ECG Benchmark

ECGBench

Reproducible ECG benchmark data from open-access PhysioNet datasets. A PyTorch-native library with standardised fold splits, quality-checked metadata, and a curated index of 64 publicly available ECG datasets.

64
Total Datasets
24
12-Lead (PhysioNet)
15
12-Lead (Other)
10
2-Lead
10
1-Lead
11M+
Max Records
13+
Countries

Installation
Install ECGBench from PyPI or source to start loading ECG benchmark data in PyTorch

Quick Install

# Using pip
pip install ecgbench

# Using uv (recommended)
uv pip install ecgbench

From Source (Development)

git clone https://github.com/vlbthambawita/ECGBench.git
cd ECGBench
uv pip install -e ".[dev]"
Quick Start
Load ECG data with a standardised PyTorch Dataset in just a few lines. Fold assignments are downloaded from huggingface.co/datasets/vlbthambawita/ECGBench; the waveforms stay on your disk, since the source datasets are not redistributed.

Load a split and iterate batches

from ecgbench import ECGDataset, ecg_collate_fn
from torch.utils.data import DataLoader

# PTB-XL train split (folds 1-8), at 100 Hz
dataset = ECGDataset(
    "ptbxl",
    split="train",
    data_path="/path/to/ptb-xl/1.0.3/",
    sampling_rate=100,
    labels=True,          # attach diagnoses, demographics, report
)

loader = DataLoader(dataset, batch_size=32, collate_fn=ecg_collate_fn)

for batch in loader:
    signals = batch["signal"]      # (B, 12, 1000) float32, millivolts
    ids     = batch["record_id"]   # tensor of ecg_id values
    labels  = batch["labels"]      # list of per-record dicts
    break

Choose leads, units and a sample window

# Leads by NAME, not index: the order differs between datasets, so
# signal[4] is aVL in PTB-XL and aVF in MIMIC-IV-ECG.
ds = ECGDataset(
    "ptbxl", split="train", data_path="/path/to/ptb-xl/1.0.3/",
    leads=["I", "II", "V5"], units="uV",
)

# window=(start, length) in samples is read at load time, so the rest of
# the record is never decoded. Here: the first 5 s of the 500 Hz signal.
ds = ECGDataset(
    "ptbxl", split="train", data_path="/path/to/ptb-xl/1.0.3/",
    sampling_rate=500, window=(0, 2500),
)

Subsample folds, or roll your own cross-validation

# One fold instead of the whole split — useful on the large datasets.
fold1 = ECGDataset(
    "ptbxl", split="train", data_path="/path/to/ptb-xl/1.0.3/", fold_numbers=[1],
)

# split=None selects by fold number across the default train/val/test
# boundaries, which is the only way to express a custom rotation.
held_out = ECGDataset(
    "ptbxl", split=None, fold_numbers=[10], data_path="/path/to/ptb-xl/1.0.3/",
)

