Accurate detection of atrial fibrillation events with R-R intervals from ECG signals

Junbo Duan; Qing Wang; Bo Zhang; Chen Liu; Chenrui Li; Lei Wang

doi:10.1371/journal.pone.0271596

Abstract

Atrial fibrillation (AF) is a typical category of arrhythmia. Clinical diagnosis of AF is based on the detection of abnormal R-R intervals (RRIs) with an electrocardiogram (ECG). Previous studies considered this detection problem as a classification problem and focused on extracting a number of features. In this study we demonstrate that instead of using any specific numerical characteristic as the input feature, the probability density of RRIs from ECG conserves comprehensive statistical information; hence, is a natural and efficient input feature for AF detection. Incorporated with a support vector machine as the classifier, results on the MIT-BIH database indicates that the proposed method is a simple and accurate approach for AF detection in terms of accuracy, sensitivity, and specificity.

Citation: Duan J, Wang Q, Zhang B, Liu C, Li C, Wang L (2022) Accurate detection of atrial fibrillation events with R-R intervals from ECG signals. PLoS ONE 17(8): e0271596. https://doi.org/10.1371/journal.pone.0271596

Editor: Felix Albu, Valahia University of Targoviste: Universitatea Valahia din Targoviste, ROMANIA

Received: February 9, 2022; Accepted: July 3, 2022; Published: August 4, 2022

Copyright: © 2022 Duan et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Data Availability: All ECG files are freely available from the PhysioNet database with URL https://www.physionet.org/content/afdb/1.0.0/, https://www.physionet.org/content/ltafdb/1.0.0/, and https://www.physionet.org/content/nsrdb/1.0.0/.

Funding: This study was partially supported by National Science Foundation of China under grant number 61771381, and Provincial Science Foundation of Shaanxi under grant number 2021JM-128. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Competing interests: The authors have declared that no competing interests exist.

1 Introduction

Atrial fibrillation (AF or AFIB) is a type of abnormal heart rhythm (arrhythmia) characterized by the rapid, irregular beating of the heart’s upper chambers, resulting in the pooling and clotting of blood inside the heart, thereby increasing the risk of heart attack, failure, and stroke [1]. The symptoms of AF frequently begin with short periods of arrhythmia, such as abnormal beating or atrial flutter, followed by longer arrhythmia periods, sometimes even lasting for hours, accompanied occasionally with heart palpitations, fainting, lightheadedness, shortness of breath, or chest pain [2].

The clinical diagnosis of AF is based on the surface electrocardiogram (ECG), and because of the disorganized electrical activity, AF is characterised by the absence of a P wave. However, because the amplitude of the P wave is relatively low (also a heavy baseline), making its detection difficult, the R-R interval (RRI), which reflects the ventricular interbeat, was proposed as a significant biomarker for AF detection [3]. Compared with RRI in regular rhythm segments, consecutive RRIs during AF episodes exhibit low averages and high fluctuations, reflecting rapid and irregular heart beating. Fig 1 illustrates a typical ECG record (04043) from the MIT-BIH atrial fibrillation database (AFDB) [4, 5], which demonstrates the different patterns of RRIs (red line) in and off AF segments. Since AF episodes duration may change from a few seconds to hours, the chance of AF detection depends heavily on the monitoring period of the ECG. To maximize the probability of AF detection, long-term monitoring of the ECG is required, for example, the use of a Holter. However, the visual inspection of twenty-four-hour-long ECG records is time consuming for clinicians. Hence, an automated AF detection method is required. Another issue to be addressed is the distortion in the long-term record introduced by physical activities; therefore, an accurate AF detection method is needed. With the rapid development of wearable and smart devices, memory-efficient, real-time, automatic, and accurate AF detection methods have become possible [6–8].

Download:

Fig 1. Demonstration of signals and annotations of AFDB (record 04043).

Blue solid and green dashed lines indicate the ECG signals of two channels, red circled line is the RRI, and black thick line is the AF label (1 and 0 indicate AF and normal, respectively). It can be easily observed that the RRIs of AF and non-AF segments exhibit different patterns.

https://doi.org/10.1371/journal.pone.0271596.g001

As illustrated in Fig 2, the histogram of the RRI of AF events (panel (a)) exhibits a lower mean, longer tail, and is left-skewed, compared with that of normal events (panel (b)). These findings are evident from Fig 1(b). Based on these findings, mathematical expectation, variance/deviation, skewness, difference of RRIs, and other higher-order statistics were proposed as features for classification.

Download:

Fig 2. The average histogram of RRIs.

(a) represents AF records of AFDB, and (b) represents normal records from NSRDB.

https://doi.org/10.1371/journal.pone.0271596.g002

Table 1 lists 35 published studies that have been conducted to develop efficient AF detection methods. Tateno and Glass [3] first noticed the increase in the variation on AF episodes and hence proposed the coefficient of variation as a feature of RRI and ΔRRI (the first-order difference of RRI). Subsequently, they used statistical hypothesis testing to verify the existence of AF events. They also proposed the use of the Kolmogorov-Smirnov test to compare the histograms of AF RRIs and normal one. Many subsequent studies considered this detection problem as a classification problem and focused on the extraction of various features and the design of classifiers. These features include entropy [9–13], mean and/or median (with or without normalization), root mean square and/or variance [14–16], quantiles [16, 17], median absolute deviation [10, 16, 17], coefficients of wavelet transformation [12, 13], Markov score [18] of RRI and/or ΔRRI, or a combination of several features [10, 11, 16, 19, 20]. In recent studies, deep learning algorithms such as long short-term memory (LSTM) [21, 22], and others [20, 23–25] have been used to process original signals without feature extraction.

Download:

Table 1. List of published methods (in chronological order).

https://doi.org/10.1371/journal.pone.0271596.t001

In this study, from a statistical perspective, we consider that instead of employing any numerical characteristic (i.e., mean, variance, skewness, etc.) as a specific feature, the probability density function conserves comprehensive information and hence enables high-performance classification. Consequently, we propose the use of a histogram of the RRI from an ECG as a natural and general feature and the widely used support vector machine (SVM) as the classifier.

