Annotated dataset of post-ischemic ventricular tachycardia electrograms (ARGO)

@article{PhysioNet-argo-1.0.0,
  author = {Orrù, Marco and Baldazzi, Giulia and Zirolia, Davide and Bertagnolli, Livio and Viola, Graziana and Solinas, Maria Giuliana and Pani, Danilo},
  title = {{Annotated dataset of post-ischemic ventricular tachycardia electrograms (ARGO)}},
  journal = {{PhysioNet}},
  year = {2026},
  month = jul,
  note = {Version 1.0.0},
  doi = {10.13026/8gh2-e660},
  url = {https://doi.org/10.13026/8gh2-e660}
}

Abstract

The analysis and identification of abnormal ventricular potentials (AVPs) in intracardiac electrograms (EGMs) are paramount in radiofrequency catheter ablation (RFCA) of post-ischemic ventricular tachycardia (VT). The ARGO dataset represents the first open-access dataset of post-ischemic VT EGMs featuring invasive and noninvasive electrophysiological recordings, as well as electroanatomical (EA) map data, including data annotation by three expert electrophysiologists and their final consensus. The dataset includes unipolar and bipolar EGMs, 12-lead surface ECGs, and EA maps, for a total of 1962 entries collected from nine post-ischemic VT patients using the CARTO® 3 mapping system. Every entry encompasses annotations of the bipolar EGMs in three different classes, i.e., Physiological, AVP, or Unknown, as well as the delineation of the onset and end of the pathological activation in the AVPs. The dataset also incorporates detailed information about the spatial localization of ablations performed during the RFCA procedures.

Background

Intracardiac EGMs analysis is crucial in understanding and treating cardiac arrhythmias. Nowadays, RFCA is the gold standard for treating this condition. It aims to target and ablate arrhythmogenic areas, to restore the physiological heart rhythm. In post-ischemic VT, arrhythmogenic areas are commonly identified by the presence of AVPs, characterized by late activations with low-voltage, high-frequency fractionated components in local bipolar EGMs [1]. Since AVPs identification is a laborious and time-consuming task, several studies explore computational methods leveraging signal processing techniques combined with artificial intelligence to assist clinicians, thereby expediting VT ablation procedures [2-8]. The development of automatic tools to support clinicians in identifying, characterizing, and delineating AVPs is still hampered by the lack of open-access datasets. Indeed, publicly available datasets related to intracardiac recordings are very rare. The only one is the Intracardiac Atrial Fibrillation Database [9], which primarily focuses on the annotation of QRS complexes for atrial fibrillation and flutter. Conversely, other works present datasets with indications of atrial activity onset/end in different types of arrhythmias [10] or propose datasets encompassing synchronized surface and intracardiac signals for the detection of various arrhythmias (including VT) as well as for the identification of AVPs [11, 12]. However, in such cases, free and public access is not guaranteed, which poses a significant obstacle to research in post-ischemic VT algorithms. The Ablation Reinforcement by computer-aided Guidance and Optimization (ARGO) study aims at filling this gap by proposing the first publicly available and free dataset. Indeed, the ARGO Dataset represents a reusable resource for benchmarking and developing computational tools and decision-support systems for AVP detection and delineation in post-ischemic VT ablation procedures.

Methods

Nine patients affected by post-ischemic VT who underwent EA mapping and RFCA between 2017 and 2018 at the San Francesco Hospital (Nuoro, Italy) were enrolled in the study. The electrophysiological data acquisition was performed by an expert electrophysiologist during routine electrophysiological studies and RFCA procedure using the CARTO® 3 mapping system (Biosense Webster, Inc., Diamond Bar, California). Intracardiac EGMs and EA information, provided in this dataset, were recorded in sinus rhythm using PentaRay 2-6-2 mm, ThermoCool SmartTouch, and ThermoCool SmartTouch SF (Biosense Webster, Inc., Diamond Bar, California). During mapping, all spatial data of the geometry of the LV, the EA maps (i.e., voltage maps and local activation time maps), and the 12-lead surface ECG were acquired simultaneously.

Each 12-lead ECG and EGM recording is 2.5 s long and sampled at 1 kHz. Only the last beat per signal must be considered reliable. All the recordings underwent band-pass filtering by the CARTO® 3 system. Specifically, bipolar EGMs were band-pass filtered in the range 16-500 Hz, unipolar EGMs in the range 2-240 Hz, and 12-lead ECG traces between 0.5 and 120 Hz.

Three expert electrophysiologists manually and independently annotated the signals using an ad-hoc annotation graphical user interface [13], blindly with respect to the patient. The annotators were asked to label each bipolar EGM segment as “Physiological”, “AVP”, or “Unknown” and to delineate the onset and the end of the abnormal potential in the case of AVP.

