Estimation of Heart Rate Directly from ECG Spectrogram in Neonate Intensive Care Units

Area of Science:

• Biomedical Engineering
• Cardiology
• Neonatal Medicine

Background:

• Traditional neonatal heart rate (HR) monitoring relies on RR-interval analysis from ECG, which can be prone to errors from peak detection.
• Accurate HR monitoring is critical for neonates, especially in Neonatal Intensive Care Units (NICUs).

Purpose of the Study:

• To present a novel, simple method for estimating neonatal heart frequency (HF) and heart rate (HR) directly from electrocardiogram (ECG) signals.
• To avoid the inaccuracies associated with traditional peak detection methods used in RR-interval analysis.

Main Methods:

• The method utilizes the highest Power Spectral Densities (PSD) within specific frequency bins of the ECG signal, analyzed over time using spectrograms.
• This approach leverages time-varying spectral characteristics of the ECG to estimate heart rate directly.

Main Results:

• Validation with 6-day monitoring data from 52 NICU patients showed 92.4% of samples within 5 bpm of a commercial monitor.
• Median Mean Absolute Error (MAE) was 2.28 bpm, and median Root Mean Square Error (RMSE) was 5.82 bpm.
• Failure cases were linked to high PSD in lower frequencies, often coinciding with critical alarms from other physiological systems like desaturation.

Conclusions:

• The proposed ECG-based PSD method offers a promising, accurate alternative for neonatal heart rate estimation.
• The method's potential for customization to other patient populations warrants further investigation.
• Understanding error sources, such as interference from other physiological events, is crucial for refining the technique.