2 Materials and methods

2.1 Databases

This study employed the MIT-BIH atrial fibrillation database (AFDB) [4, 5], which is widely used in arrhythmia studies. This database includes 25 records of human subjects with AF, and each record includes two-channel ECG signals with a sample frequency of 250 Hz and a 12-bit A/D resolution. Furthermore, this database contains clinical annotations and QRS calls, and supports online retrieval with the easy-to-use toolbox waveform database (WFDB) [43, 44]. Note that R waves were already called by WFDB, so this paper do not cover the detection of R waves from an ECG signal. Researchers interested in this topic are referred to fruitful literature [45–48].

Fig 1 illustrates a typical record (ID 04043) of the AFDB, including two ECG channels (blue and green lines), the RRI (red line), and AF label (black line).

The MIT-BIH long-term atrial fibrillation database (LTAFDB) [49, 50] was also employed as a positive test dataset, which includes 84 long-term (24 hours) ECG recordings, with the same sampling parameters as the AFDB.

As in normal control cases, to evaluate the specificity, this study also employed the MIT-BIH normal sinus rhythm database (NSRDB) [51, 52], which includes long-term ECG records of 18 human subjects who exhibited no significant signs of arrhythmia.

All 127 (25 + 18 + 84) records were downloaded using MATLAB (the pseudo code is listed in S2 File).

2.2 Performance criteria

Three widely used criteria were employed to quantify AF detection performance: accuracy (ACC), sensitivity (SEN), and specificity (SPE).

SEN is referred to as the true positive rate, which is used to measure how well a method can identify real patients, and is defined as the proportion of true positives among all positive subjects.

SPE is referred to as the true negative rate, which is used to measure how well a method can identify a normal person and is defined as the proportion of true negatives among all negative subjects.

For diagnosis and screening, there exists a trade-off between SEN and SPE; therefore, ACC is commonly used to consider SEN and SPE integrally. ACC is defined as the proportion of the sum of true positives and true negatives among all the samples.

Notably, precision (PRE), also known as positive predictive value (PPV), is frequently employed in several studies and is defined as the proportion of true positives among all detected positive cases. However, among these four criteria (ACC, SEN, SPE, and PRE) only three are independent, and the fourth can be calculated depending on the other three (see S1 File). Therefore in our study, only ACC, SEN, and SPE were evaluated, and PRE was not considered. Among the studies listed in Table 1, a few studies provided PRE, while others provided ACC; hence, we used the formulas mentioned in S1 File to convert among them, and the resultant values have been labelled with asterisks.

2.3 Method

2.3.1 Data pre-processing.

After 127 records were downloaded, the following pre-processing steps were followed:

RRI values were re-scaled from the sample index to milliseconds by dividing with the sampling frequency;
Annotations and comments of AFDB and LTAFDB were resolved, and RRI regions with the string ‘(AFIB’ were selected as positive regions;
All RRI regions of NSRDB were selected as negative regions;
Both positive and negative regions were cut to segments, each including 30 PPIs;
A histogram with M bins of each RRI segment was calculated, and stored in a row vector of size M;
The N₀ row vectors from NSRDB were cascaded vertically to form the negative sample matrix X₀, and the same method was employed for vectors from AFDB and LTAFDB, yielding matrices X₁ and X₂ of height N₁ and N₂, respectively.

After pre-processing, we obtained three sample matrices X₀, X₁, and X₂, with width M, and height N₀ = 58742, N₁ = 16817, and N₂ = 101376, respectively.

2.3.2 Classifier.

Soft-margin support vector machine (SVM) [53, 54] was trained as the classifier, which is formally defined as the following optimization problem (the Lagrangian dual form): (1) where stores the N training samples, each sample is a row vector of length M; stores the labels of samples (1 for AF, and -1 for normal); is an unknown weight vector to be optimized; sum(α) is the sum of all elements in α; ⊙ is the point-wise multiplication (the Khatri-Rao product); c is a box constraint parameter, which controls the strength of regularization; is the kernel matrix of X, with element k_ij = κ(x_i, x_j) is the Gaussian kernel (or radial basis function): (2) where σ is a scale parameter.

To optimize problem (1), sequential minimal optimization (SMO) [55] was utilized as the solver. Once α is obtained, the bias parameter b can be calculated as: (3) For a test PPI vector t, the predicting function read: (4) and if p(t) > 0, an AF event is detected.

3 Results

3.1 Kernel function

First, we compared the performance of the Gaussian kernel function (2) with that of the linear kernel function , and the third-order polynomial kernel function . At this step, all other parameters were set to default values (scale parameter σ = 1, box constraint parameter c = 1, and number of bins M = 10).

The SVM classifier was trained with the positive and negative sample matrices X₁ and X₀ by constructing a kernel matrix K_X of size (N₁ + N₀) × (N₁ + N₀), and a training label y with N₁ ones and N₀ negative ones. The optimize problem (1) was solved with SMO solver to train the weight vector α; then, the bias parameter b was calculated based on Eq (3). Subsequently, the same samples were tested with the trained SVM, and the performance criteria were evaluated.

Fig 3 illustrates the performance with different kernel functions. It was demonstrated that the radial basis function was the best, and this kernel was chosen in the sequel.

Download:

Fig 3. The performance with different kernel functions.

https://doi.org/10.1371/journal.pone.0271596.g003

3.2 Number of histogram bins

Subsequently, we tested the impact of the histogram bin number M on detection performance. Because approximately 99% of the RRI values lie within the region of 50 ms and 2e3 ms, the centres of the first and last bins were set to 50 ms and 2e3 ms, respectively. Other M − 2 bin centres were located linearly within this region. RRI values beyond this region were assigned to either the first or the last bin.