Dataset Catalogue

12-Lead ECG Datasets

24 datasets
# Dataset Format Patients Records Access Status Origin Paper Related
1 PTB-XL physionet.org ↗ 12-lead · 10 s · 500 Hz (also 100 Hz) 18,869 21,799 Open
CC BY 4.0
Completed Physikalisch-Technische Bundesanstalt
Germany
PTB-XL: A Large Publicly Available ECG Dataset
2 PTB-XL+ physionet.org ↗ features & annotations for PTB-XL · no raw ECGs 18,869 21,799 Open
CC BY 4.0
Completed Karlsruhe Institute of Technology
Germany
PTB-XL+: A Comprehensive ECG Feature Dataset
3 PTB Diagnostic ECG Database physionet.org ↗ 15-lead (12 + 3 Frank) · 32-120 s · 1,000 Hz 290 549 Open
ODC-By 1.0
Completed Univ. Clinic Benjamin Franklin
Germany — Berlin
Bousseljot et al., Biomedizinische Technik, 1995
4 MIMIC-IV-ECG physionet.org ↗ 12-lead · 10 s · 500 Hz 161,352 800,035 Credentialed
PhysioNet DUA
Completed Beth Israel Deaconess Medical Center
USA — Boston, MA
Gow et al.
5 MIMIC-IV-ECG Demo physionet.org ↗ 12-lead · 10 s · 500 Hz · WFDB 92 659 Open
ODbL 1.0
Completed Beth Israel Deaconess Medical Center
USA — Boston, MA
Dataset DOI
6 MIMIC-IV-ECG-Ext-ICD physionet.org ↗ ICD-10-CM diagnoses for MIMIC-IV-ECG · no raw ECGs 161,352 800,035 Credentialed
PhysioNet Credentialed Health Data License 1.5.0
Completed University of Oldenburg + Charité Berlin
Germany
Prospects for AI-enhanced ECG as a unified screening tool for cardiac and non-cardiac conditions
7 Chapman-Shaoxing (Arrhythmia) physionet.org ↗ 12-lead · 10 s · 500 Hz · WFDB 45,152 45,152 Open
CC BY 4.0
Completed Chapman Univ.; Shaoxing People's Hospital & Ningbo First Hospital
China / USA
Zheng et al., Scientific Reports, 2020
8 St Petersburg INCART 12-Lead Arrhythmia Database physionet.org ↗ 12-lead · 30 min · 257 Hz 32 75 Open
ODC-By 1.0
Completed St. Petersburg Institute of Cardiological Technics (INCART)
Russia
Dataset DOI
9 Lobachevsky University ECG Database (LUDB) physionet.org ↗ 12-lead · 10 s · 500 Hz · manually annotated waves 200 200 Open
ODC-By 1.0
Completed Nizhny Novgorod City Hospital No. 5 / Lobachevsky University
Russia
Kalyakulina et al., IEEE Access, 2020
10 Brugada-HUCA physionet.org ↗ 12-lead · 12 s · 100 Hz 363 363 Open
CC BY-SA 4.0
Completed Hospital Universitario Central de Asturias (HUCA)
Spain
Costa Cortez & Garcia Iglesias, PhysioNet, 2026
11 KURIAS-ECG physionet.org ↗ 12-lead · 10 s · 500 Hz · SNOMED CT + OMOP-CDM 13,862 20,000 Restricted
Pending audit
Unavailable Korea University Anam Hospital
South Korea — Seoul
Dataset DOI
12 Leipzig Heart Center ECG Database physionet.org ↗ 12-lead + intracardiac EGM · 14-20 ch · 78 s-2.5 h · 977 Hz 39 39 Open
ODC-By 1.0
Completed Leipzig Heart Center
Germany
Leipzig Heart Center ECG-Database: Arrhythmias in children and patients with congenital heart disease
13 Norwegian Endurance Athlete ECG Database physionet.org ↗ 12-lead · 10 s · 500 Hz 28 28 Open
CC BY 4.0
Completed University of Oslo
Norway
Singstad, PhysioNet, 2022
