2026 ECGI Data Challenge 

Summary:

We are excited to announce the 2026 consortium for ECGI data challenge! We seek to provide an opportunity to showcase your ECGI workflow and compete against others in the field on a set of centralized datasets. Our hope is to generate novel innovations, collaborations, and energy among the ECGI community. Challenge winners will be announced at the 2026 ECGI summit (November 2026) in Valencia, Spain. We have an excellent cohort of datasets from contributors in Utah (USA), Bordeaux (France), Bratislava (Slovakia). Auckland (New Zealand), and Halifax (Canada).

Objective:

The goal of this challenge is to bring together a variety of datasets and provide a standardized testing bed for comparison of ECGI techniques. We aim to provide incentive for further improvement and take a survey of the state of the art in ECGI applications. To that end, we intend to summarize the results of this study into a publication which can be referenced in future ECGI improvement efforts. To aid in this, we also encourage all participants to 1) make their code open-source whenever possible to aid in contemporary  and future validation efforts and 2) contribute if invited to the summary publication following the challenge.

We will split the datasets described below into a training, and testing set. The entire training set (inputs and ground truth) will be provided to challenge participants. The testing set will be kept hidden. Later in the challenge we will release the inputs to the testing set and open a submission site for answers on the test set. The winners of the challenge will be the ones who score highest on the testing dataset. Throughout the challenge we will post a leaderboard.

2026 Challenge Description:

Premature ventricular contractions are a phenomena where activation initiates spontaneously at a point in the ventricles of the heart rather than through the standard conductive pathway initiating at the sinus node. Similarly, premature ventricular beats can be initiated through electrical stimulation of the myocardium. Tracking the location of either natural PVCs or paced ventricular beats has been a common validation task for ECGI techniques. The focus of the score metric for the 2026 ECGI Data Challenge will be the localization the side of initiation for either natural PVCs or paced ventricular beats across the datasets. When present, invasively recorded electrogram data will be compared to inverse reconstructions, but the ranking of submissions will be based on their localization of the pacing site or PVC initiation location. The datasets organized for this year's challenge contain a mix of natural PVCs, paced beats, clinical, and pre-clinical data. Thanks again to all of the contributing groups and organizations.

Links:

• Challenge Repo (all the code for the challenge, with example scrips and the scoring scripts): TO BE PUBLISHED (5/15/2026)

• Challenge Registration Link :  TO BE PUBLISHED 

• Challenge datasets (you will need to register an EDGAR account)  : https://edgar.sci.utah.edu/2026/05/02/2026-ecgi-data-challenge/

• Challenge Submission Site : TO BE PUBLISHED (7/1/2026) 

Datasets:

• • Utah torso tank dataset:
    Simultaneous epicardial (240 electrode sock) and torso tank (192 electrodes) recordings during a langgendorf canine heart preparation. Data consists of signals recorded during intramural ventricular pacing, sinus rhythm, and atrial pacing. Data is recorded from two separate preparations.

  • Auckland In Vivo Porcine dataset:
    Simultaneous epicardial (239 electrode sock) and torso surface (170 electrodes) recordings during epicardial paced beats in a porcine in-vivo model. Data was collected from 4 subjects with 5-15 pacing sites per subject .

  • Bordeaux Torso Tank dataset:
    Simultaneous epicardial (128 electrode sock) and torso tank (256 electrodes) recordings during a langgendorf porcine heart preparation. Data consists of signals recorded during intramural ventricular pacing, and sinus rhythm. Data is recorded from ?? separate preparations.

  • Bratislava Human PVC Ablation Dataset
    Body surface and 12 lead ECG recordings from 13 human subjects who underwent ablation for premature ventricular contractions along with the ablations sites used to terminate PVCs.

  • Halifax Clinical Epicardial Pacing Dataset
    Body surface mapping data during epicardial mapping and pacing of a patient during a clinical VT mapping procedure.
    
    

Key Dates:

• Release of the challenge and datasets: 5/5/2026

• Submit challenge abstract : 9/30/2026  

• Initial date to submit on the test set: 10/1/2026

• Final date to submit on the test set: 11/10/2026

• Award announcement: 11/25/2026 (at the ECGI summit)

Rules:

These rules and guidelines as well as other challenge information will be updated as needed. This is out first year running the challenge, have patience and let us know if there are problems. To be ranked in the challenge it is expected that at least one member of each team will attend the conference in person. If unforeseen circumstances prevent someone from your team being able to attend, please reach out to us and we will see if there is an accommodation we can arrange. We do not guarantee that non-in-person participants can be ranked.

How to participate:

1. Register (either as an individual or a team) for the challenge using the link above

2. Get the data from the link above (you will need to register for an account on EDGAR)

3. Develop using the training data

4. Submit your abstract by the submission deadline (??)

5. Submit results on the Testing data once it has been released

6. Register and attend the ECGI summit

7. Attend, present, and discuss

8. Consider contributing to the followup proceedings paper after the conference