1MSEIT, MATS University, Raipur, Chhattisgarh, India. Email: anurag020981@gmail.com
2VC, MATS University, Raipur, Chhattisgarh, India.
3Department of E&TC, BIT, Bhilai House, Durg, Chhattisgarh, India.
*Corresponding Author: Anurag Kumar Mishra, MSEIT, MATS University, Raipur, Chhattisgarh, India. Email: anurag020981@gmail.com
Electrocardiogram (ECG) signals are vital for detecting cardiac arrhythmias. Medical professionals typically do this diagnosis to detect certain malfunctioning of heart. In such a detection subjectivity arises and also the detection may not be accurate therefor it is important to develop automatic detection and classification techniques. In the past most of the researchers have classified up to 6 cardiac arrythmias and reported their classifier capabilities. It is difficult to accurately classify ECG when there are more arrhythmias to classify. A deep Long Short-Term Memory (LSTM) network is proposed in this paper which classifies 7 arrhythmia types using hybrid features extracted from temporal and frequency domains. The hybrid approach combines Discrete Wavelet Transform (DWT)-based features with statistical descriptors to enhance discriminative power. Training and validation were done on the proposed network using the MIT-BIH Arrhythmia Database (ADB), achieving overall average accuracy, precision, recall and F-score of 99.43%, 99.44%, 99.45% and 99.41% respectively, outperforming traditional CNN and standalone LSTM architectures. These results demonstrate that integrating hybrid features with deep LSTM networks significantly improves ECG classification performance, supporting automated clinical diagnosis. Proposed method's performance is compared with some of the standard existing methods and it was observed that the presented method outperformed numerous existing methods. The hardware requirement and processing time is also less as compared to other methods.
Keywords: ECG, Arrhythmia, LSTM, Deep Learning, Hybrid Features, DWT, MIT-BIH
How to cite this article: Mishra AK, Yadav KP, Dewangan NK. ECG Arrhythmia Classification using Deep LSTM Network with Hybrid Feature Extraction. Int J Drug Deliv Technol. 2026;16(11s): 254-268; DOI: 10.25258/ijddt.16.11s.24.
Source of support: Nil.
Conflict of interest: Nil