14 MHD Effect on 12-Lead ECGs in MRI Scanners physionet.org ↗ 12-lead + 3-lead · 24 s–12 min · 1,024 Hz 26 53 Open
ODC-By 1.0
Completed Otto-von-Guericke University of Magdeburg
Germany
Krug et al., CinC 2017
15 Wilson Central Terminal ECG Database physionet.org ↗ 37 ch (I/II/III + V1–V6 + LA/RA/LL + UV1–UV6, raw & filtered, + WCT) · 10 s · 800 Hz 92 540 Open
ODC Attribution
Completed MARCS Institute, Western Sydney Univ.; Campbelltown Hospital
Australia
Machines, 2016
16 CiPA ECG Validation Study physionet.org ↗ 12-lead · 10 s · 1000 Hz · + derived median beats 60 5,749 segments Open
ODC Attribution
Completed US FDA / CDER; Phase I Clinical Pharmacology Study (NCT03070470)
USA
Clin Pharmacol Ther, 2019
17 ECG Effects of Dofetilide, Moxifloxacin and Combinations (ECGDMMLD) physionet.org ↗ 12-lead · 10 s · 500 Hz → 1 kHz · + derived median beats 22 4,211 segments Open
ODC Attribution
Completed US FDA / CDER; Phase I Crossover Study (NCT02308748)
USA
Clin Pharmacol Ther, 2016
18 ECG Effects of Ranolazine, Dofetilide, Verapamil, Quinidine (ECGRDVQ) physionet.org ↗ 12-lead · 10 s · 500 Hz → 1 kHz · + derived median beats 22 5,232 segments Open
ODC Attribution
Completed US FDA / CDER; Phase I Crossover Study (NCT01873950)
USA
Clin Pharmacol Ther, 2014
19 Eye Tracking Dataset for 12-Lead ECG Interpretation physionet.org ↗ gaze metrics over 10 printed ECGs · no raw ECGs 63 readers 630 reader sessions Open
ODC ODbL 1.0
Completed Qatar Biomedical Research Institute, Hamad bin Khalifa Univ.
Qatar
Understanding Cardiology Practitioners' Interpretations of Electrocardiograms: An Eye-Tracking Study
20 EchoNext physionet.org ↗ 12-lead · 10 s · 250 Hz · NumPy arrays, z-scored 36,286 100,000 (82,543 split) Restricted
PhysioNet Restricted Health Data License 1.5.0
Completed Columbia University Irving Medical Center
USA — New York, NY
Poterucha et al., Nature, 2025
21 Symile-MIMIC physionet.org ↗ multimodal cohort over MIMIC-IV-ECG · ECG + CXR + labs 9,573 11,622 Credentialed
PhysioNet Credentialed Health Data License 1.5.0
Completed New York University
USA
Contrasting with Symile: Simple Model-Agnostic Representation Learning for Unlimited Modalities
22 PhysioNet/CinC Challenge 2020 physionet.org ↗ 12-lead · 5–1,800 s · 257/500/1,000 Hz 43,101 Open
CC BY 4.0
Completed CPSC, INCART, PTB, PTB-XL, Georgia
Multi-national (China, Russia, Germany, USA)
Alday et al., Physiological Measurement, 2020
23 PhysioNet/CinC Challenge 2021 physionet.org ↗ 12-lead · 5–1,800 s · 257/500/1,000 Hz 88,253 Open
CC BY 4.0
Completed CPSC, INCART, PTB, PTB-XL, Georgia, Chapman-Shaoxing, Ningbo
Multi-national (China, Russia, Germany, USA)
Reyna et al., Computing in Cardiology, 2021
24 STAFF III Database physionet.org ↗ 9-lead (12-lead derivable) · 94.5–960 s · 1,000 Hz · 0.625 µV resolution · WFDB 104 520 Open
ODC Attribution
Completed Charleston Area Medical Center; Lund University
USA / Sweden
Martínez et al., CinC 2017

