Real-Time Comparison of Machine Learning-Enabled Devices for Measuring Compensatory Reserve Status

Area of Science:

• Physiology
• Biomedical Engineering
• Medical Devices

Background:

• Central hypovolemia is a critical condition requiring accurate monitoring.
• Existing methods for assessing compensatory reserve may have limitations in real-time application.
• Noninvasive photoplethysmography offers a potential avenue for continuous physiological monitoring.

Purpose of the Study:

• To compare the real-time performance of a novel compensatory reserve measurement (CRM) device against an established FDA-cleared compensatory reserve index (CRI) device.
• To validate the CRM device using data from a simulated central hypovolemia study.
• To assess the potential of CRM for early detection of compensatory changes.

Main Methods:

• A clinical research study involving simulated central hypovolemia induced by lower body negative pressure (LBNP).
• Data acquisition using noninvasive photoplethysmography sensors from 20 participants.
• Comparison of CRM and CRI device outputs using median error, median absolute error, and pooled Pearson correlations.
• Bland-Altman analysis to assess bias between the two devices.

Main Results:

• Pooled correlation analysis demonstrated strong linear agreement between CRM and CRI (R² = 0.859).
• Bland-Altman analysis indicated a near-zero mean bias (0.13%) between the devices.
• CRM provided early measurement at a mean of 20.52 minutes, comparable to CRI's 17.92 minutes.

Conclusions:

• The CRM device exhibits statistically similar performance to the CRI device in tracking compensatory status during simulated central hypovolemia.
• Both devices effectively monitor compensatory changes associated with progressive hypovolemia.
• Real-time tracking capabilities of CRM hold promise for continuous monitoring and triage in prehospital trauma settings.