Related Experiment Videos
A breathing circuit alarm system based on neural networks
1Department of Anesthesiology, University of Utah, Salt Lake City 84132.
Journal of Clinical Monitoring
|March 1, 1994
Summary
This study developed an intelligent anesthesia alarm system using neural networks to detect breathing circuit faults. The system demonstrated high accuracy in detecting faults, potentially improving patient safety by reducing false alarms.
Area of Science:
- Biomedical Engineering
- Artificial Intelligence in Medicine
- Anesthesiology
Background:
- Anesthesia breathing circuits are critical for patient ventilation.
- Current alarm systems often generate false alarms, leading to alarm fatigue.
- Intelligent alarm systems are needed to improve specificity and reduce false alarms.
Purpose of the Study:
- To implement an intelligent respiratory alarm system utilizing a neural network.
- To enhance alarm specificity and decrease false alarm rates compared to existing systems.
Main Methods:
- A neural network was trained to identify 13 distinct faults in anesthesia breathing circuits.
- Thirty breath-to-breath features were extracted from airway CO2, flow, and pressure signals.
- Training data involved introducing 616 faults in 5 dogs, with the network trained using the backward error propagation algorithm.
Main Results:
- The trained neural network achieved 95.0% accuracy in fault detection during controlled ventilation in animals.
- Accuracy was 86.9% during spontaneous breathing in animals.
- In clinical testing, the system identified 54 of 57 real-world faults over 43.6 hours, with 74 false alarms.
Conclusions:
- Neural networks show promise for developing intelligent anesthesia alarm systems.
- The implemented system demonstrated potential for improved fault detection and reduced false alarms.