The training of SVM and performance evaluation were the same as in the previous experiment. Fig 4 demonstrates the results, which indicate that the detection performance increases with an increase in M and reaches the ceiling at 30. Therefore, in the following experiments, M was fixed at 30.

Download:

Fig 4. The performance with different histogram bin numbers.

https://doi.org/10.1371/journal.pone.0271596.g004

3.3 Cross-validation with scale and box constraint parameters

The scale parameter σ and box constraint parameter impacted the training significantly; hence, we used ten-fold cross-validation to optimize these two parameters. Both σ and c were sampled on a two-dimensional logarithmic grid. The training of SVM, performance evaluation, and training dateset were the same as in the previous experiment. Fig 5 demonstrates the results in which panel (b) indicates that a high SEN performance requires large-scale and box constraint parameters, and panel (c) indicates that a high SPE performance requires small-scale and box constraint parameters. As mentioned, ACC considers both SEN and SPE, and the best performance was achieved at σ = 3.2, and c = 1 (the yellow star in panel (a)). Table 2 lists the ten-fold cross-validation performance in this setting.

Download:

Fig 5. Ten-fold cross validation performance with different scale and box constraint parameters.

https://doi.org/10.1371/journal.pone.0271596.g005

Download:

Table 2. Performance of ten-fold cross validation at yellow star point in Fig 5.

https://doi.org/10.1371/journal.pone.0271596.t002

3.4 Independent dataset testing

In the last experiment, the SVM model was trained with AFDB (N₁ = 16817) and NSRDB (N₀ = 58742) as positive and negative samples, respectively. Model parameters were set according to the results of previous experiments. Subsequently, LTAFDB (N₂ = 101376) was used as the independent positive testing dataset. The confusion matrix is shown in Table 3, and detection results for ACC, SEN, and SPE were 0.9697, 0.9524, and 0.9994, respectively, thus indicating a good generalization performance.

Download:

Table 3. Confusion matrix of independent dataset testing.

https://doi.org/10.1371/journal.pone.0271596.t003

4 Conclusion and discussion

We conclude that an accurate detection method for atrial fibrillation events based on the RR interval measured from an ECG signal was proposed in this paper. The advantage of the proposed method over the methods described in literature is that: instead of using any specific numerical characteristic (e.g., entropy, mean, median, root mean square, variance, quantiles, etc. or a combination of several characteristics) as the input feature, the probability density conserves all statistical information; hence, is natural, comprehensive, easy-computing and efficient as the input features. On the MIT-BIH databases, the proposed method achieved 0.9843±0.0019, 0.9848±0.0029, and 0.9840±0.0024, in terms of ACC, SEN, and SPE, respectively, for a ten-fold cross-validation, and 0.9697, 0.9524, and 0.9994, respectively, for an independent testing, indicating that the proposed method is effective in AF detection.

Note that some studies highlighted the difference between the histograms of PPIs of AF and normal one, and proposed the use of a histogram to detect AF, but the manner in which they utilize histograms is quite different from that in this study. For example, Tateno and Glass [3] observed an increase in variation in AF episodes, and proposed using the Kolmogorov-Smirnov test to compare the histogram of AF RRI and normal RRI; Petrucci et al. [26] calculated several statistics, such as the distribution width based on the histogram of RRI prematurity and ΔRRI, and used a geometric test to detect AF. Alternatively, this study proposes using the histogram as the feature vector and an input to the support vector machine for classification, which is the main contribution of this study.

Supporting information

S1 File. Relations between accuracy (ACC), sensitivity (SEN), specificity (SPE), and precision (PRE).

https://doi.org/10.1371/journal.pone.0271596.s001

(PDF)

S2 File. Dataset downloading.

https://doi.org/10.1371/journal.pone.0271596.s002

(PDF)

