Lightweight ECG signal classification via linear law-based feature extraction

This paper introduces LLT-ECG, a novel semi-supervised method for electrocardiogram (ECG) signal classification that leverages principles from theoretical physics to generate features without relying on backpropagation or hyperparameter tuning. The method identifies linear laws that capture shared p...

Full description

Saved in:
Bibliographic Details
Main Authors: Péter Pósfay, Marcell T Kurbucz, Péter Kovács, Antal Jakovác
Format: Article
Language:English
Published: IOP Publishing 2025-01-01
Series:Machine Learning: Science and Technology
Subjects:
Online Access:https://doi.org/10.1088/2632-2153/ade6c3
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849428783564587008
author Péter Pósfay
Marcell T Kurbucz
Péter Kovács
Antal Jakovác
author_facet Péter Pósfay
Marcell T Kurbucz
Péter Kovács
Antal Jakovác
author_sort Péter Pósfay
collection DOAJ
description This paper introduces LLT-ECG, a novel semi-supervised method for electrocardiogram (ECG) signal classification that leverages principles from theoretical physics to generate features without relying on backpropagation or hyperparameter tuning. The method identifies linear laws that capture shared patterns within a reference class, enabling compact and verifiable representations of time series data. We evaluate the method on two PhysioNet datasets, TwoLeadECG and variable projection networks (VPNet). On TwoLeadECG, a minimal configuration—using only the linear law-based transformation (LLT) and a linear decision rule—reaches 73.1% accuracy using just two features. On VPNet, LLT-ECG combined with classifiers like k-nearest neighbors and support vector machines achieves up to 96.4% accuracy, comparable to deep learning models. These results highlight LLT-ECG’s promise for lightweight, interpretable, and high-performing ECG classification.
format Article
id doaj-art-e6ba87a663be4b4697c4491c1749b934
institution Kabale University
issn 2632-2153
language English
publishDate 2025-01-01
publisher IOP Publishing
record_format Article
series Machine Learning: Science and Technology
spelling doaj-art-e6ba87a663be4b4697c4491c1749b9342025-08-20T03:28:34ZengIOP PublishingMachine Learning: Science and Technology2632-21532025-01-016303500110.1088/2632-2153/ade6c3Lightweight ECG signal classification via linear law-based feature extractionPéter Pósfay0https://orcid.org/0000-0002-6769-3302Marcell T Kurbucz1https://orcid.org/0000-0002-0121-6781Péter Kovács2https://orcid.org/0000-0002-0772-9721Antal Jakovác3https://orcid.org/0000-0002-7410-0093Department of Computational Sciences, Institute for Particle and Nuclear Physics, HUN-REN Wigner Research Centre for Physics , 29-33 Konkoly-Thege Miklós Street, H-1121 Budapest, HungaryInstitute for Global Prosperity, The Bartlett, University College London , 149 Tottenham Court Road, W1T 7NE London, United KingdomDepartment of Numerical Analysis, Eötvös Loránd University , 1/c Pázmány Péter sétány, H-1117 Budapest, HungaryDepartment of Computational Sciences, Institute for Particle and Nuclear Physics, HUN-REN Wigner Research Centre for Physics , 29-33 Konkoly-Thege Miklós Street, H-1121 Budapest, Hungary; Department of Statistics, Institute of Data Analytics and Information Systems, Corvinus University of Budapest , 8 Fövám Square, H-1093 Budapest, HungaryThis paper introduces LLT-ECG, a novel semi-supervised method for electrocardiogram (ECG) signal classification that leverages principles from theoretical physics to generate features without relying on backpropagation or hyperparameter tuning. The method identifies linear laws that capture shared patterns within a reference class, enabling compact and verifiable representations of time series data. We evaluate the method on two PhysioNet datasets, TwoLeadECG and variable projection networks (VPNet). On TwoLeadECG, a minimal configuration—using only the linear law-based transformation (LLT) and a linear decision rule—reaches 73.1% accuracy using just two features. On VPNet, LLT-ECG combined with classifiers like k-nearest neighbors and support vector machines achieves up to 96.4% accuracy, comparable to deep learning models. These results highlight LLT-ECG’s promise for lightweight, interpretable, and high-performing ECG classification.https://doi.org/10.1088/2632-2153/ade6c3ECG classificationlinear lawrepresentation learninganomaly detectionmachine learning
spellingShingle Péter Pósfay
Marcell T Kurbucz
Péter Kovács
Antal Jakovác
Lightweight ECG signal classification via linear law-based feature extraction
Machine Learning: Science and Technology
ECG classification
linear law
representation learning
anomaly detection
machine learning
title Lightweight ECG signal classification via linear law-based feature extraction
title_full Lightweight ECG signal classification via linear law-based feature extraction
title_fullStr Lightweight ECG signal classification via linear law-based feature extraction
title_full_unstemmed Lightweight ECG signal classification via linear law-based feature extraction
title_short Lightweight ECG signal classification via linear law-based feature extraction
title_sort lightweight ecg signal classification via linear law based feature extraction
topic ECG classification
linear law
representation learning
anomaly detection
machine learning
url https://doi.org/10.1088/2632-2153/ade6c3
work_keys_str_mv AT peterposfay lightweightecgsignalclassificationvialinearlawbasedfeatureextraction
AT marcelltkurbucz lightweightecgsignalclassificationvialinearlawbasedfeatureextraction
AT peterkovacs lightweightecgsignalclassificationvialinearlawbasedfeatureextraction
AT antaljakovac lightweightecgsignalclassificationvialinearlawbasedfeatureextraction