12-Lead ECG Datasets (Other Repositories)

15 datasets
# Dataset Format Patients Records Access Status Origin Paper Related
1 CPSC 2018 (China Physiological Signal Challenge 2018) icbeb.org ↗ 12-lead · 6–144 s · 500 Hz · WFDB (.hea/.mat) 6,877 Open
CC BY 4.0
Completed 11 hospitals (ICBEB, Nanjing)
China
Liu et al., J. Med. Imaging Health Inform., 2018
2 Chapman-Shaoxing ECG Database (10,646 patients) figshare.com ↗ 12-lead · 10 s · 500 Hz · CSV (µV) 10,646 10,646 Open
CC0 1.0
Completed Chapman University; Shaoxing People's Hospital
China / USA
Zheng et al., Scientific Data, 2020
3 Ningbo First Hospital ECG Database (Idiopathic Ventricular Arrhythmia) figshare.com ↗ 12-lead · 2.9–59 s · 2000 Hz · CSV 334 334 Open
CC BY 4.0
Completed Chapman University; Ningbo First Hospital, Zhejiang University
China / USA
Zheng et al., Scientific Data, 2020
4 Shandong Provincial Hospital ECG Database (SPHDB) figshare.com ↗ 12-lead · 10–56 s · 500 Hz · HDF5 24,666 25,770 Open
CC BY 4.0
Completed Shandong Provincial Hospital
China
Liu et al., Scientific Data, 2022
5 CODE-15% (Telehealth Network of Minas Gerais, 15% subset) zenodo.org ↗ 12-lead · 10.24 s · 400 Hz · HDF5 233,770 345,779 Open
CC BY 4.0
Completed Telehealth Network of Minas Gerais (TNMG)
Brazil
Ribeiro et al., Nature Communications, 2020
6 CODE-test (827-record hold-out test set) zenodo.org ↗ 12-lead · 7–10 s padded to 10.24 s · 400 Hz · HDF5 827 827 Open
CC BY 4.0
Completed Universidade Federal de Minas Gerais / TNMG
Brazil
Ribeiro et al., Nature Communications, 2020
7 CODE (Full Dataset, ~2.3M records) scilifelab.se ↗ 12-lead · 400 Hz · HDF5 ~1,676,384 ~2,322,513 Restricted
DUA required
Not started Telehealth Network of Minas Gerais (TNMG)
Brazil
Ribeiro et al., Nature Communications, 2020
8 SaMi-Trop (Chagas Cardiomyopathy Cohort) zenodo.org ↗ 12-lead · 10.24 s · 400 Hz · HDF5 1,631 1,631 Open
CC BY 4.0
Completed UFMG; Uppsala University; EPFL
Brazil / Sweden / Switzerland
Lima et al., medRxiv, 2021
9 IKEM Dataset (Institute for Clinical and Experimental Medicine, Prague) zenodo.org ↗ 8 reduced leads · 8.192 s · 500 Hz · HDF5 30,290 98,130 Open
CC BY-NC-ND 4.0
Completed IKEM (Institute for Clinical and Experimental Medicine)
Czech Republic — Prague
Seják et al., Knowledge-Based Systems, 2023
10 MedalCare-XL (Synthetic 12-Lead ECGs from Simulations) zenodo.org ↗ 12-lead · 10 s · 500 Hz · CSV (leads in rows, no header) · raw/noise/filtered variants 0 (synthetic) 16,842 Open
CC BY 4.0
Completed Medical Univ. of Graz; KIT; PTB; Univ. of Edinburgh
Austria / Germany / UK
Gillette et al., Scientific Data, 2023
11 Harvard-Emory ECG Database (HEEDB) bdsp.io ↗ 12-lead · 10 s · 250/500 Hz · WFDB 2,167,795 11,607,261 Credentialed
DUA (BDSP)
Not started Massachusetts General Hospital; Emory University Hospital
USA — Boston & Atlanta
Koscova et al., Scientific Data, 2026
12 Nightingale BWH Emergency Dept ECG Dataset ngsci.org ↗ 12-lead · 100 Hz · NumPy arrays 30,933 103,952 Restricted
Institutional credentials
Not started Brigham and Women's Hospital
USA — Boston, MA
Mullainathan & Obermeyer, QJE, 2021
13 Nightingale NTUH Cardiac Arrest ECG Dataset ngsci.org ↗ 12-lead · ~500 Hz · XML/array 10,950 18,072 Restricted
Institutional credentials
Not started National Taiwan University Hospital, Emergency Dept
Taiwan
Obermeyer et al., Nature Medicine, 2022
14 GU-ECG (Gazi University, PTCA-induced Ischaemia) mendeley.com ↗ 12-lead continuous · 8,800 Hz · 24-bit · .ekg format 74 222 Open
CC BY 4.0
Not started Gazi University Faculty of Medicine; Bilkent University
Turkey
Dataset DOI
15 ZZU pECG (Zhengzhou University Pediatric ECG Database) figshare.com ↗ 12-lead + 9-lead · 5-120 s · 500 Hz · WFDB 11,643 children 14,190 Open
CC BY 4.0
Completed First Affiliated Hospital of Zhengzhou University
China
Tan et al., Scientific Data, 2025