References

1. Centers for Disease Control and Prevention. https://www.cdc.gov/heartdisease/other conditions.htm;.
2. Munger TM, Wu LQ, Shen WK. Atrial fibrillation. The Journal of Biomedical Research. 2014;28(1):1–17. pmid:24474959
- View Article
- PubMed/NCBI
- Google Scholar
3. Tateno K, Glass L. Automatic detection of atrial fibrillation using the coefficient of variation and density histograms of RR and ΔRR intervals. Medical Biological Engineering Computing. 2001;39(6):664–671. pmid:11804173
- View Article
- PubMed/NCBI
- Google Scholar
4. PhysioNet. https://www.physionet.org/content/afdb/1.0.0/;.
5. Moody G, Mark R. A new method for detecting atrial fibrillation using R-R intervals. Computers in Cardiology. 1983;10:227–230.
- View Article
- Google Scholar
6. Tison GH, Sanchez JM, Ballinger B, Singh A, Olgin JE, Pletcher MJ, et al. Passive detection of atrial fibrillation using a commercially available smartwatch. JAMA cardiology. 2018;3(5):409–416. pmid:29562087
- View Article
- PubMed/NCBI
- Google Scholar
7. Dunn J, Runge R, Snyder M. Wearables and the medical revolution. Personalized medicine. 2018;15(5):429–448. pmid:30259801
- View Article
- PubMed/NCBI
- Google Scholar
8. Faust O, Ciaccio EJ, Acharya UR. A review of atrial fibrillation detection methods as a service. International journal of environmental research and public health. 2020;17(9):3093.
- View Article
- Google Scholar
9. Lake DE, Moorman JR. Accurate estimation of entropy in very short physiological time series: the problem of atrial fibrillation detection in implanted ventricular devices. American Journal of Physiology Heart and Circulatory Physiology. 2011;300(1):319–325. pmid:21037227
- View Article
- PubMed/NCBI
- Google Scholar
10. Kennedy A, Finlay DD, Guldenring D, Bond RR, Moran K, McLaughlin J. Automated detection of atrial fibrillation using R-R intervals and multivariate-based classification. Journal of Electrocardiology. 2016;49(6):871–876. pmid:27717571
- View Article
- PubMed/NCBI
- Google Scholar
11. Andersen R, Poulsen E, Puthusserypady S. A novel approach for automatic detection of Atrial Fibrillation based on Inter Beat Intervals and Support Vector Machine. vol. 2017; 2017. p. 2039–2042.
- View Article
- Google Scholar
12. Kumar M, Pachori RB, Acharya UR. Automated diagnosis of atrial fibrillation ECG signals using entropy features extracted from flexible analytic wavelet transform. Biocybernetics and Biomedical Engineering. 2018;38(3):564–573.
- View Article
- Google Scholar
13. Hu Y, Zhao Y, Liu J, Pang J, Zhang C, Li P. An effective frequency domain feature of atrial fibrillation based on time-frequency analysis. BMC Medical Informatics and Decision Making. 2020; p. 308–318. pmid:33239025
- View Article
- PubMed/NCBI
- Google Scholar
14. Logan B, Healey JH. Robust detection of atrial fibrillation for a long term telemonitoring system. vol. 32; 2005. p. 619–622.
- View Article
- Google Scholar
15. Nuryani N, Harjito B, Yahya I, Lestari A. Atrial fibrillation detection using support vector machine. In: Proceedings of the Joint International Conference on Electric Vehicular Technology and Industrial, Mechanical, Electrical and Chemical Engineering (ICEVT IMECE); 2015. p. 215–218.
16. Wrobel J, Horoba K, Matonia A, Kupka T, Henzel N, Sobotnicka E. Optimizing the Automated Detection of Atrial Fibrillation Episodes in Long-term Recording Instrumentation. In: International Conference on Mixed Design of Integrated Circuits and Systems; 2018. p. 460–464.
17. Czabanski R, Horoba K, Wrobel J, Matonia A, Martinek R, Kupka T, et al. Detection of Atrial Fibrillation Episodes in Long-Term Heart Rhythm Signals Using a Support Vector Machine. Sensors. 2020;20(3):765–788. pmid:32019220
- View Article
- PubMed/NCBI
- Google Scholar
18. Babaeizadeh S, Gregg RE, Helfenbein ED, Lindauer JM, Zhou SH. Improvements in atrial fibrillation detection for real-time monitoring. Journal of Electrocardiology. 2009;42(6):522–526. pmid:19608194
- View Article
- PubMed/NCBI
- Google Scholar
19. Colloca R, Johnson AEW, Mainardi L, Clifford GD. A Support Vector Machine approach for reliable detection of atrial fibrillation events. Computing in Cardiology. 2013;40:1047–1050.
- View Article
- Google Scholar
20. Shao M, Zhou Z, Bin G, Bai Y, Wu S. A Wearable Electrocardiogram Telemonitoring System for Atrial Fibrillation Detection. Sensors. 2020;20(3):606–621. pmid:31979184
- View Article
- PubMed/NCBI
- Google Scholar
21. Faust O, Shenfield A, Kareem M, San TR, Fujita H, Acharya UR. Automated detection of atrial fibrillation using long short-term memory network with RR interval signals. Computers in Biology and Medicine. 2018;102:327–335. pmid:30031535
- View Article
- PubMed/NCBI
- Google Scholar
22. Andersen RS, Peimankar A, Puthusserypady S. A deep learning approach for real-time detection of atrial fibrillation. Expert Systems with Applications. 2019;115:465–473.
- View Article
- Google Scholar
23. Ivanovic MD, Atanasoski V, Shvilkin A, Hadzievski L, Maluckov A. Deep Learning Approach for Highly Specific Atrial Fibrillation and Flutter Detection based on RR Intervals. In: 2019 41st Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC); 2019. p. 1780–1783.
24. Ghosh S, Tripathy R, Arrieta Paternina M, Arrieta J, Zamora-Mendez A, Naik G. Detection of Atrial Fibrillation from Single Lead ECG Signal Using Multirate Cosine Filter Bank and Deep Neural Network. Journal of Medical Systems. 2020;44:114–128. pmid:32388733
- View Article
- PubMed/NCBI
- Google Scholar
25. Mousavi S, Afghah F, Acharya UR. HAN-ECG: An interpretable atrial fibrillation detection model using hierarchical attention networks. Computers in Biology and Medicine. 2020;127:104057–104065. pmid:33126126
- View Article
- PubMed/NCBI
- Google Scholar
26. Petrucci E, Balian V, Filippini G, Mainardi L. Atrial Fibrillation Detection Algorithms for Very Long Term ECG Monitoring. vol. 2005; 2005. p. 623–626.
- View Article
- Google Scholar
27. Dash S, Chon K, Lu S, Raeder E. Automatic real time detection of atrial fibrillation. Annals of Biomedical Engineering. 2009;37(9):1701–1709. pmid:19533358
