Key result
Deep learning with a BILSTM layer using STFT and DCT feature extraction achieved 99.1% testing accuracy for classifying ECG signals as normal or atrial fibrillation.
Why the study?
Physicians rely on ECG to evaluate cardiac electrical activity, motivating the development of new feature extraction methods to classify ECG signals into normal and AF.
Does Deep Learning with BILSTM improve the classification accuracy of normal vs Atrial Fibrillation ECG signals compared to other classifiers?
Does Deep Learning with BILSTM improve the classification accuracy of normal vs Atrial Fibrillation ECG signals compared to other classifiers?
A novel feature extraction and deep learning algorithm using BILSTM achieved over 99% accuracy in classifying normal versus atrial fibrillation ECG signals.
Physicians use ECG to evaluate the electrical activity of patients’ hearts to know whether their hearts working effectively or not. In this paper, ECG is classified into normal andAtrial Fibrillation AF patients using new methods of extracting features from ECG signals. Extracted features from ECG signals are conducted as follows: in the first step ECG signals are normalized and detrended. Then, 24 algorithms are examined. The best performance algorithm is obtained using short time Fourier transform STFT. After that, power is calculated by squaring the signal. Then discrete cosine transform DCT is considered. First and second derivative are computed for the DCT signal. Finally statistical calculations are applied for DCT signal, 1st derivative and 2nd derivative. Many classifiers are compared as Artificial Neural Network, KNN, Support Vector Machine SVM, ANFIS, Deep Learning DL with bi-long short term memory BILSTM and long short term memory LSTM. The maximum obtained accuracy is achieved by using DL with BILSTM layer after extracting features from ECG signals using the best algorithm. The obtained training and testing accuracies are 99.5% and 99.1% respectively. Receiver operating characteristics (ROC) of the selected algorithm are approaching to 1. So, the novelty of this research is obtained by applying this algorithm for extracting ECG signals.
No takes yet. Share an insight, caveat, or question.
Mohamed Moustafa Azmy (2022) studied Atrial Fibrillation. Deep Learning with BILSTM layer and STFT/DCT feature extraction vs. Other classifiers (Artificial Neural Network, KNN, SVM, ANFIS, LSTM) was evaluated on Classification accuracy for normal vs Atrial Fibrillation ECG signals. Deep learning with a BILSTM layer using STFT and DCT feature extraction achieved 99.1% testing accuracy for classifying ECG signals as normal or atrial fibrillation.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: