Post-Resuscitation ECGs for Neurological Outcome Prediction After Cardiac Arrest

976 post-resuscitation 12-lead ECGs from the Lombardy Cardiac Arrest Registry, collected between January 2015 and December 2023. The dataset was used to develop a deep neural network model for predicting neurological outcome after out-of-hospital cardiac arrest. The model achieved an AUROC of 0.86 and an accuracy of 80.8%.

Use Cases

• Training machine learning models for early neurological outcome prognostication based on post-ROSC ECG features.
• Validating computer vision techniques for feature extraction from 12-lead ECG waveforms.
• Benchmarking model interpretability methods like SHAP values in a clinical prediction context.
• Investigating the association between ECG characteristics, patient demographics, and neurological recovery.

Strengths

• Dataset includes 976 ECGs with associated neurological outcome labels (641 poor, 335 favorable).
• Model performance is documented with multiple metrics, including an AUROC of 0.86 and an MCC of 0.61.
• Data was prospectively collected from a regional registry over an 8-year period (2015-2023).

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown for the specific data file, which may limit suitability assessment.
• Data may reflect geographic bias inherent to the Lombardy Cardiac Arrest Registry.

Provenance

Source

Lombardy Cardiac Arrest Registry

Collection Method

Retrospective analysis of prospectively collected post-ROSC ECGs.

Time Range

January 2015 to December 2023

Freshness

Last updated 2026-05-11 05:25:50; freshness should be verified.

Geography

Lombardy region