A method for automatic identification of reliable heart rates calculated from ECG and PPG waveforms

Area of Science:

• Biomedical Engineering
• Physiological Monitoring
• Data Science in Healthcare

Background:

• Data-driven decision-support systems require reliable physiologic data for accurate medical triage, diagnostics, and prognostics.
• Current vital-signs monitors generate heart rate (HR) data from electrocardiogram (ECG) and photoplethysmogram (PPG) waveforms, but their reliability is often unquantified.
• Ensuring the accuracy of HR data is critical for meaningful interpretation and clinical decision-making.

Purpose of the Study:

• To develop and validate an automated method for assessing the reliability of reference heart rates (HRr) derived from ECG and PPG waveforms.
• To quantitatively express the reliability of HRr using a quality index (QI).
• To improve the trustworthiness of physiological data used in medical applications.

Main Methods:

• Assessed the quality of ECG and PPG waveforms using a Support Vector Machine classifier.
• Independently computed heart rates from both ECG and PPG waveforms using an adaptive peak identification technique (ADAPIT) designed to filter motion noise.
• Combined waveform quality and HR computation into a single Quality Index (QI), considering signal redundancy and independence of calculations.

Main Results:

• The automated method was evaluated on 158 trauma patient samples from helicopter transport.
• The algorithm's QI matched or was more conservative than human expert analysis in 92% of cases.
• Discrepancies in the remaining 8% were often due to ambiguous waveform quality, with a potential misclassification rate as low as 3% upon relabeling.

Conclusions:

• The developed method offers a robust and automated approach to evaluate the reliability of heart rate data.
• It provides a reliable quality index for HRr derived from vital-signs monitors.
• This facilitates the use of large quantities of heart rate data in data-driven medical systems.