Can Supervised Learning Be Used to Classify Cardiac Rhythms?

Vollmer, Marcus; Sodmann, Philipp; Caanitz, Leonard; Nath, Neetika; Kaderali, Lars

doi:10.22489/cinc.2017.347-176

Cited by 11 publications

(10 citation statements)

References 8 publications

(10 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For a given frame the classification method must predict if the rhythm is normal, atrial fibrillation (AFib) 3 , or atrial flutter, based on the input representation. AFib is indicated by irregular RR intervals, no distinct P waves and usually variable intervals between two atrial activations (Vollmer et al, 2018). Flutter appears as a saw-tooth pattern of R waves.…”

Section: Quantitative Evaluationmentioning

confidence: 99%

See 1 more Smart Citation

Icentia11K: An Unsupervised Representation Learning Dataset for Arrhythmia Subtype Discovery

Tan,

Androz,

Chamseddine

et al. 2019

Preprint

View full text Add to dashboard Cite

We release the largest public ECG dataset of continuous raw signals for representation learning containing 11 thousand patients and 2 billion labelled beats. Our goal is to enable semi-supervised ECG models to be made as well as to discover unknown subtypes of arrhythmia and anomalous ECG signal events. To this end, we propose an unsupervised representation learning task, evaluated in a semi-supervised fashion. We provide a set of baselines for different feature extractors that can be built upon. Additionally, we perform qualitative evaluations on results from PCA embeddings, where we identify some clustering of known subtypes indicating the potential for representation learning in arrhythmia sub-type discovery.

show abstract

Section: Quantitative Evaluationmentioning

confidence: 99%

“…It is possible that, because the Periodogram and FFT captures periodicity in the signal, it performs better than the other feature extraction methods. Work by Vollmer et al (2018) has shown that it is possible in a supervised setting. The results also shows the issues with using MLPs as a classification method for this task.…”

Section: Quantitative Evaluationmentioning

confidence: 99%

Icentia11K: An Unsupervised Representation Learning Dataset for Arrhythmia Subtype Discovery

Tan,

Androz,

Chamseddine

et al. 2019

Preprint

View full text Add to dashboard Cite

show abstract

“…However, CNNs-based algorithms for stratifying cardiovascular diseases are not well-established due to limited availability of ECG database. Though a few previous studies have applied CNNs to detect cardiac arrhythmias ( Rajpurkar et al, 2017 ; Vollmer et al, 2017 ), it still remains a challenge to develop an effective algorithm for detecting AF based on short episodes of ECGs.…”

Section: Introductionmentioning

confidence: 99%

Automatic Detection of Atrial Fibrillation Based on Continuous Wavelet Transform and 2D Convolutional Neural Networks

et al. 2018

View full text Add to dashboard Cite

Atrial fibrillation (AF) is the most common cardiac arrhythmias causing morbidity and mortality. AF may appear as episodes of very short (i.e., proximal AF) or sustained duration (i.e., persistent AF), either form of which causes irregular ventricular excitations that affect the global function of the heart. It is an unmet challenge for early and automatic detection of AF, limiting efficient treatment strategies for AF. In this study, we developed a new method based on continuous wavelet transform and 2D convolutional neural networks (CNNs) to detect AF episodes. The proposed method analyzed the time-frequency features of the electrocardiogram (ECG), thus being different to conventional AF detecting methods that implement isolating atrial or ventricular activities. Then a 2D CNN was trained to improve AF detection performance. The MIT-BIH Atrial Fibrillation Database was used for evaluating the algorithm. The efficacy of the proposed method was compared with those of some existing methods, most of which implemented the same dataset. The newly developed algorithm using CNNs achieved 99.41, 98.91, 99.39, and 99.23% for the sensitivity, specificity, positive predictive value, and overall accuracy (ACC) respectively. As the proposed algorithm targets the time-frequency feature of ECG signals rather than isolated atrial or ventricular activity, it has the ability to detect AF episodes for using just five beats, suggesting practical applications in the future.

show abstract

“…However, this produces a large amount of data, for which reason automated methods of AF detection are needed. The search for automated algorithms for AF detection during the PhysioNet Challenge 2017 [4] showed many promising methods, usually focused on machine learning approaches such as random forests [5], recurrent and convolutional neural networks [6], [7], supported vector machines [8] and others. However, these robust methods usually use tens or hundreds of features and may be difficult to implement in miniature processing units.…”

Section: Introductionmentioning

confidence: 99%

Shape Analysis of Consecutive Beats May Help in the Automated Detection of Atrial Fibrillation

Plešinger

Andrla

Viščor

et al. 2018

2018 Computing in Cardiology Conference (CinC)

View full text Add to dashboard Cite

Background: Atrial fibrillation (AF) is associated with a higher risk of heart failure or death. AF may be episodic and patients with suspected AF are equipped with Holter ECG devices for several days. However, automated detection of AF in an ECG signal remains problematic, as was shown by the results of the PhysioNet Challenge 2017. Here, we introduce a simple yet robust logistic regression model for AF detection. Method: The detrended signal is filtered (1-35 Hz) and normalized. QRS detection based on envelograms (10-35 Hz) reveals QRS complexes. Five features are extracted from the ECG signal describing RR stability as well as the shape stability of areas preceding QRS complexes. Features were extracted for 1,517 recordings from the PhysioNet Challenge 2017 public dataset (758 AF recordings and 759 recordings with normal rhythm, other arrhythmia or noisy signal). The recordings were split in a 70/30 % ratio for the purposes of training and testing. Results: The results showed a sensitivity and specificity of 93 % and 90 %, respectively (AUC 0.96). The presented model was also tested on the MIT-AFDB public database, showing sensitivity and specificity of 89 % and 88 %, respectively. However, tests on an independent private dataset revealed lower specificity when pathologies which are not widely present in the training dataset are common in the tested ECG signal.

show abstract

Can Supervised Learning Be Used to Classify Cardiac Rhythms?

Abstract: Background:This

Cited by 11 publications

References 8 publications

Icentia11K: An Unsupervised Representation Learning Dataset for Arrhythmia Subtype Discovery

Icentia11K: An Unsupervised Representation Learning Dataset for Arrhythmia Subtype Discovery

Automatic Detection of Atrial Fibrillation Based on Continuous Wavelet Transform and 2D Convolutional Neural Networks

Shape Analysis of Consecutive Beats May Help in the Automated Detection of Atrial Fibrillation

Contact Info

Product

Resources

About