Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Reducing False Ventricular Tachycardia Alarms in ICU Settings: A Machine Learning Approach

Grace Funmilayo Farayola, Akinyemi Sadeeq Akintola, Oluwole Fagbohun, Chukwuka Michael Oforgu, Bisola Faith Kayode, Christian Chimezie, Temitope Kadri, Abiola Oludotun, Nelson Ogbeide, Mgbame Michael, Adeseye Ifaturoti, Toyese Oloyede

TL;DR

False ventricular tachycardia alarms in ICUs contribute to alarm fatigue and patient safety risks. The authors implement a machine learning pipeline using the VTaC dataset, combining time-domain and frequency-domain features with two neural architectures—a 1D CNN with multi-head attention and an FCNN for engineered features—to classify alarms as true or false. They achieve ROC-AUC scores above 0.96 across configurations, with SMOTE balancing providing the best gains for the FCNN; the system can trigger alerts only when confidence is high. The work demonstrates a clinically relevant reduction in false VT alarms and offers a scalable approach to improving patient safety in ICUs.

Abstract

False arrhythmia alarms in intensive care units (ICUs) are a significant challenge, contributing to alarm fatigue and potentially compromising patient safety. Ventricular tachycardia (VT) alarms are particularly difficult to detect accurately due to their complex nature. This paper presents a machine learning approach to reduce false VT alarms using the VTaC dataset, a benchmark dataset of annotated VT alarms from ICU monitors. We extract time-domain and frequency-domain features from waveform data, preprocess the data, and train deep learning models to classify true and false VT alarms. Our results demonstrate high performance, with ROC-AUC scores exceeding 0.96 across various training configurations. This work highlights the potential of machine learning to improve the accuracy of VT alarm detection in clinical settings.

Reducing False Ventricular Tachycardia Alarms in ICU Settings: A Machine Learning Approach

TL;DR

False ventricular tachycardia alarms in ICUs contribute to alarm fatigue and patient safety risks. The authors implement a machine learning pipeline using the VTaC dataset, combining time-domain and frequency-domain features with two neural architectures—a 1D CNN with multi-head attention and an FCNN for engineered features—to classify alarms as true or false. They achieve ROC-AUC scores above 0.96 across configurations, with SMOTE balancing providing the best gains for the FCNN; the system can trigger alerts only when confidence is high. The work demonstrates a clinically relevant reduction in false VT alarms and offers a scalable approach to improving patient safety in ICUs.

Abstract

False arrhythmia alarms in intensive care units (ICUs) are a significant challenge, contributing to alarm fatigue and potentially compromising patient safety. Ventricular tachycardia (VT) alarms are particularly difficult to detect accurately due to their complex nature. This paper presents a machine learning approach to reduce false VT alarms using the VTaC dataset, a benchmark dataset of annotated VT alarms from ICU monitors. We extract time-domain and frequency-domain features from waveform data, preprocess the data, and train deep learning models to classify true and false VT alarms. Our results demonstrate high performance, with ROC-AUC scores exceeding 0.96 across various training configurations. This work highlights the potential of machine learning to improve the accuracy of VT alarm detection in clinical settings.

Paper Structure

This paper contains 11 sections, 4 equations, 3 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Methodology
  3. Dataset
  4. Preprocessing
  5. Feature Extraction
  6. Model Architecture
  7. 1D Convolutional Neural Network with Multi-Head Attention
  8. Fully Connected Neural Network
  9. Handling Class Imbalance
  10. Results
  11. Conclusion

Figures (3)

  • Figure 1: Pipeline for false arrhythmia alarm detection in ICU settings. The system processes multiple physiological signals through feature engineering before classification using deep learning models.
  • Figure 2: Pipeline for VT Alarm Classification.
  • Figure 3: Architectural overview of the proposed models: (Left) A 1D Convolutional Neural Network (CNN) with Multi-Head Attention for processing time-series signals, comprising convolutional layers, attention mechanisms, and fully connected layers. (Right) A Fully Connected Neural Network (FCNN) for feature-based classification, featuring batch normalization and dropout regularization. Layer types and dimensions are annotated, with a legend clarifying component representations.