- View Article
- PubMed/NCBI
- Google Scholar
28. Huang C, Ye S, Chen H, Li D, He F, Tu Y. A Novel Method for Detection of the Transition Between Atrial Fibrillation and Sinus Rhythm. IEEE Transactions on Biomedical Engineering. 2011;58(4):1113–1119. pmid:21134807
- View Article
- PubMed/NCBI
- Google Scholar
29. Lian J, Wang L, Muessig D. A Simple Method to Detect Atrial Fibrillation Using RR Intervals. The American Journal of Cardiology. 2011;107(10):1494–1497. pmid:21420064
- View Article
- PubMed/NCBI
- Google Scholar
30. Du X, Rao N, Qian M, Liu D, Li J, Feng W, et al. A Novel Method for Real-Time Atrial Fibrillation Detection in Electrocardiograms Using Multiple Parameters. Annals of Noninvasive Electrocardiology. 2013;19:217–225. pmid:24252119
- View Article
- PubMed/NCBI
- Google Scholar
31. Zhou X, Ding H, Ung B, Pickwell-MacPherson E, Zhang Y. Automatic online detection of atrial fibrillation based on symbolic dynamics and Shannon entropy. Biomedical Engineering Online. 2014;13(1):18–35. pmid:24533474
- View Article
- PubMed/NCBI
- Google Scholar
32. Ladavich S, Ghoraani B. Rate-independent detection of atrial fibrillation by statistical modeling of atrial activity. Biomedical Signal Processing and Control. 2015;18:274–281.
- View Article
- Google Scholar
33. Petrėnas A, Marozas V, Sörnmo L. Low-complexity detection of atrial fibrillation in continuous long-term monitoring. Computers in Biology and Medicine. 2015;65:184–191. pmid:25666902
- View Article
- PubMed/NCBI
- Google Scholar
34. García M, Ródenas J, Alcaraz R, Rieta JJ. Application of the relative wavelet energy to heart rate independent detection of atrial fibrillation. Computer Methods and Programs in Biomedicine. 2016;131:157–168. pmid:27265056
- View Article
- PubMed/NCBI
- Google Scholar
35. Islam MS, Ammour N, Alajlan N, Aboalsamh H. Rhythm-based heartbeat duration normalization for atrial fibrillation detection. Computers in Biology and Medicine. 2016;72:160–169. pmid:27043858
- View Article
- PubMed/NCBI
- Google Scholar
36. Afdala A, Nuryani N, Nugroho AS. Automatic Detection of Atrial Fibrillation Using Basic Shannon Entropy of RR Interval Feature. Journal of Physics: Conference Series. 2017;795:012038–012043.
- View Article
- Google Scholar
37. Cui X, Chang E, Yang WH, Jiang B, Yang AC, Peng CK. Automated Detection of Paroxysmal Atrial Fibrillation Using an Information-Based Similarity Approach. Entropy. 2017;19:677–690.
- View Article
- Google Scholar
38. Li Y, Tang X, Wang A, Tang H. Probability density distribution of ΔRR intervals: a novel method for the detection of atrial fibrillation. Australasian Physical and Engineering Sciences in Medicine. 2017;40(6):707–716. pmid:28620839
- View Article
- PubMed/NCBI
- Google Scholar
39. Chen Z, Li J, Li Z, Peng Y, Gao X. Automatic detection and classification of atrial fibrillation using RR intervals and multi-eigenvalue. Journal of Biomedical Engineering. 2018;35(4):56–62. pmid:30124017
- View Article
- PubMed/NCBI
- Google Scholar
40. Lan T, Yang C. Prediction of recurrence of paroxysmal atrial fibrillation based on RR interval. Journal of Biomedical Engineering. 2019;36(4):521–530. pmid:31441251
- View Article
- PubMed/NCBI
- Google Scholar
41. Buscema PM, Grossi E, Massini G, Breda M, Della Torre F. Computer Aided Diagnosis for atrial fibrillation based on new artificial adaptive systems. Computer Methods and Programs in Biomedicine. 2020;191:105401–105412. pmid:32146212
- View Article
- PubMed/NCBI
- Google Scholar
42. Marsili IA, Biasiolli L, Masè M, Adami A, Andrighetti AO, Ravelli F, et al. Implementation and validation of real-time algorithms for atrial fibrillation detection on a wearable ECG device. Computers in Biology and Medicine. 2020;116:103540–103549. pmid:31751811
- View Article
- PubMed/NCBI
- Google Scholar
43. PhysioNet. https://physionet.org/content/wfdb-matlab/0.10.0/;.
44. Silva I, Moody G. An Open-source Toolbox for Analysing and Processing PhysioNet Databases in MATLAB and Octave. Journal of Open Research Software. 2014;2(1):e27. pmid:26525081
- View Article
- PubMed/NCBI
- Google Scholar
45. Benitez D, Gaydecki P, Zaidi A, Fitzpatrick A. The use of the Hilbert transform in ECG signal analysis. Computers in biology and medicine. 2001;31(5):399–406. pmid:11535204
- View Article
- PubMed/NCBI
- Google Scholar
46. Manikandan MS, Soman K. A novel method for detecting R-peaks in electrocardiogram (ECG) signal. Biomedical Signal Processing and Control. 2012;7(2):118–128.
- View Article
- Google Scholar
47. Ravanshad N, Rezaee-Dehsorkh H, Lotfi R, Lian Y. A level-crossing based QRS-detection algorithm for wearable ECG sensors. IEEE J Biomed Health. 2013;18(1):183–192.
- View Article
- Google Scholar
48. Merah M, Abdelmalik T, Larbi B. R-peaks detection based on stationary wavelet transform. Computer methods and programs in biomedicine. 2015;121(3):149–160. pmid:26105724
- View Article
- PubMed/NCBI
- Google Scholar
49. PhysioNet. https://www.physionet.org/content/ltafdb/1.0.0/;.
50. Petrutiu S, Sahakian A, Swiryn S. Abrupt changes in fibrillatory wave characteristics at the termination of paroxysmal atrial fibrillation in humans. Europace. 2007;9:466–470. pmid:17540663
- View Article
- PubMed/NCBI
- Google Scholar
51. PhysioNet. https://www.physionet.org/content/nsrdb/1.0.0/;.
52. Goldberger A, Amaral L, Glass L, Hausdorff J, Ivanov PC, Mark R, et al. PhysioBank, PhysioToolkit, and PhysioNet: Components of a new research resource for complex physiologic signals. Circulation. 2000;101(23):e215–e220. pmid:10851218
- View Article
- PubMed/NCBI
- Google Scholar
53. Vapnik VN. Statistical Learning Theory. New York: Wiley; 1998.
54. Vapnik VN. An Overview of Statistical Learning Theory. IEEE Trans Neural Networks. 1999;10(5):988–999. pmid:18252602
- View Article
- PubMed/NCBI
- Google Scholar
55. Platt J. Sequential Minimal Optimization: A Fast Algorithm for Training Support Vector Machines. Microsoft Research; 1998. MSR-TR-98-14.
- View Article
- Google Scholar

