Machine Learning Models for Tracking Blood Loss and Resuscitation in a Hemorrhagic Shock Swine Injury Model

Area of Science:

• Biomedical Engineering
• Physiology
• Machine Learning

Background:

• Hemorrhage causes life-threatening shock, often detected too late by traditional vital signs.
• Early detection of hemorrhagic shock is critical for timely intervention and improved patient outcomes.
• Existing vital signs may not adequately reflect the severity of blood loss due to compensatory mechanisms.

Purpose of the Study:

• To evaluate advanced blood loss metrics in a swine model of hemorrhage.
• To determine if previously developed metrics from canine models are applicable to swine.
• To assess the potential for these metrics in human medicine for early shock detection.

Main Methods:

• Utilized feature extraction and machine learning on arterial waveform data.
• Developed advanced metrics including Blood Loss Volume Metric, Percent Estimated Blood Loss, and Hemorrhage Area.
• Applied metrics developed in a canine hemorrhage model to data from a swine hemorrhage model.

Main Results:

• The advanced metrics accurately detected impending shock in a swine model.
• Performance of the metrics in swine validated the framework developed in canine models.
• These metrics identified shock earlier than traditional vital signs, such as blood pressure.

Conclusions:

• The developed metrics framework is applicable across different animal models (canine and swine).
• Advanced metrics show promise for early and accurate detection of blood loss and shock in humans.
• Successful application in swine suggests potential for improved casualty outcomes in civilian and military medicine.