New Artificial Intelligence Tool Predicts Hemorrhagic Shock for Injured Warfighters

Research Team Uses Machine Learning to Detect Severe Bleeding Before Standard Vital Signs Drop

June 16, 2026 by Wendy Li

Hemorrhagic shock is the leading cause of preventable death in both civilian trauma and battlefield settings. Standard vital signs, including heart rate and blood pressure, can mask significant blood loss until it is too late to act. Dr. Victor Convertino, an adjunct faculty member in the Department of Medicine at the Uniformed Services University of the Health Sciences (USU), and the U.S. Army Institute of Surgical Research have developed an artificial intelligence (AI) tool that detects the crisis before a monitor ever shows it.

Convertino, also a Senior Scientist for Combat Casualty Care at the U.S. Army Institute of Surgical Research at Joint Base San Antonio-Fort Sam Houston, has spent more than a decade studying how the body responds to hemorrhage and how to predict shock before it sets in. He and his colleagues recently published on the development of a novel medical capability called "compensatory reserve measurement" (CRM), a continuous index of the body's remaining capacity to offset blood loss before vital signs collapse.

Vital signs, Convertino explained, are lagging indicators. They reflect bleeding's secondary effects, not the underlying hemodynamic shift that precedes clinical deterioration. The CRM reads the body's actual compensatory state directly. "Vital signs change very little in the early phases of hemorrhage because they are secondary effects of bleeding, rather than primary mechanisms of compensation that should be measured," he said.

The tool is built on a deep Convolutional Neural Network (CNN), the same class of machine learning architecture behind facial recognition, trained on physiological data from healthy volunteers. Those volunteers underwent controlled, progressive reductions in central blood volume using Lower Body Negative Pressure (LBNP), a noninvasive laboratory technique that safely mimics hemorrhagic stress by pulling blood away from the upper body. A finger cuff recorded each participant's blood pressure waveform continuously. From those waveforms, the CNN analyzed 20-second intervals and extracted 57 distinct physiological features that reflect how the body adapts to reduced blood volume.

Once trained, the algorithm can evaluate a patient it has never seen before, detect subtle warning signs in the pulse, and calculate a CRM score in real time - giving a medic an important head start before vital signs begin to shift. The score runs from 100%, indicating full compensatory capacity, down to 0%, indicating decompensated shock.

At the point of injury, where Role 1 medics, the first echelon of care on the battlefield, operate under pressure, that information must be immediate and readable. The CRM's display translates the score into a color-coded bar that functions like a fuel gauge: green (100% to 75%) signals adequate reserve, amber (75% to 40%) signals moderate reserve, and red (40% to 0%) signals impending shock. The algorithm's precision is what makes the approach clinically viable. "Accuracy analysis shows a correlation of 0.95 or greater for estimating compensatory reserve at any point in time using this approach while all standard vital signs remain virtually unchanged," Convertino said.

The Army Medical Department's Future Capabilities Directorate has already mandated integration of the CRM into a wearable medical trauma sensor under development. A medic treating a blast casualty could, in the near future, read a patient's compensatory reserve from a wrist-worn device and initiate resuscitation before a single vital sign crosses a clinical threshold - long before the window for intervention closes.