Switching Kalman filter based methods for apnea bradycardia detection from ECG signals

Area of Science:

• Biomedical Engineering
• Neonatal Monitoring
• Signal Processing

Background:

• Apnea bradycardia (AB) is a common complication in preterm infants, posing a challenge for timely clinical intervention.
• Early detection of AB is crucial for effective nursing care and improved infant outcomes.
• Cardiovascular signals, particularly electrocardiograms (ECG), offer potential for non-invasive AB monitoring.

Purpose of the Study:

• To develop and evaluate novel signal processing algorithms for the early detection of apnea bradycardia (AB) in preterm infants.
• To introduce two switching Kalman filter (SKF) based methods utilizing ECG signals for AB detection.
• To assess the efficacy of these methods in improving the speed and accuracy of AB identification.

Main Methods:

• Two SKF methods were developed: one integrating McSharry's ECG dynamical model with Kalman filters, and another using RR sequences with autoregressive (AR) models.
• Both SKF approaches employed a discrete state variable (switch) to select between normal and AB interval models during real-time analysis.
• Model probabilities determined by the switch assigned observation labels, facilitating continuous AB detection.

Main Results:

• The SKF method based on the ECG dynamical model demonstrated effective AB detection.
• Performance evaluation showed high sensitivity (94.74%) and specificity (94.17%) for the proposed method.
• The algorithm achieved rapid detection of AB events compared to annotated onsets.

Conclusions:

• The developed switching Kalman filter methods, particularly the ECG dynamical model approach, offer a promising tool for monitoring neonates at risk of AB.
• These algorithms can aid healthcare professionals in the early identification and management of apnea bradycardia.
• The proposed methods represent an effective algorithmic solution for continuous neonatal monitoring, potentially improving clinical outcomes.