The ARGO dataset: Annotated and delineated intracardiac electrograms of post-ischemic ventricular tachycardia

The identification of abnormal ventricular potentials (AVPs) in intracardiac electrograms of post-ischemic ventricular tachycardia patients is challenging, and the development of automated tools is hampered by the lack of annotated benchmarking datasets. This work presents the ARGO dataset, the first open-access collection of manually annotated left-ventricle post-ischemic ventricular tachycardia electrograms. The dataset includes 1962 signals from nine patients, including unipolar and bipolar electrograms, 12-lead surface ECGs, and the local activation time and voltage maps, with annotations of the bipolar EGMs in terms of their nature (i.e., Physiological, AVP, or Unknown) and the delineation of the onset/end of the AVPs. The annotation is the result of three independent annotations by as many expert electrophysiologists, including a final consensus, for enhanced reliability and accuracy of the dataset. Leveraging good inter-rater and intra-rater reliability, the proposed validation analysis also provides quantitative estimates of the tolerance associated with AVP delineation error. Even considering the limitations of the proposed dataset affecting generalizability, ARGO provides a unique resource to support the development and validation of computational tools for AVP identification and characterization.