pith. sign in

arxiv: 2604.15822 · v1 · submitted 2026-04-17 · 💻 cs.LG · cs.AI· cs.CE· cs.NE· eess.SP

ECG-Lens: Benchmarking ML & DL Models on PTB-XL Dataset

Saloni Garg , Ukant Jadia , Amit Sagtani , Kamal Kant Hiran This is my paper

Pith reviewed 2026-05-10 08:31 UTC · model grok-4.3

classification 💻 cs.LG cs.AIcs.CEcs.NEeess.SP
keywords modelsclassificationsignalslearningmodelaccuracyautomatedclassifier
0
0 comments X

The pith

ECG-Lens, a complex CNN, achieves 80% accuracy and 90% ROC-AUC on PTB-XL ECG classification, outperforming Decision Tree, Random Forest, Logistic Regression, simple CNN, and LSTM.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The work tests six models on raw 12-lead ECG recordings from the PTB-XL dataset, which covers normal hearts and several cardiac conditions. Three traditional machine learning algorithms—Decision Tree Classifier, Random Forest Classifier, and Logistic Regression—are compared against three deep learning models: a simple CNN, an LSTM network, and a more complex CNN architecture named ECG-Lens. All deep learning models receive the raw signals directly and learn features automatically. To increase training variety, the authors apply Stationary Wavelet Transform augmentation that aims to preserve core signal traits while adding diversity. Evaluation uses accuracy, precision, recall, F1-score, and ROC-AUC. The ECG-Lens model records the top scores at 80% accuracy and 90% ROC-AUC, indicating that deeper convolutional structures handle the raw multi-lead data more effectively than the simpler networks or classical algorithms in this benchmark.

Core claim

The ECG-Lens model achieved the highest performance, with 80% classification accuracy and a 90% ROC-AUC. These findings demonstrate that deep learning architectures, particularly complex CNNs substantially outperform traditional ML methods on raw 12-lead ECG data.

Load-bearing premise

That Stationary Wavelet Transform augmentation increases diversity without distorting essential ECG characteristics and that the reported metrics reflect genuine generalization rather than overfitting or favorable data splits.

read the original abstract

Automated classification of electrocardiogram (ECG) signals is a useful tool for diagnosing and monitoring cardiovascular diseases. This study compares three traditional machine learning algorithms (Decision Tree Classifier, Random Forest Classifier, and Logistic Regression) and three deep learning models (Simple Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), and Complex CNN (ECGLens)) for the classification of ECG signals from the PTB-XL dataset, which contains 12-lead recordings from normal patients and patients with various cardiac conditions. The DL models were trained on raw ECG signals, allowing them to automatically extract discriminative features. Data augmentation using the Stationary Wavelet Transform (SWT) was applied to enhance model performance, increase the diversity of training samples, and preserve the essential characteristics of the ECG signals. The models were evaluated using multiple metrics, including accuracy, precision, recall, F1-score, and ROC-AUC. The ECG-Lens model achieved the highest performance, with 80% classification accuracy and a 90% ROC-AUC. These findings demonstrate that deep learning architectures, particularly complex CNNs substantially outperform traditional ML methods on raw 12-lead ECG data, and provide a practical benchmark for selecting automated ECG classification models and identifying directions for condition-specific model development.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 1 invented entities

No mathematical axioms or derivations are present. The only new element is the named ECG-Lens architecture, which functions as an empirical model variant without independent falsifiable evidence beyond the reported metrics.

invented entities (1)
  • ECG-Lens no independent evidence
    purpose: Complex CNN architecture for ECG classification
    Presented as the top-performing model in the benchmark.

pith-pipeline@v0.9.0 · 5543 in / 1201 out tokens · 35658 ms · 2026-05-10T08:31:26.610728+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

24 extracted references · 24 canonical work pages

  1. [1]

    Machine Learning Implementation a) Decision Tree Decision Trees can be used to recursively partition the feature space in order to make accurate classifications, and they are particularly effective for identifying intricate non - linear patterns in ECG data [17]. The fundamental idea behind Decision Trees involves dividing the feature space based on the f...

    work page

  2. [2]

    Deep Learning Implementation Convolutional Neural Networks (CNNs) are a type of deep learning model that is particularly well-suited for processing and analyzing large data. The key components of a CNN are: Convolutional Layers (Conv1D): A series of trainable filters are applied to the input data by these layers, which enables the model to extract importa...

    work page

  3. [3]

