A Fast Machine Learning Model for ECG-Based Heartbeat Classification and Arrhythmia Detection

Abstract

We present a fully automatic and fast ECG arrhythmia classifier based on a simple brain-inspired machine learning approach known as Echo State Networks. Our classifier has a low-demanding feature processing that only requires a single ECG lead. Its training and validation follows an inter-patient procedure. Our approach is compatible with an online classification that aligns well with recent advances in health-monitoring wireless devices and wearables. The use of a combination of ensembles allows us to exploit parallelism to train the classifier with remarkable speeds. The heartbeat classifier is evaluated over two ECG databases, the MIT-BIH AR and the AHA. In the MIT-BIH AR database, our classification approach provides a sensitivity of 92.7% and positive predictive value of 86.1% for the ventricular ectopic beats, using the single lead II, and a sensitivity of 95.7% and positive predictive value of 75.1% when using the lead V1'. These results are comparable with the state of the art in fully automatic ECG classifiers and even outperform other ECG classifiers that follow more complex feature-selection approaches.

A hybrid unsupervised—Deep learning tandem for electrooculography time series analysis

“…electrocardiograms (ECG). An ensemble of echo state networks uses ECG data for the classification of arrhythmias [22], as for this type of data a recurrent architecture is appropriate. A standard NN is appointed in [23] for the detection of ischemia from ECG.…”

“…So, once the training saccades are relabelled, the CNN-LSTM architecture is applied on the 'cleaned' collection (line 16) and the most balanced model on the validation set, with respect to the sensitivity (recall) for classes C and P, is selected (lines [16][17]. The test accuracy is done register-wise, however, at this point, it is the majority label of its inner saccades (predicted by the DL model) that give the final class (lines [20][21][22][23][24].…”

From Biodata to Somadata

“…In the past decade, biosignals found many applications in healthcare and HCI [42]. They allowed for applications that support self-awareness by displaying properties of our bodies as they unfold in real-time, or applications that are able to track one's mood over certain periods of time [18], or even to assess one's health status [3]. On the other hand, design researchers and artists bring creative potential to the field, driving further development of biosensing technologies to fit outside the narrow settings they were originally designed for [59,24,25].…”

Regression Heuristics by Optimal Tridimensional Features of Electrocardiogram for Arrhythmia Detection

“…The performance significance of the proposed model has scaled by comparing the observed results of the classification assessment metrics and they are "Precision, Specificity, Sensitivity, Accuracy, F-measure, False Alarming, and Matthews's correlation coefficient (MCC)". The results evinced from the proposed model RH-TOF, and the other contemporary models "A fast machine learning model (FMLM)" [44] and "Multiclass Classification of Cardiac Arrhythmia (MCCA) to Predict Advanced Arrhythmia" [45] has compared and concluded the significance of the proposed model towards arrhythmia detection.…”

“…The recent contributions "Fast Machine Learning Model (FMLM) [44] for ECG-Based Heartbeat Classification" and "Multiclass Classification of Cardiac Arrhythmia (MCCA) [45] Using Improved Feature Selection and SVM Invariants" have considered all possible features and variance in arrhythmia projection format (multiple classes) of the electrocardiograms in respective order. However, these contemporary methods still evincing the considerable false alarming and poor sensitivity, which is due to the improper feature adaption and ignorance of the feature dimensionality.…”