Battlefield Burn
Triage
Spectral MD × MTEC
DeepView’s AI burn assessment is moving from a hospital cart to a combat medic’s handheld — funded by MTEC, pushing the tool further forward in the care chain, to the point of injury, before a patient ever reaches an ER. The core question changes with it: not will this wound heal, but how severe is this burn, and how fast do we need to evacuate.
Context
One device.
Two very different moments of care.
DeepView already serves two customers: hospitals and burn centers, and BARDA/DoD-funded mass-casualty triage — through a cart-based device deployed anywhere from a battlefield ER to routine hospital rounds.
MTEC is funding the next step: pushing the same underlying AI even further forward. Not a hospital cart anymore — a handheld device a combat medic carries at the point of injury, before evacuation, before any ER.
This shift redefines the core question. The cart product asks will this wound heal? Does it need grafting? The handheld product asks how severe is this burn? How fast do we need to evacuate? That makes TBSA (Total Body Surface Area burned) — previously one data point among many — the hero metric, since it directly drives fluid resuscitation and evacuation priority.
2
care-chain positions served by one AI
—[confirm]
field interviews with medics conducted
—[confirm]
prototype iterations
—[confirm]
reduction in data-entry time
The Pivot
Same AI.
A completely different job.
Three things changed at once: who uses it, who it’s built for, and the question it’s being asked.
Research
Understanding the medic,
not just the wound.
When the user shifts from a physician in a clinic to a medic under fire, research has to establish what actually changes.
01 — SME & Medic Interviews
[Working title — confirm research method & sources][confirm]
[Add: specifics of interviews or field research conducted with medics / military SMEs][confirm]
What carried over from the hospital product’s research: hands are often gloved, one is frequently unavailable, light is inconsistent, and there is no time to read a manual. Every interaction has to survive being learned once, under stress, and used correctly the first time.
[Add: specific findings or quotes from field research][confirm]
Design
From cart workflow
to combat glove.
Two interaction decisions defined how this tool works when a medic has one hand free, ten seconds, and no time to read a manual.
01 — Data Entry Under Fire
“[Illustrative — confirm/replace with a real medic quote about data entry under pressure]”[confirm]
Dogtags were the first idea. Field-specific keyboards shipped.
Early on, we explored scanning a soldier’s dogtag to auto-populate patient identity and basic info — the obvious fix for slow manual typing in combat conditions. It was a reasonable idea, but it solved only one field and added a hardware dependency the device couldn’t always guarantee.
The solution that shipped instead: field-specific input keyboards. Rather than one generic text keyboard for every field, each data type — numeric vitals, body-region selection, TBSA percentage — gets a purpose-built input control suited to gloved, one-handed, high-stress use. Fewer keystrokes, fewer mistakes, no fumbling for the right key on a cramped standard layout.
[Add: measured or estimated reduction in data-entry time/errors][confirm]
02 — One-Take Photo Capture
“[Illustrative — confirm/replace with a real medic quote about retaking photos under fire]”[confirm]
No time for a retake.
The cart product could afford a slower, more deliberate capture process — a clinician in a stable room, retaking a photo if the first one was blurry or poorly framed. A medic under fire doesn’t get that luxury; every extra attempt is time not spent on the patient or on cover.
We added real-time on-screen guidance during the capture step: a “stay still” message that appears when motion would blur the shot, and real-time feedback on capture distance (too close / too far) so the medic can correct before pressing the shutter instead of after. The goal was a usable photo in one take, not several.
[Add: measured or estimated reduction in failed-capture retries][confirm]
Reflection
What changes when the user
is holding a rifle, not a stylus.
Design for stress, not just expertise
The difference between a physician and a medic isn’t just depth of training — it’s available attention in the moment. Cognitive overhead that’s tolerable in a clinic — one more tap, one more line to read — becomes an untreated patient on a battlefield. Every interaction has to assume the user gets one shot at getting it right.
One AI model, two very different moments of care
The cart product and the handheld product share the same underlying prediction model — what actually changed is the interaction layer: who’s using it, what they need to know in that moment, and how they need to be told. Pushing a product into a new context is often less about retraining the model and more about redesigning the experience wrapped around it.
[Add a third reflection — e.g. a finding that surprised you, or a constraint you didn’t anticipate][confirm]
[Add the supporting detail for this reflection][confirm]