A final expert consensus was performed to ensure robustness for the whole data annotation process and enhance annotation reliability. Consequently, for each EGM, a final annotation was derived from such expert consensus and made available in the presented dataset, along with the three individual annotations.

The electrophysiological signals from the CARTO® 3 system and the related annotation data were formatted into WFDB-compatible files, validated with WFDB tools, and can be imported and analyzed using functions such as rdsamp and rdann from the WFDB Software Package.

Data Description

The dataset is structured hierarchically, with a directory assigned to each patient. In each patient folder, the electrophysiological data, the information on the EA map reconstruction of the LV, and the annotation in terms of class and delineation of the AVP are provided, for a total of 1962 entries.

The documented electrophysiological data, representing the heart’s electrical activity, includes bipolar EGM, a pair of unipolar EGMs, and the 12-lead ECG for each data point Pn. Adhering to the standard WFDB software package format, each n-th data point Pn is associated with two files: the data file and the header file. The data file, Pn.dat, contains the electrophysiological measurements column-wise, in a structured 2500x15 matrix format stored as dimensionless integer values with a gain factor of 0.003 mV/a.u. Supplementing this data file, the Pn.hea header contains metadata detailing technical aspects.

For each data point, distinct WFDB-compatible annotation files are reported for every annotator and expert consensus, namely, Pn.annotation_ann1, Pn.annotation_ann2, Pn.annotation_ann3 and Pn.annotation_consensus. These files report the class annotation (i.e., “P”, “A” or “U”), the onset and end delineation of the AVP in terms of sample index, respectively indicated as “(”and “)”. In cases where the n-th signal is not identified as AVP, the corresponding delineation information will be documented using a single “"” with a default out-of-range sample index set to 9999.

Lastly, for the reconstruction of the LV EA map as a triangulated mesh, all required details were reported in text files (.txt); specifically, for each patient, five files were included:

Each patient folder therefore includes six electrophysiological files for data point (Pn.dat, Pn.hea, Pn.annotation_ann1, Pn.annotation_ann2, Pn.annotation_ann3, and Pn.annotation_consensus), plus five unique text files describing the EA maps (XYZmesh.txt, ConnectivityList.txt, MESHcoloring.txt, POS_POINTS.txt, and AblationPoints.txt). Based on the number of annotated EGMs per patient (Pt1–Pt9: 157, 104, 90, 471, 46, 839, 76, 129, 50), the total number of files per patient folder is summarized below:

All identifying information that could link data to a given participant has been removed by anonymization before public release. The dataset contains no protected health information (PHI) or direct identifiers. The accompanying file Additional_subject_data.csv provides non-identifiable demographic and clinical information for the nine enrolled subjects, including patient code (Pt1–Pt9), sex, age, ejection fraction, and number of data points, solely for statistical reference. No identifying elements are included.

Usage Notes

Electrophysiological data from the CARTO® 3 system were converted into WFDB-compatible files, validated using WFDB tools, and are fully compatible with analysis functions (e.g., rdsamp, rdann) from the WFDB Software Package.

Additionally, the same information provided in WFDB format is made available in a unique MATLAB struct variable (i.e., ARGODataset_MATLAB.mat). The .mat file duplicates the WFDB signals for convenience, but WFDB-compatible files represent the primary, open-source data format.

To get started, users can load any record and its annotations using standard WFDB or MATLAB commands, as shown in the following MATLAB example:

% Read the signals, sample frequency, and the sampling intervals:
[signal, Fs, tm] = rdsamp('ARGODataset_Folder/Pt1/P71')

% Read the annotations:
[ann, anntype, ~, ~, ~, comments] = rdann('ARGODataset_Folder/Pt1/P71','annotation_consensus')

Beyond the dataset, a set of custom MATLAB functions (The MathWorks, MA, USA) has been developed to facilitate the reading and visualization of both electrophysiological and EA data. Specifically, the first function (VisualizeRecordingsARGO.p) enables the simultaneous display of 12-lead ECG recordings, the bipolar electrogram and corresponding unipolar leads, along with delineation markers provided by expert annotators and the final consensus (in case of AVP only). This visualization occurs in a separate window, allowing for detailed inspection. Additionally, a graphical user interface (GUI) (VisualizeMapsARGO_GUI.p) has been implemented and provided to visualize EA maps, including all routinely acquired maps during such electrophysiological studies, i.e., voltage and local activation time (LAT) maps, along with all targeted sites for ablation during the clinical procedure. These tools enable efficient browsing of both electrophysiological and EA data.

While the dataset provides synchronized and annotated recordings, its intended use is exclusively for the development and validation of computational algorithms. Therefore, no clinical conclusions should be drawn from its usage. Its scope is limited by the small patient cohort and the absence of detailed pharmacological data. Moreover, recordings were acquired using solely the CARTO® 3 V6 mapping system, ensuring homogeneity but limiting variability across different mapping platforms.

