Classification of Single-Lead Electrocardiograms: TDA Informed Machine Learning

Ignacio, Paul Samuel; Dunstan, Christopher; Escobar, Esteban; Trujillo, Luke; Uminsky, David

doi:10.1109/icmla.2019.00204

Cited by 15 publications

(12 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Our entry to this challenge leverages mathematically computable topological signatures of 12-lead ECGs as proxy for features in-formed by medical expertise to train a random forest model in a multi-class classification task. As has been shown for detecting Atrial Fibrillation using single-lead ECGs [6], this approach verifies the existence and viability of signal in the topology of ECGs for improving diagnosis of cardiac conditions. Upscaling this to the use of all 12 leads of a standard ECG to diagnose multiple heart conditions improves accessibility to automated diagnostics by reducing expert-dependent input in feature extraction.…”

Section: Introductionsupporting

confidence: 56%

“…We examine each point cloud embedding using topological data analysis, particularly via persistent homology [7]. A quick introduction to this approach with accompanying similar application is found in [6]. Succinctly, we induce a distance-parameterized monotonic sequence of abstract simplicial complexes revealing a multi-scale record of the evolving topological signatures of the underlying point cloud.…”

Section: Feature Extractionmentioning

confidence: 99%

See 1 more Smart Citation

A Topology Informed Random Forest Classifier for ECG Classification

Ignacio¹,

Bulauan²,

Manzanares³

2020

2020 Computing in Cardiology Conference (CinC)

View full text Add to dashboard Cite

This paper accompanies Team Cordi-Ak's entry to the classification of 12-lead ECGs for the Phys-ioNet/Computing in Cardiology Challenge 2020. Our approach leverages mathematically computable topological signatures of 12-lead ECGs as proxy for features informed by medical expertise to train a two-level random forest model in a multi-class classification task. We view ECGs as multivariate time series data and convert different segments and groupings of leads to point cloud embeddings. This stores both local and global structures of ECGs, and encodes periodic information as attractor cycles in high-dimensional space. We then employ topological data analysis on these embeddings to extract topological features based on different summaries available in the literature. We supplement these features with demographic data and statistical moments of RR intervals based on the Pan-Tompkins algorithm for each lead to train the classifier. Our classifier achieved a challenge validation score of 0.304, and a final score of -0.113 on the full hidden test data, placing us 37th out of 41 officially ranked teams that participated in this year's Challenge.

show abstract

Section: Introductionsupporting

confidence: 56%

Section: Feature Extractionmentioning

confidence: 99%

A Topology Informed Random Forest Classifier for ECG Classification

Ignacio¹,

Bulauan²,

Manzanares³

2020

2020 Computing in Cardiology Conference (CinC)

View full text Add to dashboard Cite

show abstract

“…A difficulty in this is that it may be difficult to know what a “correct” subset of a signal is in order for it to be considered a P,Q,S, or T-wave. Additionally, information about optimal 1-cycles identified as P,Q,S, and T-waves could be combined with other existing approaches such as analyzing other persistent homology statistics, wavelet decompositions, and machine learning for automated arrhythmia detection in future work [12] [13] [17] [19] [21].…”

Section: Resultsmentioning

confidence: 99%

“…Topological data analysis (TDA) is concerned with the study of shapes constructed from a dataset which are invariant under continuous deformations such as stretching and twisting. Applications of TDA to ECG signals have used persistence statistics [20] [18], fractal dimension [18], and machine learning [17] [19]. Using cycle reconstructions has shown utility in various applications outside of ECG analysis such as analyzing structures on the atomic scale [24] and in structural engineering [25].…”

Section: Introductionmentioning

confidence: 99%

“…Thus the analysis of ECGs is important for accurate diagnosis and proper treatment of cardiovascular diseases. Several approaches to automated ECG analysis have been performed, including machine learning [4] [5] [6] [7] [8] [9] [10] [11], wavelet transforms [12] [13] [14] [15] [16], and persistent homology [17] [18] [19] [20] [21] [22] [23]. Due to the high accuracy required of ECG-analysis software and the fact that the bulk of ECG analysis is carried out by healthcare providers, the development of algorithms that identifies P,Q,R,S, and T-waves, measures intervals of interest, and/or detects arrhythmias is an active area of research.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Electrocardiogram feature extraction and interval measurements using optimal representative cycles from persistent homology

Dlugas

2022

Preprint

View full text Add to dashboard Cite

Cardiovascular diseases are among the leading causes of death, and their early detection and treatment is important for lowering their prevalence and mortality rate. Electrocardiograms (ECGs) record electrical activity of the heart to provide information used to diagnose and treat various cardiovascular diseases. Many approaches to computer-aided ECG analysis have been performed, including Fourier analysis, principal component analysis, analyzing morphological changes, and machine learning. Due to the high accuracy required of ECG-analysis software, there is no universally-agreed upon algorithm to identify P,Q,R,S, and T-waves and measure intervals of interest. Topological data analysis uses tools from algebraic topology to quantify hole-like shapes within data, and methods using persistence statistics and fractal dimension with machine learning have been applied to ECG signals in the context of detecting arrhythmias within recent years. To our knowledge, there does not exist a method of identifying P,Q,S, and T-waves and measuring intervals of interest which relies on topological features of the data, and we propose a novel topological method for performing these aspects of ECG analysis. Specifically, we establish criteria to identify cardinality-minimal and area-minimal 1-cycles with certain properties as P,Q,S, and T-waves. This yields a procedure for measuring the PR-interval, QT-interval, ST-segment, QRS-duration, P-wave duration, and T-wave duration in Lead II ECG data. We apply our procedure to 400 sets of simulated Lead II ECG signals and compare with the interval values set by the model. Additionally, the algorithm is used to identify cardinality-minimal and area-minimal 1-cycles as P,Q,S, and T-waves in two sets of 200 randomly sampled Lead II ECG signals of real patients with 11 common rhythms. Analysis of optimal 1-cycles identified as P,Q,S, and T-waves and comparison of interval measurements shows that 1-cycle reconstructions can provide useful information about the ECG signal and could hold utility in characterizing arrhythmias.

show abstract