A Feature Selection Method Based on Shapley Value to False Alarm Reduction in ICUs A Genetic-Algorithm Approach

Area of Science:

• Biomedical Engineering
• Medical Informatics
• Artificial Intelligence in Healthcare

Background:

• High false alarm rates in intensive care units (ICUs) pose significant clinical challenges.
• False alarms overwhelm staff, reduce care quality, and negatively impact patient well-being (delirium, stress, sleep deprivation).
• Individual processing of noisy or artifact-corrupted signals from monitoring devices contributes to false alarms.

Purpose of the Study:

• To develop a computationally efficient and accurate method for reducing false alarms in ICUs.
• To identify the most informative biomarkers across multiple monitoring device signals.
• To mitigate the negative consequences associated with high false alarm rates.

Main Methods:

• Proposed a game-theoretic feature selection method.
• Utilized a genetic algorithm for biomarker identification.
• Focused on low-computational complexity for practical clinical application.

Main Results:

• Successfully identified informative biomarkers across diverse signals.
• Demonstrated a considerable reduction in the false alarm rate.
• The proposed method is accurate and computationally efficient.

Conclusions:

• The developed game-theoretic feature selection method effectively reduces ICU false alarms.
• Integrating biomarker identification across multiple signals improves alarm accuracy.
• This approach offers a promising solution to a critical healthcare challenge.