[ref1] 1. Centers for Disease Control and Prevention. https://www.cdc.gov/heartdisease/other conditions.htm;.

[ref2] 2. Munger TM, Wu LQ, Shen WK. Atrial fibrillation. The Journal of Biomedical Research. 2014;28(1):1–17. pmid:24474959
View Article
PubMed/NCBI
Google Scholar

[3] View Article

[4] PubMed/NCBI

[5] Google Scholar

[ref3] 3. Tateno K, Glass L. Automatic detection of atrial fibrillation using the coefficient of variation and density histograms of RR and ΔRR intervals. Medical Biological Engineering Computing. 2001;39(6):664–671. pmid:11804173
View Article
PubMed/NCBI
Google Scholar

[7] View Article

[8] PubMed/NCBI

[9] Google Scholar

[ref4] 4. PhysioNet. https://www.physionet.org/content/afdb/1.0.0/;.

[ref5] 5. Moody G, Mark R. A new method for detecting atrial fibrillation using R-R intervals. Computers in Cardiology. 1983;10:227–230.
View Article
Google Scholar

[12] View Article

[13] Google Scholar

[ref6] 6. Tison GH, Sanchez JM, Ballinger B, Singh A, Olgin JE, Pletcher MJ, et al. Passive detection of atrial fibrillation using a commercially available smartwatch. JAMA cardiology. 2018;3(5):409–416. pmid:29562087
View Article
PubMed/NCBI
Google Scholar

[15] View Article

[16] PubMed/NCBI

[17] Google Scholar

[ref7] 7. Dunn J, Runge R, Snyder M. Wearables and the medical revolution. Personalized medicine. 2018;15(5):429–448. pmid:30259801
View Article
PubMed/NCBI
Google Scholar

[19] View Article

[20] PubMed/NCBI

[21] Google Scholar

[ref8] 8. Faust O, Ciaccio EJ, Acharya UR. A review of atrial fibrillation detection methods as a service. International journal of environmental research and public health. 2020;17(9):3093.
View Article
Google Scholar

[23] View Article

[24] Google Scholar

[ref9] 9. Lake DE, Moorman JR. Accurate estimation of entropy in very short physiological time series: the problem of atrial fibrillation detection in implanted ventricular devices. American Journal of Physiology Heart and Circulatory Physiology. 2011;300(1):319–325. pmid:21037227
View Article
PubMed/NCBI
Google Scholar

[26] View Article

[27] PubMed/NCBI

[28] Google Scholar

[ref10] 10. Kennedy A, Finlay DD, Guldenring D, Bond RR, Moran K, McLaughlin J. Automated detection of atrial fibrillation using R-R intervals and multivariate-based classification. Journal of Electrocardiology. 2016;49(6):871–876. pmid:27717571
View Article
PubMed/NCBI
Google Scholar

[30] View Article

[31] PubMed/NCBI

[32] Google Scholar

[ref11] 11. Andersen R, Poulsen E, Puthusserypady S. A novel approach for automatic detection of Atrial Fibrillation based on Inter Beat Intervals and Support Vector Machine. vol. 2017; 2017. p. 2039–2042.
View Article
Google Scholar

[34] View Article

[35] Google Scholar

[ref12] 12. Kumar M, Pachori RB, Acharya UR. Automated diagnosis of atrial fibrillation ECG signals using entropy features extracted from flexible analytic wavelet transform. Biocybernetics and Biomedical Engineering. 2018;38(3):564–573.
View Article
Google Scholar

[37] View Article

[38] Google Scholar

[ref13] 13. Hu Y, Zhao Y, Liu J, Pang J, Zhang C, Li P. An effective frequency domain feature of atrial fibrillation based on time-frequency analysis. BMC Medical Informatics and Decision Making. 2020; p. 308–318. pmid:33239025
View Article
PubMed/NCBI
Google Scholar

[40] View Article

[41] PubMed/NCBI

[42] Google Scholar

[ref14] 14. Logan B, Healey JH. Robust detection of atrial fibrillation for a long term telemonitoring system. vol. 32; 2005. p. 619–622.
View Article
Google Scholar

[44] View Article

[45] Google Scholar

[ref15] 15. Nuryani N, Harjito B, Yahya I, Lestari A. Atrial fibrillation detection using support vector machine. In: Proceedings of the Joint International Conference on Electric Vehicular Technology and Industrial, Mechanical, Electrical and Chemical Engineering (ICEVT IMECE); 2015. p. 215–218.

[ref16] 16. Wrobel J, Horoba K, Matonia A, Kupka T, Henzel N, Sobotnicka E. Optimizing the Automated Detection of Atrial Fibrillation Episodes in Long-term Recording Instrumentation. In: International Conference on Mixed Design of Integrated Circuits and Systems; 2018. p. 460–464.

[ref17] 17. Czabanski R, Horoba K, Wrobel J, Matonia A, Martinek R, Kupka T, et al. Detection of Atrial Fibrillation Episodes in Long-Term Heart Rhythm Signals Using a Support Vector Machine. Sensors. 2020;20(3):765–788. pmid:32019220
View Article
PubMed/NCBI
Google Scholar

[49] View Article

[50] PubMed/NCBI

[51] Google Scholar

[ref18] 18. Babaeizadeh S, Gregg RE, Helfenbein ED, Lindauer JM, Zhou SH. Improvements in atrial fibrillation detection for real-time monitoring. Journal of Electrocardiology. 2009;42(6):522–526. pmid:19608194
View Article
PubMed/NCBI
Google Scholar

[53] View Article

[54] PubMed/NCBI

[55] Google Scholar

[ref19] 19. Colloca R, Johnson AEW, Mainardi L, Clifford GD. A Support Vector Machine approach for reliable detection of atrial fibrillation events. Computing in Cardiology. 2013;40:1047–1050.
View Article
Google Scholar

[57] View Article

[58] Google Scholar

[ref20] 20. Shao M, Zhou Z, Bin G, Bai Y, Wu S. A Wearable Electrocardiogram Telemonitoring System for Atrial Fibrillation Detection. Sensors. 2020;20(3):606–621. pmid:31979184
View Article
PubMed/NCBI
Google Scholar

