Key result
Deep learning using continuous PPG predicts ICU cardiac arrest with ~0.79 AUROC.
Why the study?
Non-invasive patient monitoring to track and predict adverse acute health events is an emerging area of research, prompting investigation into IHCA prediction using only single-channel finger PPG signals.
Does a Feature Extractor-Aggregator Network using continuous PPG signals predict in-hospital cardiac arrest in ICU patients?
Does a Feature Extractor-Aggregator Network using continuous PPG signals predict in-hospital cardiac arrest in ICU patients?
Effect estimate: AUROC 0.79
A novel deep learning model using continuous single-channel PPG signals demonstrated strong predictive performance for in-hospital cardiac arrest in ICU patients up to 24 hours prior to the event.
Requires prospective validation before ICU deployment; leaves open PPG-based deep learning for cardiac arrest prediction.
Non-invasive patient monitoring for tracking and predicting adverse acute health events is an emerging area of research. We pursue in-hospital cardiac arrest (IHCA) prediction using only single-channel finger photoplethysmography (PPG) signals. Our proposed two-stage model Feature Extractor-Aggregator Network (FEAN) leverages powerful representations from pre-trained PPG foundation models (PPG-GPT of size up to 1 Billion) stacked with sequential classification models. We propose two FEAN variants ("1H", "FH") which use the latest one-hour and (max) 24-hour history to make decisions respectively. Our study is the first to present IHCA prediction results in ICU patients using only unimodal (continuous PPG signal) waveform deep representations. With our best model, we obtain an average of 0.79 AUROC over 24 h prediction window before CA event onset with our model peaking performance at0.82 one hour before CA. We also provide a comprehensive analysis of our model through architectural tuning and PaCMAP visualization of patient health trajectory in latent space.
No takes yet. Share an insight, caveat, or question.
Kataria et al. (2025) studied In-hospital cardiac arrest (IHCA). Feature Extractor-Aggregator Network (FEAN) using single-channel PPG signals was evaluated on In-hospital cardiac arrest (IHCA) prediction (AUROC 0.79). A deep learning model using single-channel continuous PPG signals predicted in-hospital cardiac arrest in ICU patients with an average AUROC of 0.79 over a 24-hour prediction window.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: