Physics informed ai improves false heart alarm detection in icu
Physics-Informed Deep Learning for False Ventricular Tachycardia Alarm Reduction in the ICU
Machine Learning
Summary
False alarms for dangerous heart rhythms in intensive care units create too many alerts, causing alarm fatigue for staff. The authors designed a deep learning method that not only analyzes heart signals but also checks if the signals match realistic blood pressure patterns, based on a simple physics model. This approach helps the system ignore false alarms caused by noisy data while catching real cases better. Tested on a standard challenge, it outperforms older methods and needs fewer examples to learn accurately.
What this means in practice
- •For icu technology teams: Improve ICU heart monitor software to reduce false ventricular tachycardia alarms by integrating physiologically constrained deep learning models.
- •For medical device developers: Design cardiac monitoring devices that incorporate physics-based deep learning to filter out noisy heart signal artifacts and enhance alarm reliability.
Authors
Athanasios Papastathopoulos-Katsaros, Alexandra Stavrianidi, Zhandong Liu
Abstract
False ventricular tachycardia (VT) alarms are a leading contributor to alarm fatigue in intensive care units. We propose a deep learning framework combining a 1D SE-ResNet with ICU-realistic data augmentations and a physics-informed auxiliary reconstruction task based on the three-element Windkessel hemodynamic model, implemented as a differentiable forward simulation. By requiring the network's latent representation to produce physiologically plausible arterial pressure waveforms, artifact-driven ECG patterns are penalized while true VT remains coherent across modalities. Evaluated on the VTaC benchmark under a strict real-time protocol (10-second pre-alarm window), our method achieves a 5-point Challenge Score improvement over prior state-of-the-art. Ablation studies confirm that the physics-informed objective is the primary performance driver, providing gains in accuracy, 2x label efficiency, and more localized and clinically meaningful ECG segments.