provide the greatest performance. The vgg16-based convolutional method achieves 0.8710 accuracy and 0.9448 AUC when applied to the ECG dataset. The ensemble model achieves the accuracy of 0.8690 and AUC of 0.9431. In the PTB-XL dataset, we used a 100 Hz sample rate for our training purposes. We split out the PTB-XL dataset into training, which is approx. 70% of the data, testing, which occupies 15% of the dataset, and the remaining 15% is for the validation set. The findings demonstrate that the strategy presented here is not only effective but also accurate when it comes to classifying ECG signals.