How to train a device to monitor vital signs

We enrolled 30 participants in a calibration study to train our sensor for the accurate monitoring of heart rate, respiratory rate, and oxygen saturation.

Vital Signs to Detect Opioid Overdose

Opioid overdose is manifested by respiratory depression, which is a complex set of physiologic conditions defined by Medtronic as either (A) “respiratory rate ≤5 breaths/min (bpm), SpO2 ≤85%, or end-tidal carbon dioxide ≤15 or ≥60 mm Hg” for ≥3 minutes of (B) an “apnea episode lasting >30 seconds”.

To detect opioid overdose, the Resilient Monitor measures a user’s vital signs, specifically heart rate, respiratory rate, and oxygen saturation. We’ve built a custom patch-based sensor that is worn on the abdomen and uses the proven sensing technologies of photoplethysmography (PPG) and accelerometry for this purpose.

Study Protocol

Digital signal processing algorithms transform the raw output from PPG and accelerometry into vital sign measurements. These algorithms need to be calibrated on human data, since the input signal changes at different locations on the body. We enrolled 30 participants, each of whom wore:

  • four of our devices at different locations on the abdomen,
  • an industry-standard pulse oximeter for reference heart rate and SpO2 measurements,
  • a pneumotachometer for reference respiratory rate measurements.

Participants engaged in different activities throughout the study to modulate the vital signs of interest:

  1. Baseline Phase: Participants sat in a relaxed position.
  2. Respiratory Rate Phase: Participants breathed at different rates, following a series of onscreen instructions.
  3. Oxygen Saturation Phase: Participants followed a breathing pattern that included extended breath holds, so as to decrease their oxygen saturation.
  4. Heart Rate Phase: Participants walked on a treadmill to increase their heart rate.

Subject Enrollment

Since PPG can vary across different skin tones, and since our device is worn on the abdomen, we made sure to include a variety of different skin tones and BMIs in our pool of participants:

Over the past few months, we’ve been using the data collected from this study to calibrate our sensing algorithms for vital sign monitoring. 

For a look behind the scenes, see video here: Pittsburgh Region. Next is Now.

Contributor(s)
Bradford Holden
Charlie Proctor
Posted:
April 1, 2024
Stay up to date with our monthly newsletter
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
Connect with us