ECG dataset usage

Most-used ECG datasets.

PPG-DaLiA is the most-used ECG dataset in the ChatPPG research hub: 28 papers and 19 models published here link to it. The rest of the top five are WESAD (44), VitalDB (42), MESA Sleep (28) and PulseDB v2.0 (28). This ranking covers the 54 of 117 ranked datasets that record ECG, with 397 links from published papers and models.

54 ECG datasets · 266 paper links · 131 model links

Last updated

Ranking

54 datasets

#1
PPG-DaLiA

47 total · 28 papers · 19 models

Pretrain 1 · Train 0 · Eval 22

PPG · ECG · Accel · Skin temp · Resp

Licence: Scientific, non-commercial use with credit to the owners

#2
WESAD

44 total · 28 papers · 16 models

Pretrain 1 · Train 0 · Eval 21

PPG · ECG · Accel · Skin temp · EDA · Resp

Licence: Scientific, non-commercial use with credit to the owners

#3
VitalDB

42 total · 30 papers · 12 models

Pretrain 5 · Train 2 · Eval 7

PPG · ECG · Resp · ABP · CO₂

Licence: CC BY 4.0 (PhysioNet release)

#4
MESA Sleep

28 total · 21 papers · 7 models

Pretrain 5 · Train 0 · Eval 5

PSG · PPG · ECG · EEG · Resp · Accel

Licence: NSRR Data Access and Use Agreement

#4
PulseDB v2.0

28 total · 20 papers · 8 models

Pretrain 2 · Train 0 · Eval 4

PPG · ECG · ABP

Licence: Split by folder: PulseDB_MIMIC files under ODbL 1.0 (contents under DbCL 1.0); PulseDB_Vital files and the subset/Info files under CC BY-NC-SA 4.0

#6
PTB-XL

24 total · 15 papers · 9 models

Pretrain 0 · Train 1 · Eval 5

ECG

Licence: Not stated

#7
Sleep Heart Health Study (SHHS, NSRR)

18 total · 15 papers · 3 models

Pretrain 2 · Train 1 · Eval 2

PSG · ECG · EEG · Resp · SpO₂

Licence: NSRR Data Access and Use Agreement

#8
BIDMC PPG and Respiration (Pimentel 2016 / PhysioNet)

16 total · 10 papers · 6 models

Pretrain 1 · Train 0 · Eval 3

PPG · ECG · Resp · SpO₂

Licence: ODC-By 1.0

#9
MIMIC-IV-ECG

15 total · 10 papers · 5 models

Pretrain 2 · Train 0 · Eval 1

ECG

Licence: Not stated

#10
Brno University of Technology Smartphone PPG Database (BUT PPG)

14 total · 9 papers · 5 models

Pretrain 1 · Train 0 · Eval 4

PPG · ECG

Licence: CC BY 4.0

#11
MrOS Sleep Study (Osteoporotic Fractures in Men, NSRR)

10 total · 7 papers · 3 models

Pretrain 2 · Train 0 · Eval 1

PSG · ECG · EEG · Resp · SpO₂

Licence: Not stated

#12
UCI Cuff-Less Blood Pressure Estimation (UCI 340; Kachuee 2015)

8 total · 6 papers · 2 models

Pretrain 0 · Train 0 · Eval 6

PPG · ABP · ECG

Licence: Not stated

#13
Chapman–Shaoxing–Ningbo 12-lead ECG arrhythmia database (PhysioNet)

7 total · 4 papers · 3 models

Pretrain 1 · Train 0 · Eval 0

ECG

Licence: Creative Commons Attribution 4.0 International (CC BY 4.0)

#14
Long-term ECG and wrist PPG with annotated AF episodes (AFPPGECG)

6 total · 4 papers · 2 models

Pretrain 0 · Train 0 · Eval 3

PPG · ECG · Accel

Licence: Zenodo lists the licence as Other (Non-Commercial)

#14
WildPPG

6 total · 3 papers · 3 models

Pretrain 0 · Train 0 · Eval 3

PPG · ECG · Accel · Skin temp · BARO

Licence: Data: CC BY-NC-SA 4.0. Code: GPL-3.0

#16
MIMIC-BP

4 total · 4 papers · 0 models

Pretrain 0 · Train 0 · Eval 1

PPG · ABP · ECG · Resp

Licence: Open Data Commons Open Database License v1.0.