2-Lead ECG Datasets

12 datasets
# Dataset Format Patients Records Access Status Origin Paper Related
1 MIT-BIH Arrhythmia Database physionet.org ↗ 2-lead (usually MLII + V1) · 30 min · 360 Hz · WFDB 47 48 Open
ODC-By 1.0
Completed Beth Israel Hospital / MIT
USA
Moody & Mark, IEEE EMBS 2001
2 MIT-BIH Atrial Fibrillation Database physionet.org ↗ 2-lead (ECG1 + ECG2, unnamed) · 10 h · 250 Hz · WFDB 25 25 (23 with signals) Open
ODC-By 1.0
Completed Beth Israel Hospital
USA
Moody & Mark, CinC 1983
3 Long-Term AF Database (LTAFDB) physionet.org ↗ 2-lead (both named ECG) · 6.1–26.4 h · 128 Hz · WFDB 84 84 Open
ODC-By 1.0
Completed Northwestern University; MEDICALgorithmics
USA / Poland
Petrutiu et al., Europace 2007
4 MIT-BIH Normal Sinus Rhythm Database physionet.org ↗ 2-lead (ECG1 + ECG2, unnamed) · 23.1–26.0 h · 128 Hz · WFDB 18 18 Open
ODC-By 1.0
Completed Beth Israel Hospital
USA
No database paper — cite PhysioNet
5 MIT-BIH Supraventricular Arrhythmia Database physionet.org ↗ 2-lead (ECG1 + ECG2, unnamed) · 30 min · 128 Hz · WFDB 78 Open
ODC-By 1.0
Completed Harvard-MIT HST
USA
Greenwald, PhD thesis, Harvard-MIT 1990
6 European ST-T Database (EDB) physionet.org ↗ 2-lead ambulatory · 15 lead layouts · 2 h · 250 Hz · WFDB 79 90 Open
ODC-By 1.0
Completed CNR Institute of Clinical Physiology, Pisa; European Society of Cardiology
Italy
Taddei et al., Eur Heart J 1992
7 BIDMC Congestive Heart Failure Database physionet.org ↗ 2-lead (ECG1 + ECG2, unnamed) · 19.8–20.0 h · 250 Hz · WFDB 15 15 Open
ODC-By 1.0
Completed Beth Israel Hospital
USA — Boston, MA
Baim et al., Survival of patients with severe congestive heart failure treated with oral milrinone, J Am Coll Cardiol 1986
8 Sudden Cardiac Death Holter Database physionet.org ↗ 2-lead (ECG1 + ECG2, both unnamed) · 3.9–25.1 h · 250 Hz · WFDB 23 23 Open
ODC Attribution
Completed MIT · Boston-area hospitals
USA — Boston, MA
Greenwald, Development and analysis of a ventricular fibrillation detector, MS thesis, MIT 1986
9 QT Database (QTDB) physionet.org ↗ 2-lead · 20 lead layouts · 15 min · 250 Hz · WFDB 103 105 Open
ODC-By 1.0
Completed Universidad de Zaragoza; MIT; Beth Israel Deaconess Medical Center
Spain / USA
Laguna et al., Comput Cardiol 1997
10 SHDB-AF (Saitama Holter Database — Atrial Fibrillation) physionet.org ↗ 2-lead (mod. CC5 + NASA) · 9-24 h · 200 Hz · WFDB 122 128 Open
ODC-By 1.0
Completed Saitama Medical University International Medical Center; Technion
Japan
Tsutsui et al., Scientific Data 2025
11 MIT-BIH ST Change Database physionet.org ↗ 2-lead on 18 records, 1-lead on 10 (ECG1/ECG2, both unnamed) · 13.1–67.2 min · 360 Hz · WFDB · mostly exercise stress ECGs 28 Open
ODC Attribution
Completed MIT
USA
Albrecht, S-T segment characterization for long-term automated ECG analysis, MS thesis, MIT 1983
12 Long-Term ST Database (LTSTDB) physionet.org ↗ 2- or 3-lead ambulatory · 12 lead layouts · 17–48 h · 250 Hz · WFDB · 3 sets of ST episode annotations 80 86 Open
ODC-By 1.0
Completed University of Ljubljana; CNR Institute of Clinical Physiology, Pisa; MIT
Multi-national (Slovenia, Italy, USA)
Jager et al., Med Biol Eng Comput 2003