    Performance Matrices The performance metrics used in this study consist of a comprehensive set of measures, including accuracy, area under the curve (ROC-AUC), F1-score, precision, and recall. These metrics are highly utilized and recognized in the domain of machine learning and deep learning, enabling a thorough evaluation of the model ’s performance fro...

    work page

  4. [4]

    Towards AI Based Diagnosis of Rheumatic Heart Disease: Data Annotation and View Classification,

    L. Mugambi, L. ZÜHLKE and C. W. Maina, "Towards AI Based Diagnosis of Rheumatic Heart Disease: Data Annotation and View Classification," 2022 IST -Africa Conference (IST -Africa), Ireland, 2022, pp. 1-8, doi: 10.23919/IST-Africa56635.2022.9845657

    work page doi:10.23919/ist-africa56635.2022.9845657 2022

  5. [5]

    Diabetes and cardiovascular disease,

    B. V. Howard and M. F. Magee, “Diabetes and cardiovascular disease,” Current Atherosclerosis Reports, vol. 2, no. 6, pp. 476–481, Nov. 2000, doi: https://doi.org/10.1007/s11883-000-0046-8

    work page doi:10.1007/s11883-000-0046-8 2000

  6. [6]

    Study of a new model of normal ECG wave,

    A. Dutta and S. C. Bera, "Study of a new model of normal ECG wave," 2014 First International Conference on Automation, Control, Energy , and Systems (ACES), Adisaptagram, India, 2014, pp. 1 -4, doi: 10.1109/ACES.2014.6808008

    work page doi:10.1109/aces.2014.6808008 2014

  7. [7]

    Detection of abnormal heart conditions based on characteristics of ECG signals,

    M. Hammad, A. Maher, K. Wang, F. Jiang, and M. Amrani, “Detection of abnormal heart conditions based on characteristics of ECG signals,” Measurement, vol. 125, pp. 634 –644, Sep. 2018, doi: https://doi.org/10.1016/j.measurement.2018.05.033

    work page doi:10.1016/j.measurement.2018.05.033 2018

  8. [8]

    A Machine Learning Model for the Early Prediction of Cardiovascular Disease in Patients,

    Abhishek, H. V. Bhagat , and M. Singh, "A Machine Learning Model for the Early Prediction of Cardiovascular Disease in Patients," 2023 Second International Conference on Advances in Computational Intelligence and Communication (ICACIC), Puducherry, India, 2023, pp. 1-5, doi: 10.1109/ICACIC59454.2023.10435210

    work page doi:10.1109/icacic59454.2023.10435210 2023

  9. [9]

    Recommendation to Use Wearable -Based mHealth in Closed-Loop Management of Acute Cardiovascular Disease Patients During the COVID-19 Pandemic,

    N. Ji et al., "Recommendation to Use Wearable -Based mHealth in Closed-Loop Management of Acute Cardiovascular Disease Patients During the COVID-19 Pandemic," in IEEE Journal of Biomedical and Health Informatics, vol. 25, no. 4, pp. 903 -908, April 2021, doi : 10.1109/JBHI.2021.3059883

    work page doi:10.1109/jbhi.2021.3059883 2021

  10. [10]

    Scalogram Based Heart Disease Classification using Hybrid CNN -Naive Bayes Classifier,

    A. S. B., S. S., S. S. R., A. R. Nair and M. Raju, "Scalogram Based Heart Disease Classification using Hybrid CNN -Naive Bayes Classifier," 2022 International Conference on Wireless Communications Signal Processing and Networking (WiSPNET), Chennai, India, 2022, pp. 345 -348, doi: 10.1109/WiSPNET54241.2022.9767153

    work page doi:10.1109/wispnet54241.2022.9767153 2022

  11. [11]

    Explaining ECG Biometrics: Is It All In The QRS?,

    J. R. Pinto and J. S. Cardoso, "Explaining ECG Biometrics: Is It All In The QRS?," 2020 International Conference of the Biometrics Special Interest Group (BIOSIG), Darmstadt, Germany, 2020, pp. 1-5

    work page 2020

  12. [12]

    Wagner, P., Strodthoff, N., Bousseljot, R., Samek, W., & Schaeffter, T. (2022). PTB -XL, a large publicly available electrocardiography dataset (version 1.0.3). PhysioNet. https://doi.org/10.13026/kfzx- aw45

    work page doi:10.13026/kfzx- 2022

  13. [13]

    R., Fujita, H., Lih, O

    Acharya, U. R., Fujita, H., Lih, O. S., Hagiwara, Y., Tan, J. H., & Adam, M. (2017). Automated detection of arrhythmias using different intervals of tachycardia ECG segments with a convolutional neural network. Information Sciences, 405, 81-90

    work page 2017

  14. [14]

