Software under Development
Qualification of ECG Waveform, Pulse Oximetry Waveform and Heart Rates [1]

This invention, a computer software system, automatically determines the reliability of heart rates (HR) derived from electro-cardiogram (ECG) and photoplethysmogram (PPG, pulse oximeter) waveforms generated by vital signs monitors. While enabling the selection of high-quality HRs from archived data for study, it also applies to data-driven decision-support systems for military and civilian medical triage, diagnostics, and prognostics. It can provide reliable HRs to automated care systems or to caregivers either at the bedside, or monitoring from a remote location.

The figure shows how the software determines the reliability of HRs and provides a quality index (QI) through a series of steps. The HR is independently calculated from the ECG and PPG waveforms (left side of figure). The HR algorithm for ECG waveforms uses a filter to eliminate the baseline of the signal and then uses a threshold computation to identify the peaks of the R waves. A set of moving square waves with varying periods is used to match the R waves to where the values are maximized. The HR rate is computed based on the number of rectangular waves optimized to the signal (blue box in figure). A similar approach is used for PPG-derived HR, with some modifications. A machine learning classifer, implemented by a Support Vector Machine, automates the categorization of the waveform quality by mimicking the performance of human experts who rely on visual inspection of the waveforms and the application of some implicit or explicit rules of thumb (pink box in figure). Finally, an algorithm compares the newly-computed and original monitor-derived (reference) heart rates, and the waveform quality, to formulate the final QI (orange box in figure).

Features and advantages:

  • Unique ECG and PPG waveform classification values based on time- and frequency-domain parameters

  • Novel ECG R-wave detection for computing HRs

  • Independently generates HR from ECG and PPG waveforms and compares to monitor values

  • Uses redundancy checks to generate QI

  • In 90% of trauma patient data, the algorithm matches or is more conservative than human-derived QI values