[60] View Article

[61] PubMed/NCBI

[62] Google Scholar

[ref21] 21. Faust O, Shenfield A, Kareem M, San TR, Fujita H, Acharya UR. Automated detection of atrial fibrillation using long short-term memory network with RR interval signals. Computers in Biology and Medicine. 2018;102:327–335. pmid:30031535
View Article
PubMed/NCBI
Google Scholar

[64] View Article

[65] PubMed/NCBI

[66] Google Scholar

[ref22] 22. Andersen RS, Peimankar A, Puthusserypady S. A deep learning approach for real-time detection of atrial fibrillation. Expert Systems with Applications. 2019;115:465–473.
View Article
Google Scholar

[68] View Article

[69] Google Scholar

[ref23] 23. Ivanovic MD, Atanasoski V, Shvilkin A, Hadzievski L, Maluckov A. Deep Learning Approach for Highly Specific Atrial Fibrillation and Flutter Detection based on RR Intervals. In: 2019 41st Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC); 2019. p. 1780–1783.

[ref24] 24. Ghosh S, Tripathy R, Arrieta Paternina M, Arrieta J, Zamora-Mendez A, Naik G. Detection of Atrial Fibrillation from Single Lead ECG Signal Using Multirate Cosine Filter Bank and Deep Neural Network. Journal of Medical Systems. 2020;44:114–128. pmid:32388733
View Article
PubMed/NCBI
Google Scholar

[72] View Article

[73] PubMed/NCBI

[74] Google Scholar

[ref25] 25. Mousavi S, Afghah F, Acharya UR. HAN-ECG: An interpretable atrial fibrillation detection model using hierarchical attention networks. Computers in Biology and Medicine. 2020;127:104057–104065. pmid:33126126
View Article
PubMed/NCBI
Google Scholar

[76] View Article

[77] PubMed/NCBI

[78] Google Scholar

[ref26] 26. Petrucci E, Balian V, Filippini G, Mainardi L. Atrial Fibrillation Detection Algorithms for Very Long Term ECG Monitoring. vol. 2005; 2005. p. 623–626.
View Article
Google Scholar

[80] View Article

[81] Google Scholar

[ref27] 27. Dash S, Chon K, Lu S, Raeder E. Automatic real time detection of atrial fibrillation. Annals of Biomedical Engineering. 2009;37(9):1701–1709. pmid:19533358
View Article
PubMed/NCBI
Google Scholar

[83] View Article

[84] PubMed/NCBI

[85] Google Scholar

[ref28] 28. Huang C, Ye S, Chen H, Li D, He F, Tu Y. A Novel Method for Detection of the Transition Between Atrial Fibrillation and Sinus Rhythm. IEEE Transactions on Biomedical Engineering. 2011;58(4):1113–1119. pmid:21134807
View Article
PubMed/NCBI
Google Scholar

[87] View Article

[88] PubMed/NCBI

[89] Google Scholar

[ref29] 29. Lian J, Wang L, Muessig D. A Simple Method to Detect Atrial Fibrillation Using RR Intervals. The American Journal of Cardiology. 2011;107(10):1494–1497. pmid:21420064
View Article
PubMed/NCBI
Google Scholar

[91] View Article

[92] PubMed/NCBI

[93] Google Scholar

[ref30] 30. Du X, Rao N, Qian M, Liu D, Li J, Feng W, et al. A Novel Method for Real-Time Atrial Fibrillation Detection in Electrocardiograms Using Multiple Parameters. Annals of Noninvasive Electrocardiology. 2013;19:217–225. pmid:24252119
View Article
PubMed/NCBI
Google Scholar

[95] View Article

[96] PubMed/NCBI

[97] Google Scholar

[ref31] 31. Zhou X, Ding H, Ung B, Pickwell-MacPherson E, Zhang Y. Automatic online detection of atrial fibrillation based on symbolic dynamics and Shannon entropy. Biomedical Engineering Online. 2014;13(1):18–35. pmid:24533474
View Article
PubMed/NCBI
Google Scholar

[99] View Article

[100] PubMed/NCBI

[101] Google Scholar

[ref32] 32. Ladavich S, Ghoraani B. Rate-independent detection of atrial fibrillation by statistical modeling of atrial activity. Biomedical Signal Processing and Control. 2015;18:274–281.
View Article
Google Scholar

[103] View Article

[104] Google Scholar

[ref33] 33. Petrėnas A, Marozas V, Sörnmo L. Low-complexity detection of atrial fibrillation in continuous long-term monitoring. Computers in Biology and Medicine. 2015;65:184–191. pmid:25666902
View Article
PubMed/NCBI
Google Scholar

[106] View Article

[107] PubMed/NCBI

[108] Google Scholar

[ref34] 34. García M, Ródenas J, Alcaraz R, Rieta JJ. Application of the relative wavelet energy to heart rate independent detection of atrial fibrillation. Computer Methods and Programs in Biomedicine. 2016;131:157–168. pmid:27265056
View Article
PubMed/NCBI
Google Scholar

[110] View Article

[111] PubMed/NCBI

[112] Google Scholar

[ref35] 35. Islam MS, Ammour N, Alajlan N, Aboalsamh H. Rhythm-based heartbeat duration normalization for atrial fibrillation detection. Computers in Biology and Medicine. 2016;72:160–169. pmid:27043858
View Article
PubMed/NCBI
Google Scholar

[114] View Article

[115] PubMed/NCBI

[116] Google Scholar

[ref36] 36. Afdala A, Nuryani N, Nugroho AS. Automatic Detection of Atrial Fibrillation Using Basic Shannon Entropy of RR Interval Feature. Journal of Physics: Conference Series. 2017;795:012038–012043.
View Article
Google Scholar

[118] View Article

[119] Google Scholar

[ref37] 37. Cui X, Chang E, Yang WH, Jiang B, Yang AC, Peng CK. Automated Detection of Paroxysmal Atrial Fibrillation Using an Information-Based Similarity Approach. Entropy. 2017;19:677–690.
View Article
Google Scholar