1-Lead ECG Datasets

10 datasets
# Dataset Format Patients Records Access Status Origin Paper Related
1 Icentia11k Single Lead Continuous ECG physionet.org ↗ 1-lead (modified Lead I) · ~70 min/seg · 250 Hz 11,000 541,794 segments Open
CC BY-NC-SA 4.0
Not started Université de Montréal; Icentia Inc.
Canada
Tan et al., CinC 2021
2 PhysioNet/CinC Challenge 2017 (AF Classification) physionet.org ↗ 1-lead (ECG) · 9.05–60.95 s · 300 Hz · WFDB 8,528 Open
ODC-By 1.0
Completed AliveCor Inc.; Emory University; MIT
USA
Clifford et al., CinC 2017
3 Apnea-ECG Database physionet.org ↗ 1-lead (ECG, unnamed) · 6.75–9.62 h · 100 Hz · WFDB · per-minute apnea labels 30 70 Open
ODC-By 1.0
Completed Philipps-University Marburg
Germany
Penzel et al., The Apnea-ECG Database, CinC 2000
4 ECG-ID Database physionet.org ↗ 1-lead (Lead I, limb clamps) · stored twice: raw + filtered · 20 s · 500 Hz · WFDB 90 310 Open
ODC-By 1.0
Completed Electrotechnical University "LETI"
Russia — St. Petersburg
Lugovaya, MSc thesis, 2005
5 Post-Ictal Heart Rate Oscillations in Partial Epilepsy physionet.org ↗ 1-lead ("ECG", unnamed) · 1.50–3.77 h · 200 Hz · 8-bit · WFDB · seizure times in a text file 5 7 Open
ODC Attribution
Completed Beth Israel Deaconess Medical Center / Harvard Medical School
USA — Boston, MA
Al-Aweel et al., Post-Ictal Heart Rate Oscillations in Partial Epilepsy, Neurology 53(7):1590-1592, 1999
6 tOLIet (Thigh-based ECG, toilet seat) physionet.org ↗ 1-lead (thigh, dry polymer electrodes) · 4 electrode textures per sitting · 14–197 s · 1,000 Hz · OpenSignals text 86 580 Open
CC BY 4.0
Completed Centro Hospitalar Universitário de Lisboa Central (CHULC)
Portugal — Lisbon
Silva et al., Scientific Data 2026
7 Brno University of Technology ECG Quality Database (BUT QDB) physionet.org ↗ 1-lead (Bittium Faros 180, chest) + 3-axis accel. · 24.0–38.7 h · 1,000 Hz · WFDB · sample-by-sample quality labels 15 18 Open
CC BY 4.0
Completed Brno University of Technology
Czech Republic
Nemcova et al., PhysioNet 2020 (method: Smital et al., IEEE TBME 2020)
8 VitalDB Arrhythmia Database physionet.org ↗ beat & rhythm annotations for VitalDB Lead II · no raw ECGs · 500 Hz 473 482 annotated cases Open
CC BY 4.0
Completed Seoul National University Hospital
South Korea
An Anesthesiologist-Validated Large-Scale Intraoperative Arrhythmia Dataset with Beat and Rhythm Labels
9 Preterm Infant Cardio-Respiratory Signals Database (PICSDB) physionet.org ↗ 1-lead (bedside monitor) · 20.3–70.3 h · 500 Hz (8 records) / 250 Hz (2) · WFDB · bradycardia onsets + R peaks 10 10 Open
ODC Attribution
Completed UMass Memorial Healthcare NICU
USA — Worcester, MA
Gee et al., Predicting Bradycardia in Preterm Infants Using Point Process Analysis of Heart Rate, IEEE TBME 64(9):2300-2308, 2017
10 ECG-Capable Smartwatches Dataset physionet.org ↗ 1-lead watches + 12-lead reference · 11–30 s · 250/300/500/512 Hz · WFDB 0 (synthetic — a patient simulator) 915 (736 in the clean partition) Restricted
PhysioNet Restricted Health Data License 1.5.0
Completed Complutense University of Madrid / Hospital Universitario Ramón y Cajal
Spain — Madrid
Recas et al., PhysioNet, 2025 (companion paper pending revision)

3-Lead ECG Datasets

2 datasets
# Dataset Format Patients Records Access Status Origin Paper Related
1 St. Vincent's / UCD Sleep Apnea Database (UCDDB) physionet.org ↗ 3-lead Holter (V5, CC5, V5R) · 7.52–8.68 h · 128 Hz · EDF · AHI + sleep stages + respiratory events 25 25 Open
ODC Attribution
Completed St. Vincent's University Hospital / University College Dublin
Ireland — Dublin
Dataset DOI
2 MIMIC-III Waveform Database Matched Subset physionet.org ↗ 1–5 ECG leads · typically 3-lead ICU (Lead II, V, AVR) · continuous · 125 Hz · WFDB 10,282 22,317 Open
ODbL
Not started Beth Israel Deaconess Medical Center
USA — Boston, MA
Johnson et al., Scientific Data, 2016

BSPM / ECGI Datasets

1 dataset
# Dataset Format Patients Records Access Status Origin Paper Related
1 EDGAR (Experimental Data & Geometric Analysis Repository) edgar.sci.utah.edu ↗ 24 ECGI experiments · body-surface + epicardial/endocardial/intramural · 54–2223 electrodes · 500–2048 Hz · MATLAB 20 subjects (13 human, 5 animal, 2 simulated) 2,943 Open
Free registration
Completed SCI Institute & CVRTI, University of Utah (curator); ten contributing groups
Multi-national (USA / Czechia / Germany / Spain / Canada / Netherlands / New Zealand / France / Slovakia / Switzerland)
Aras et al., J Electrocardiol, 2015

Dataset Statistics
Interactive visualisations derived from the full 64-dataset catalogue

Datasets by Lead Category

Distribution across the four lead-count groups

Access Type Breakdown

Open vs credentialed vs restricted across all datasets

Datasets by Country of Origin

Primary country attributed to each dataset (multi-national datasets counted once)

Top Datasets by Record Count (log scale)

Largest datasets ranked by number of records or segments

Access Type by Lead Category

How open access varies across lead-count categories

Sampling Frequency Distribution

How many datasets use each common sampling rate