Capstone
helm
A surgical instrument-readiness system that uses infrared-responsive ink and tray scanners to track when and where each instrument was last handled, so OR teams catch missing equipment before it becomes a delay.
ROLE
Project Lead
TIMELINE
20 Weeks
TEAM
6 UX Designers
UX Design
UX Research
System Design
AI Integration
Autonomous Systems
THE PROBLEM
Surgical delays rarely come from one obvious cause.
Early research pointed to three equally plausible directions: equipment readiness, team turnover alignment, and cognitive load. Equipment readiness won out, confirmed by the people doing the work.
RESEARCH
Two experts who’d never spoken pointed to the same fix.
An OR researcher independently ranked equipment tracking as the highest-impact, most solvable area, before we'd told her which direction we were leaning. That alignment, between two people who'd never met, confirmed the direction.
“Humans being humans is probably one of the biggest reasons for the smaller delays. The bigger ones? Probably equipment problems.”
— CHIEF OF ANESTHESIA · DIRECT INTERVIEW
45.9%
of OR delays happened specifically because an instrument wasn’t available when needed. (Wubben et al., 2010)
56%
higher risk of in-hospital death when surgery is delayed. (McIsaac et al., 2017)
$21–133
cost of every minute of OR time, depending on case complexity. (Shippert, 2005)
DIRECTION & PROCESS
Build detection into the tray that’s already at the center of every case.
We reframed the problem: put instrument detection into the Mayo tray itself, avoiding the need for another device to track. From there the concept survived three real pivots, each forced by what the work actually demanded.
Automatic Sensing
Practitioners didn't want another task, so the system recognizes what's in the room automatically.
RFID → NIR ink
RFID meant competing head-on with entrenched hospital tagging systems. Dr. Maged Henary's near-infrared ink is patented and clinically validated for tracking retained surgical items in real ORs, though never applied to instrument readiness before. The ink can also be custom-written, giving hospitals flexibility RFID tags don't.
A layered pipeline
A single force-sensitive resistor drifted on its own, so we reframed it as a first-pass signal, narrowed by a YOLOv8n detection layer and confirmed with a final contextual check.
A fraction of RFID's per-tag cost
$2+
SURGICAL RFID · COST PER TAG
Sterilizable, autoclave-rated tags used in systems like STANLEY Healthcare's STM.
INDUSTRY PRICING DATA, 2026
5–10¢
NIR INK · COST PER TAG
Custom-writable, applied per instrument, and clinically validated for real surgical environments.
DR. MAGED HENARY · DIRECT INTERVIEW
THE SYSTEM
Detection in the tray, answers in the room.
ON-TRAY DETECTION
Recognized as it lands.
Near-infrared tagging lets the tray see which instruments are present without anyone scanning or logging a thing. A force-sensitive first pass narrows candidates, confirmed by a YOLOv8n layer and a final contextual check.
VOICE-FIRST ALERTS
Heard before it’s a bottleneck.
The team hears what's missing the moment it matters, without another screen to watch.
THE AI BOUNDARY
Recommend, never decide.
Every clinician drew the same line on AI, independently. Knowing where the system shouldn't act mattered as much as knowing where it should, so Helm surfaces what it detects and leaves the decision to the OR team.
MY ROLE
What I owned across the project.
Helm was a full end-to-end effort, spanning the framing of an ambiguous problem through validating the concept with real practitioners.
Research & synthesis
Led primary research across a 4-hour OR shadowing session and interviews spanning a Chief of Anesthesia, OR and scrub nurses, a surgical researcher, and a health-tech UX researcher, then synthesized the findings into three testable problem directions.
Concept & framing
Reframed instrument readiness around the tray already at the center of every case, avoiding yet another device for the team to monitor.
System design
Designed the detection pipeline and the voice-first alert model so guidance stays ambient and the final call stays human.
Testing & validation
Planned and ran 8 usability sessions with 10 OR staff, checked color contrast in Figma throughout, and validated the concept live at Demo Day.
TESTING & DEMO DAY
Tested with 10 OR staff, then run live at Demo Day.
Across 8 sessions, real-time tracking felt intuitive and alerts improved response time. We presented the working prototype using OpenCV and Arduino tools, a 3D-printed tray running the full sensor pipeline, at SCAD's UX capstone Demo Day.
WHAT TESTING CAUGHT
The dashboard needed a rebuild mid-testing.
The early version was too text-heavy, and detection feedback wasn't clear to OR staff during sessions. We simplified the layout and rebuilt the alert hierarchy before Demo Day. The clearest lesson: legibility drove trust in the system more than raw detection accuracy did.
OUTCOMES
Validated with the people who’d actually use it.
Before committing to build, the direction was pressure-tested with real OR staff and confirmed live, the strongest signal that Helm addresses a problem worth solving.
10
OR staff in usability testing
8
sessions run before Demo Day
1
direction validated from three candidates
REFLECTION
Practitioner validation proved the project's strongest asset.
Confirming the direction with real practitioners before committing engineering effort mattered most, reinforced by realizing people had already been imagining a version of this on their own. What I'd test next is the physical prototype in real hands, since designing for adoption matters as much as designing for function.