#16
Icentia11k single-lead continuous raw ECG (PhysioNet)

4 total · 3 papers · 1 model

Pretrain 0 · Train 0 · Eval 1

ECG

Licence: Not stated

#16
MIMIC-III-Ext-PPG

4 total · 3 papers · 1 model

Pretrain 0 · Train 0 · Eval 2

PPG · ECG · Resp · ABP

Licence: PhysioNet Credentialed Health Data License 1.5.0

#16
CODE-15%: large annotated 12-lead ECG dataset (Telehealth Network of Minas Gerais)

4 total · 2 papers · 2 models

Pretrain 1 · Train 0 · Eval 0

ECG

Licence: CC BY 4.0

#16
Gyro-Acc-PPG (BAMI-I, Wonkwang)

4 total · 2 papers · 2 models

Pretrain 0 · Train 0 · Eval 3

PPG · Accel · Gyro · ECG

Licence: Not stated

#16
IEEE Signal Processing Cup 2015 (TROIKA)

4 total · 2 papers · 2 models

Pretrain 1 · Train 0 · Eval 2

PPG · Accel · ECG

Licence: CC BY 4.0

#16
MIMIC PERform

4 total · 2 papers · 2 models

PPG · ECG · Resp · ABP

Licence: ODbL

#16
UCSF ICU bedside-monitor waveforms (private; GPT-PPG / SiamQuality / SQUWA pretraining)

4 total · 2 papers · 2 models

PPG · ECG · Resp

Licence: Not stated

#24
ECSMP

3 total · 3 papers · 0 models

PPG · ECG · Accel · Skin temp · EDA · EEG

Licence: CC BY 4.0

#24
CODE-test: annotated 12-lead ECG test set (Ribeiro 2020, Zenodo)

3 total · 2 papers · 1 model

Pretrain 0 · Train 0 · Eval 2

ECG

Licence: CC BY 4.0

#24
Harvard-Emory ECG Database (HEEDB v5.0)

3 total · 2 papers · 1 model

Pretrain 0 · Train 1 · Eval 1

ECG

Licence: Not stated

#24
St Petersburg INCART 12-lead Arrhythmia Database (PhysioNet)

3 total · 2 papers · 1 model

ECG

Licence: Not stated

#24
EarSet

3 total · 1 paper · 2 models

Pretrain 0 · Train 0 · Eval 1

PPG · ECG · Accel · Gyro · Resp

Licence: CC BY 4.0

#24
Pulse Transit Time PPG

3 total · 1 paper · 2 models

Pretrain 0 · Train 0 · Eval 1

PPG · ECG · Accel · Gyro

Licence: ODbL 1.0

#30
EchoNext-Mini

2 total · 2 papers · 0 models

ECG

Licence: PhysioNet Restricted Health Data License 1.5.0

#30
K-EmoCon

2 total · 2 papers · 0 models

PPG · Accel · EDA · Skin temp · ECG · EEG · HR

Licence: Restricted access (request on Zenodo); no license listed

#30
Pulsewatch

2 total · 2 papers · 0 models

Pretrain 0 · Train 1 · Eval 2

PPG · ECG · Accel

Licence: Not stated

#30
Bed-Based BCG (Carlson, K-State)

2 total · 1 paper · 1 model

BCG · ECG · PPG · Resp · Bp continuous

Licence: CC BY 4.0

#30
MIMIC-IV Waveform Database (PhysioNet preview)

2 total · 1 paper · 1 model

Pretrain 0 · Train 0 · Eval 1

PPG · ECG · Resp · ABP

Licence: Open Data Commons Open Database License v1.0 (ODbL 1.0)

#30
STAGES

2 total · 1 paper · 1 model

Pretrain 1 · Train 0 · Eval 0

PSG · EEG · ECG · Resp · SpO₂ · PPG · Accel

Licence: Not stated

#30
UQVitalSigns

2 total · 1 paper · 1 model

PPG · ECG · Resp · SpO₂ · CO₂ · BP

Licence: CC BY-NC 3.0

#30
UTSA-PPG

2 total · 1 paper · 1 model

PPG · ECG · Accel

Licence: Not stated

#30
WenXinWuYang

2 total · 1 paper · 1 model

PPG · ECG

Licence: Not stated

#39
CapnoBase IEEE TBME RR benchmark (Karlen 2013)

1 total · 1 paper · 0 models

PPG · ECG · Resp

Licence: Custom terms specific to this dataset

#39
GalaxyPPG

1 total · 1 paper · 0 models

PPG · ECG · Accel · Skin temp

Licence: Not stated

#39
Long-Term AF Database (LTAFDB, PhysioNet)

1 total · 1 paper · 0 models

ECG

Licence: Not stated

#39
Long-term ECG + wrist PPG with annotated AF episodes (Bacevičius 2024, Zenodo)

1 total · 1 paper · 0 models

PPG · ECG · Accel

Licence: Other (Non-Commercial)

#39
MIT-BIH Arrhythmia Database (PhysioNet)

1 total · 1 paper · 0 models

ECG

Licence: Open Data Commons Attribution License v1.0 (ODC-By)

#39
MIT-BIH Atrial Fibrillation Database (PhysioNet)

1 total · 1 paper · 0 models

ECG

Licence: Not stated

#39
MIT-BIH Supraventricular Arrhythmia Database (PhysioNet)