[121] View Article

[122] Google Scholar

[ref38] 38. Li Y, Tang X, Wang A, Tang H. Probability density distribution of ΔRR intervals: a novel method for the detection of atrial fibrillation. Australasian Physical and Engineering Sciences in Medicine. 2017;40(6):707–716. pmid:28620839
View Article
PubMed/NCBI
Google Scholar

[124] View Article

[125] PubMed/NCBI

[126] Google Scholar

[ref39] 39. Chen Z, Li J, Li Z, Peng Y, Gao X. Automatic detection and classification of atrial fibrillation using RR intervals and multi-eigenvalue. Journal of Biomedical Engineering. 2018;35(4):56–62. pmid:30124017
View Article
PubMed/NCBI
Google Scholar

[128] View Article

[129] PubMed/NCBI

[130] Google Scholar

[ref40] 40. Lan T, Yang C. Prediction of recurrence of paroxysmal atrial fibrillation based on RR interval. Journal of Biomedical Engineering. 2019;36(4):521–530. pmid:31441251
View Article
PubMed/NCBI
Google Scholar

[132] View Article

[133] PubMed/NCBI

[134] Google Scholar

[ref41] 41. Buscema PM, Grossi E, Massini G, Breda M, Della Torre F. Computer Aided Diagnosis for atrial fibrillation based on new artificial adaptive systems. Computer Methods and Programs in Biomedicine. 2020;191:105401–105412. pmid:32146212
View Article
PubMed/NCBI
Google Scholar

[136] View Article

[137] PubMed/NCBI

[138] Google Scholar

[ref42] 42. Marsili IA, Biasiolli L, Masè M, Adami A, Andrighetti AO, Ravelli F, et al. Implementation and validation of real-time algorithms for atrial fibrillation detection on a wearable ECG device. Computers in Biology and Medicine. 2020;116:103540–103549. pmid:31751811
View Article
PubMed/NCBI
Google Scholar

[140] View Article

[141] PubMed/NCBI

[142] Google Scholar

[ref43] 43. PhysioNet. https://physionet.org/content/wfdb-matlab/0.10.0/;.

[ref44] 44. Silva I, Moody G. An Open-source Toolbox for Analysing and Processing PhysioNet Databases in MATLAB and Octave. Journal of Open Research Software. 2014;2(1):e27. pmid:26525081
View Article
PubMed/NCBI
Google Scholar

[145] View Article

[146] PubMed/NCBI

[147] Google Scholar

[ref45] 45. Benitez D, Gaydecki P, Zaidi A, Fitzpatrick A. The use of the Hilbert transform in ECG signal analysis. Computers in biology and medicine. 2001;31(5):399–406. pmid:11535204
View Article
PubMed/NCBI
Google Scholar

[149] View Article

[150] PubMed/NCBI

[151] Google Scholar

[ref46] 46. Manikandan MS, Soman K. A novel method for detecting R-peaks in electrocardiogram (ECG) signal. Biomedical Signal Processing and Control. 2012;7(2):118–128.
View Article
Google Scholar

[153] View Article

[154] Google Scholar

[ref47] 47. Ravanshad N, Rezaee-Dehsorkh H, Lotfi R, Lian Y. A level-crossing based QRS-detection algorithm for wearable ECG sensors. IEEE J Biomed Health. 2013;18(1):183–192.
View Article
Google Scholar

[156] View Article

[157] Google Scholar

[ref48] 48. Merah M, Abdelmalik T, Larbi B. R-peaks detection based on stationary wavelet transform. Computer methods and programs in biomedicine. 2015;121(3):149–160. pmid:26105724
View Article
PubMed/NCBI
Google Scholar

[159] View Article

[160] PubMed/NCBI

[161] Google Scholar

[ref49] 49. PhysioNet. https://www.physionet.org/content/ltafdb/1.0.0/;.

[ref50] 50. Petrutiu S, Sahakian A, Swiryn S. Abrupt changes in fibrillatory wave characteristics at the termination of paroxysmal atrial fibrillation in humans. Europace. 2007;9:466–470. pmid:17540663
View Article
PubMed/NCBI
Google Scholar

[164] View Article

[165] PubMed/NCBI

[166] Google Scholar

[ref51] 51. PhysioNet. https://www.physionet.org/content/nsrdb/1.0.0/;.

[ref52] 52. Goldberger A, Amaral L, Glass L, Hausdorff J, Ivanov PC, Mark R, et al. PhysioBank, PhysioToolkit, and PhysioNet: Components of a new research resource for complex physiologic signals. Circulation. 2000;101(23):e215–e220. pmid:10851218
View Article
PubMed/NCBI
Google Scholar

[169] View Article

[170] PubMed/NCBI

[171] Google Scholar

[ref53] 53. Vapnik VN. Statistical Learning Theory. New York: Wiley; 1998.

[ref54] 54. Vapnik VN. An Overview of Statistical Learning Theory. IEEE Trans Neural Networks. 1999;10(5):988–999. pmid:18252602
View Article
PubMed/NCBI
Google Scholar

[174] View Article

[175] PubMed/NCBI

[176] Google Scholar

[ref55] 55. Platt J. Sequential Minimal Optimization: A Fast Algorithm for Training Support Vector Machines. Microsoft Research; 1998. MSR-TR-98-14.
View Article
Google Scholar

[178] View Article

[179] Google Scholar

Figures

Abstract

1 Introduction

2 Materials and methods

2.1 Databases

2.2 Performance criteria

2.3 Method

2.3.1 Data pre-processing.

2.3.2 Classifier.

3 Results

3.1 Kernel function

3.2 Number of histogram bins

3.3 Cross-validation with scale and box constraint parameters

3.4 Independent dataset testing

4 Conclusion and discussion

Supporting information

S1 File. Relations between accuracy (ACC), sensitivity (SEN), specificity (SPE), and precision (PRE).

S2 File. Dataset downloading.

References