Arrhythmia detection from heart rate variability by SDFA method

Linhares, Raquel Romes

doi:10.1016/j.ijcard.2016.08.286

Cited by 9 publications

(3 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To investigate the LOO-DFA performance compared to the traditional DFA method, we generate time series from fractional Gaussian noise (FGN) models, with 100 replications and length n ∈ {1000, 5000, 10000, 15000}. Here, we adopt the largest window as g(n) = ln(n) 2 ; see [15] for details. For each method, we calculate the empirical values of the mean, the bias, the mean square error (MSE), and the variance (Var) values.…”

Section: Loo-dfa Simulation Resultsmentioning

confidence: 99%

Leave One Out Detrended Fluctuation Analysis

2023

View full text Add to dashboard Cite

A relevant issue in time series analysis is the estimation of long-range dependence, that is, how much future values of a time series depend on current values. One of the ways to verify this dependence is by estimating the Hurst exponent using methods such as detrended fluctuation analysis. Here, we propose a new methodology to estimate the Hurst exponent, named leave one out detrended fluctuation analysis. Furthermore, based on this new estimator for the Hurst exponent, we propose the noise reduction by the leave one out detrended fluctuation analysis method. We apply this new denoising method to electrocardiogram noise reduction. The results presented in this work show that this new methodology outperforms the SureShrink and universal noise reduction methods. topics: detrended fluctuation analysis (DFA), denoising, electrocardiogram (ECG), wavelets

show abstract

Section: Loo-dfa Simulation Resultsmentioning

confidence: 99%

Leave One Out Detrended Fluctuation Analysis

2023

View full text Add to dashboard Cite

show abstract

“…The gradient in DNN was a major issue, which was unstable and tends to either vanish or explode in earlier layers. [12] presented an automatic system for arrhythmia classification, which was the combination of smooth detrended fluctuation analysis (SDFA) and principal of wavelet shrinkage scaling analysis to represent the time series correlation properties. This research was carried out on the online databases; MIT-BIH arrhythmia dataset and MIT-BIH Normal Sinus Rhythm dataset.…”

Section: Literature Reviewmentioning

confidence: 99%

Infinite Feature Selection Based Arrhythmia Classification Using Semantic Features and Random Forest

Rangappa¹,

Prasad²,

Agarwal³

2019

IJIES

View full text Add to dashboard Cite

In this research study, arrhythmia classification was assessed by using Massachusetts Institute of Technology-Beth Israel Hospital (MIT-BIH) arrhythmia database. After signal acquisition, semantic feature extraction (combination of statistical, entropy, linear and non-linear features) was used for extracting the feature values from acquired electrocardiogram (ECG) signals. Here, twenty-six features were combined to perform feature extraction on acquired ECG signals. Then, the infinite feature selection was used for rejecting the irrelevant feature vectors or selecting the optimal feature subsets. After selecting the optimal feature vectors, a binary classifier: random forest was used to classify the eight arrhythmia heartbeat conditions: Normal sinus rhythm (N), Sinoauricular Heart Block (SHB), Premature Ventricular Contraction (PVC), Auricular Fibrillation (AF), Left Bundle Branch Block (LBBB), Premature Atrial Contraction (PAC), Supra-Ventricular-Tachycardia (SVT) and Right Bundle Branch Block (RBBB). This research work includes numerous advantages; earlier detection of arrhythmia diseases, assists clinicians during surgery, and cost efficient compared to the existing machine learning methodologies. The experimental outcome shows that the proposed methodology effectively distinguishes the heartbeat conditions by means of sensitivity, specificity, Negative Predictive Value (NPV), Positive Predictive Value (PPV) and accuracy. The proposed methodology enhances the classification accuracy up to 1.4-1.5% compared to the existing methodologies.

show abstract

“…By varying the length l, the F (l) can be characterized by a scaling exponent, more precisely, the slope coefficient of the line obtained by the regression of F (l) on ln(l), with l ∈ {4,5, ⋯ , g(n)}. Linhares (2016b) determine an optimal choice for the number of regressors in the SDFA method given by g(n) = ⌊(ln(n)) ⌋, where ⌊⋅⌋ indicates the integer part function and n is the length of the time series.…”

Section: Introductionmentioning

confidence: 99%