Role of a Deep-Learning Based Convolutional Neural Network Model for Real-Time Ventricular Tachycardia Alarm...

Area of Science:

• Critical care medicine
• Biomedical engineering
• Artificial intelligence in healthcare

Background:

• Ventricular tachycardia (VT) is a dangerous heart rhythm requiring prompt detection in intensive care units (ICUs).
• Current monitoring systems generate a high rate of false positive alarms, leading to alarm fatigue among healthcare professionals.
• Alarm fatigue can result in desensitization, delayed responses, and increased cognitive load for clinicians.

Purpose of the Study:

• To develop and evaluate a deep learning model for accurate classification of VT alarms.
• To reduce the incidence of false positive alarms in continuous cardiac monitoring.
• To alleviate alarm fatigue in the ICU setting.

Main Methods:

• A deep learning-based, one-dimensional convolutional neural network (1D-CNN) was developed.
• The model utilized multiple raw waveform inputs: two electrocardiogram (ECG) leads, photoplethysmogram (PPG), and arterial blood pressure (ABP) signals.
• Training was performed on the VTaC Arrhythmia Benchmark Dataset using 10-second pre-processed waveform segments preceding VT alarms.

Main Results:

• The 1D-CNN model achieved an Area Under the Receiver Operating Characteristic Curve (AUC) of 0.901 on the test set.
• Performance metrics included an overall accuracy of 83.22%, F1-score of 73.3%, sensitivity of 77.53%, and specificity of 85.63%.
• The model successfully detected over 75% of VT alarms while substantially decreasing false positive rates.

Conclusions:

• Deep learning, specifically a 1D-CNN model, can effectively distinguish true from false VT alarms.
• Utilizing short segments of raw waveform data enables robust VT alarm classification.
• This approach holds promise for improving the reliability of cardiac monitoring systems in ICUs.