What does this research mean for the field?

A discrete wavelet-based feature extraction method combined with a Multi-Layer Perceptron (MLP) neural network achieves 99.51% accuracy in automated ECG arrhythmia recognition. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

March 3, 2026Open Access

Development of ECG Arrhythmia Recognition from Category of Statistical Space and Discrete Wavelet Feature Extraction via Classifiers

Key Points

Accurate classification of ECG arrhythmias reached 99.51% using MLP-BP classifiers, indicating strong performance.
Results from the MIT-BIH arrhythmia database confirm high accuracy rates for SVM and PNN classifiers as well.

Structured PICO

Does a discrete wavelet-based algorithm combined with neural network classifiers accurately recognize ECG arrhythmias?

Population

48 records of the MIT-BIH arrhythmia database

Intervention

Discrete wavelet-based algorithm for feature extraction combined with Multi-Layer Perceptron-Back Propagation (MLP-BP), Probabilistic Neural Network (PNN), and support vector machine (SVM) classifiers

Comparator

Comparison among SVM, MLP, and PNN classifiers

Outcome

Accuracy of arrhythmia recognition

An automated ECG arrhythmia recognition system using discrete wavelet feature extraction and an MLP classifier achieved 99.51% accuracy on the MIT-BIH database.

Abstract

Introduction: An important medical diagnostic technique to monitor and detect abnormalities in ECG signal is heart arrhythmias recognition. The physicians workload increases because of the high numbers of heart patients. As a result, a robust automated detection system is mandatory. On the other hand, consistent or periodical heart rhythms disorders with important information, which reflect cardiac activity, lead to cardiac arrhythmias, therefore arrhythmias recognition with acceptable accuracy is required. Aims: In this study, a discrete wavelet-based algorithm for delineation of events in ECG signal is utilized and next a new fusion of MLP-BP and PNN neural networks for heart arrhythmia classification was defined. Methods: Multi resolution analysis can be extracted from any changes in the morphology of an ECG signal into time and frequency analysis. Firstly, artifact and noise are excluded by a discrete wavelet transform (DWT). Secondly, multi lead ECG signal together with QRS complexes of signal is excluded. Finally, the ECG signal is decomposed, and consequently corresponding DWT scales are segmented. Next statistical features from reconstructed ECG segments are attained. Afterwards, curve length and high order moment order-based feature extraction is computed from ECG excerpted segment. Finally, elements of feature vector for modifiable the parameters of classifiers are utilized. Next, Multi-Layer Perceptron-Back Propagation (MLP-BP) neural networks, Probabilistic Neural Network (PNN) and support vector machine (SVM) were designed and adjusted and their results were compared. Results: The proposed algorithm was verified to all 48 records of the MIT-BIH arrhythmia database. Also, the proposed topology of classifiers and their associated parameters is optimized by searching of the best value of parameters. The average value of accuracy of each classifier over all records of MIT-BIH for arrhythmias recognition is Acc=98.19, Acc=99.51 and Acc=97.53 for SVM, MLP and PNN classifiers correspondingly and obtained results were compared with similar peer-reviewed studies in this subject

Read Full Paperexternally

Bookmark

View Full Paper

Cite This Study

Asadi et al. (Wed,) studied this question.

synapsesocial.com/papers/69a75cd1c6e9836116a2600b https://doi.org/https://doi.org/10.5281/zenodo.18397360

Bookmark

View Full Paper