1 total · 1 paper · 0 models

ECG

Licence: Not stated

#39
MW-PPG-HR

1 total · 1 paper · 0 models

Pretrain 0 · Train 0 · Eval 1

PPG · ECG · Accel

Licence: Not stated

#39
OpenOximetry Repository (PhysioNet Restricted)

1 total · 1 paper · 0 models

PPG · ECG · SpO₂ · ABP · Etco2

Licence: Not stated

#39
OSASUD

1 total · 1 paper · 0 models

PPG · ECG · SpO₂ · HR · Resp · PSG

Licence: Not stated

#39
TILES-2018

1 total · 1 paper · 0 models

Pretrain 0 · Train 0 · Eval 1

PPG · IMU · ECG

Licence: Not stated

#39
VitalDB Arrhythmia

1 total · 1 paper · 0 models

Pretrain 0 · Train 1 · Eval 1

ECG

Licence: Not stated

#39
MCD-rPPG

1 total · 0 papers · 1 model

Video · PPG · ECG

Licence: CC BY 4.0

#39
MS-Drive

1 total · 0 papers · 1 model

Video · Nir video · ECG

Licence: MS-Drive academic-only licence (non-commercial; no modification or redistribution without written permission)

#39
ScientISST MOVE: Annotated Wearable Multimodal Biosignals recorded during Everyday Life Activities in Naturalistic Environments

1 total · 0 papers · 1 model

Pretrain 0 · Train 0 · Eval 1

PPG · ECG · Accel · EDA

Licence: Not stated

How the ranking works

Counts are links between published ChatPPG pages: a paper or model page that names the dataset it used, and that dataset's page. Unpublished papers and models are excluded, and nothing is inferred from titles or abstracts.

This page keeps only datasets whose ChatPPG record lists ECG among its recorded signals, and ranks them against each other. The rank column is the rank within ECG datasets; the overall ranking covers every dataset.

Datasets are ranked by total links (papers plus models); ties share a rank. Pretrain, train and eval columns count only links whose source states that role, so they can add up to less than the total; a dash means no link to that dataset states a role. The table refreshes with every research hub sync.

Licence, signals and tasks come from each dataset's ChatPPG record. Check the dataset page and its source before you use the data.

FAQ

What is the most used ECG dataset?

PPG-DaLiA, with 47 links from published ChatPPG pages (28 papers and 19 models). Next are WESAD (44), VitalDB (42) and MESA Sleep (28).

How many ECG datasets are ranked here?

54 datasets list ECG among their recorded signals and have at least one published paper or model linking to them. Together they have 266 paper links and 131 model links.

Which ECG datasets are used most for atrial fibrillation?

18 ECG datasets in this ranking list atrial fibrillation detection as a task. The most used are PTB-XL (24), Chapman–Shaoxing–Ningbo 12-lead ECG arrhythmia database (PhysioNet) (7) and Long-term ECG and wrist PPG with annotated AF episodes (AFPPGECG) (6).

Which ECG datasets do the most models use?

PPG-DaLiA (19 models), WESAD (16 models) and VitalDB (12 models). Model counts include foundation models and task models with a published ChatPPG page.

Which ECG datasets are used most for pretraining?

Where the source states a role, the most common pretraining datasets are VitalDB (5), MESA Sleep (5) and PulseDB v2.0 (2).

Which ECG datasets are used most for evaluation?

Where the source states a role, the most common evaluation datasets are PPG-DaLiA (22), WESAD (21) and VitalDB (7).

How are these counts calculated?

Each count is a link between two published ChatPPG pages: a paper or model page that lists the dataset it used, and the dataset's own page. 266 links come from papers and 131 from models. Unpublished papers and models are not counted, and no link is inferred. A dataset is included when its ChatPPG record lists ECG as a recorded signal.

Other rankings: all datasets, PPG datasets, IMU datasets, EDA datasets, rPPG datasets. Browse all datasets, or open the papers that use PPG-DaLiA. Benchmarks by dataset: leaderboards.

Cite this

Cite the page URL and the date you accessed it. Please also cite the original publications listed above.

APA

ChatPPG. (2026). Most-used ECG datasets (ChatPPG) [Data set]. ChatPPG Research Hub. https://chatppg.com/research/datasets/most-used/ecg

BibTeX
@misc{chatppg_most_used_ecg_datasets,
  author       = {{ChatPPG}},
  title        = {{Most-used ECG datasets (ChatPPG)}},
  year         = {2026},
  month        = nov,
  howpublished = {\url{https://chatppg.com/research/datasets/most-used/ecg}},
  note         = {ChatPPG Research Hub}
}
CITATION.cff
cff-version: 1.2.0
message: "If you use this resource, please cite it as below."
type: dataset
title: "Most-used ECG datasets (ChatPPG)"
authors:
  - name: "ChatPPG"
url: "https://chatppg.com/research/datasets/most-used/ecg"
date-released: "2026-11-02"