    Arrhythmia detection using a deep convolutional neural network with long duration ECG signals,

    Ö. Yıldırım, P. Pławiak, R. -S. Tan, and U. R. Acharya, "Arrhythmia detection using a deep convolutional neural network with long duration ECG signals," Computers in Biology and Medicine, vol. 102, pp. 411– 420, Nov. 2018, doi: 10.1016/j.compbiomed.2018.10.009

    work page doi:10.1016/j.compbiomed.2018.10.009 2018

  15. [15]

    Smigiel, S., Palczynski, K., & Ledzinski, D. (2021). ECG Signal Classification Using Deep Learning Techniques Based on the PTB-XL Dataset. Sensors, 21(10), 3291

    work page 2021

  16. [16]

    Classification of ECG signals using machine learning techniques: A comprehensive review,

    S. H. Jambukia, V. K. Dabhi, and H. B. Prajapati, "Classification of ECG signals using machine learning techniques: A comprehensive review," Expert Syst. Appl., vol. 145, p. 113130, Apr. 2020, doi: 10.1016/j.eswa.2019.113130

    work page doi:10.1016/j.eswa.2019.113130 2020

  17. [17]

    Luz, E. J. S., Schwartz, W. R., Cámara -Chávez, G., & Menotti, D. (2016). ECG-based heartbeat classification for arrhythmia detection: A survey. Computer Methods and Programs in Biomedicine, 127, 144 - 164

    work page 2016

  18. [18]

    Y., Haghpanahi, M., Bourn, C., & Ng, A

    Rajpurkar, P., Hannun, A. Y., Haghpanahi, M., Bourn, C., & Ng, A. Y. (2017). Cardiologist -level arrhythmia detection with convolutional neural networks. arXiv preprint arXiv:1707.01836

    work page arXiv 2017

  19. [19]

    Stationary wavelet transform based ECG signal denoising method,

    A. Kumar, H. Tomar, V. K. Mehla, R. Komaragiri, and M. Kumar, “Stationary wavelet transform based ECG signal denoising method,” ISA Transactions , Dec. 2020, doi: https://doi.org/10.1016/j.isatra.2020.12.029

    work page doi:10.1016/j.isatra.2020.12.029 2020

  20. [20]

    Classification and prediction of ECG data based on decision trees and their optimization algorithms,

    R. Xing, "Classification and prediction of ECG data based on decision trees and their optimization algorithms," 2023 International Conference on Computers, Information Processing and Advanced Education (CIPAE), Ottawa, ON, Canada, 2023, pp. 434 -439, doi: 10.1109/CIPAE60493.2023.00089

    work page doi:10.1109/cipae60493.2023.00089 2023

  21. [21]

    Malik, P

    U. J. Khan, A. Oberoi and J. Gill, "Hybrid Classfication for Heart Disease Prediction using Artificial Intelligence," 2021 5th International Conference on Computing Methodologies and Communication (ICCMC), Erode, India, 2021, pp. 1779 -1785, doi: 10.1109/ICCMC51019.2021.9418345

    work page doi:10.1109/iccmc51019.2021.9418345 2021

  22. [22]

    ECG Biometric Recognition: Review, System Proposal, and Benchmark Evaluation,

    P. Melzi, R. Tolosana , and R. Vera -Rodriguez, "ECG Biometric Recognition: Review, System Proposal, and Benchmark Evaluation," in IEEE Access, vol. 11, pp. 15555 -15566, 2023, doi: 10.1109/ACCESS.2023.3244651

    work page doi:10.1109/access.2023.3244651 2023

  23. [23]

    International Journal of Ad- vanced Computer Science and Applications, 12(6), 599–606

    Ž. Ð. Vujovic, “Classification Model Evaluation Metrics,” International Journal of Advanced Computer Science and Applications, vol. 12, no. 6, 2021, doi: https://doi.org/10.14569/ijacsa.2021.0120670

    work page doi:10.14569/ijacsa.2021.0120670 2021

  24. [24]

    Classification of ECG signals using machine learning techniques: A survey,

    S. H. Jambukia, V. K. Dabhi , and H. B. Prajapati, "Classification of ECG signals using machine learning techniques: A survey," 2015 International Conference on Advances in Computer Engineering and Applications, Ghaziabad, India, 2015, pp. 714 -721, doi: 10.1109/ICACEA.2015.7164783

    work page doi:10.1109/icacea.2015.7164783 2015