Release Notes

Version 1.0.0: Initial public release of the dataset.

Ethics

All the participants contributing with their electroanatomic recordings to this study provided their informed consent. The study was ethically approved by the Independent Ethical Committee of the Azienda Tutela Salute, Sardegna (Prot. n. 351/2021/CE, date of approval: 13/07/2021) and performed following the principles outlined in the 1975 Helsinki Declaration, as revised in 2000.

Acknowledgements

The research leading to these results has received funding from the European Union - NextGenerationEU through the Italian Ministry of University and Research under PNRR - M4C2-I1.3 Project PE_00000019 "HEAL ITALIA" to G. Baldazzi, CUP F53C22000750006. The views and opinions expressed are those of the authors only and do not necessarily reflect those of the European Union or the European Commission. Neither the European Union nor the European Commission can be held responsible for them.

Conflicts of Interest

The authors have no conflicts of interest to declare.

References

1. Baldazzi G, Orrù M, Solinas G, Matraxia M, Viola G, Pani D. Spectral characterisation of ventricular intracardiac potentials in human post-ischaemic bipolar electrograms. Sci Rep. 2022;12(1):4782. doi:10.1038/s41598-022-08743-7.
2. Pitzus A, Baldazzi G, Orrù M, Raffo L, Viola G, Djurić PM, Pani D. Arrhythmogenic sites identification in post-ischemic ventricular tachycardia electrophysiological studies by explainable deep learning. Biomed Signal Process Control. 2025;99:106844. doi:10.1016/j.bspc.2024.106844.
3. Baldazzi G, Orrù M, Viola G, Pani D. Computer-aided detection of arrhythmogenic sites in post-ischemic ventricular tachycardia. Sci Rep. 2023;13(1):6906. doi:10.1038/s41598-023-33866-w.
4. Pitzus A, Baldazzi G, Orrù M, Raffo L, Viola G, Djurić PM, Pani D. Arrhythmogenic sites mapping in post-ischemic ventricular tachycardia using a Siamese neural network. In: 2023 Computing in Cardiology (CinC). Vol. 50. IEEE; 2023. p. 1–4. doi:10.22489/CinC.2023.413.
5. Pitzus A, Baldazzi G, Orrù M, Rey AV, Viola G, Raffo L, Pani D. Exploring transfer learning for ventricular tachycardia electrophysiology studies. In: 2022 Computing in Cardiology (CinC). Vol. 49. IEEE; 2022. p. 1–4. doi:10.22489/CinC.2022.382.
6. Baldazzi G, Orrù M, Matraxia M, Viola G, Pani D. Efficacy of spectral signatures for the automatic classification of abnormal ventricular potentials in substrate-guided mapping procedures. In: 2022 Computing in Cardiology (CinC). Vol. 49. IEEE; 2022. p. 1–6. doi:10.22489/CinC.2022.351.
7. Baldazzi G, Orrù M, Matraxia M, Viola G, Pani D. Supervised classification of ventricular abnormal potentials in intracardiac electrograms. In: 2020 Computing in Cardiology (CinC). IEEE; 2020. p. 1–4. doi:10.22489/CinC.2020.397.
8. Baldazzi G, Orrù M, Matraxia M, Viola G, Pani D. Automatic recognition of ventricular abnormal potentials in intracardiac electrograms. In: 2019 Computing in Cardiology (CinC). IEEE; 2019. p. 1–4. doi:10.22489/CinC.2019.343.
9. Peterson J, Kynor D. Intracardiac Atrial Fibrillation Database. PhysioNet. 2003. Available from: https://physionet.org/content/iafdb/1.0.0/ . Accessed: 7 Mar 2026.
10. Redina R, et al. Arrhythmia database with annotated intracardiac atrial signals from pediatric patients undergoing catheter ablation. In: 2022 Computing in Cardiology (CinC). Vol. 49. IEEE; 2022. p. 1–4. doi:10.22489/CinC.2022.282.
11. Jenkins JM, Jenkins RE. Arrhythmia database for algorithm testing: surface leads plus intracardiac leads for validation. J Electrocardiol. 2003;36:157–161. doi:10.1016/j.jelectrocard.2003.09.041.
12. Yuldashev ZM, et al. Creation of a database of verified synchronous recordings of 12-channel high-resolution ECG signals and intracardiac electrograms. Biomed Eng (NY). 2021;55:273–277. doi:10.1007/s10527-021-10117-8.
13. Orrù M, Pitzalis E, Baldazzi G, Viola G, Pani D. A novel tool for the annotation of bipolar electrograms for cardiac electrophysiology. In: National Congress of Bioengineering Proceedings. Patron Editore Srl; 2023.