Spatial dermatology — from $2M rigs to a 1-minute iPhone scan

Problem

Approach

I dedicated a room in the office, sourced four iPhones with cross-polarized LED panels, and physically reproduced the $2M competitor’s capture workflow at a tenth of the price point. In parallel, when Apple announced Vision Pro and the AAD 2024 conference timing didn’t align with the normal release cadence, we ran a separate three-week sprint for the spatial-review chapter.

1. 01
   Capture POC on consumer hardware.

   iPhone + ARKit + open-source AR libraries to scan head and body, generating a high-quality 3D model in seconds. Reused existing AI models — lesion detection, lesion matching, blurriness checks — rather than rebuilding the stack.

2. 02
   The custom photo-booth rig.

   Four iPhones, tuned lamps, cross-polarized lighting. Optimal imaging conditions without medical equipment costs. Scan time target: from 20–30 minutes per patient down to ~1 minute.

3. 03
   DermVR — the spatial-review precursor.

   Before Vision Pro existed, we ran the spatial-review hypothesis on VR hardware: load a 3D body mesh, place lesion overlays at depth, let the dermatologist rotate the patient instead of the photo. Proved the model worked clinically — the immersive-review thesis was validated before we had the right device.

4. 04
   Vision Pro at AAD — three-week sprint.

   Apple Vision Pro was delayed by a month; CTO flew to Seattle to pick up the device the day it landed. Release freezes, two-hour escalation rule, single workflow scoped to read like the future. Live booth demo with team members playing doctor and patient — built on the same DermEngine SDK so AI models (DermDx, SmartSnap, lesion detection) lit up immediately.

5. 05
   Productize the capture rig.

   POC → shipping device. Magnetic iPhone (15 Pro+) mount, 48MP capture, 360° automated rotating arm, cross-polarized LED panels for shadow-free uniform illumination, support handles for stability across body types, 200 kg weight capacity, 80×80×225 cm footprint. Integrates directly with DermEngine’s 3D TBP module for AI-driven analysis. SOC 2, HIPAA, GDPR, Australia Privacy Act, NZ Medsafe, Switzerland FADP — multi-region from day one. Adjacent verticals validated beyond derm: aesthetics tracking, fitness/body transformation, fashion virtual fitting, uniform sizing. Tagline: “From hours to minutes.” dermengine.com/optimascan.

The capture rig in motion. What replaces a $2M scanner is not one clever piece of software — it’s the integration: ARKit doing depth, open-source AR doing meshing, DermEngine’s vision models doing lesion detection and matching, all wired through a four-iPhone fixture with cross-polarized lights so the output is clinically usable, not just visually impressive.

Two-million-dollar hardware is a market gap, not a moat. The real product is the workflow that makes the rig irrelevant — and the platform underneath it that makes both capture and review trivial to extend.

What shipped

• OptimaScan POC validated.

  Two government agencies, including Innovative Solutions Canada (IndigenousTech.ai), signed on as pilot partners after the consumer-hardware capture demo.

• DermVR prototype.

  Working VR build on top of DermEngine’s 3D body model — proof the immersive-review hypothesis was real before Vision Pro shipped.

• DermEngine on Vision Pro.

  Native visionOS app on the App Store, three weeks from kick-off to AAD booth demo. Lesion detection, SmartSnap, and DermDx integrated through the existing SDK.

• OptimaScan, the product.

  Magnetic iPhone mount + rotating arm device shipping under the DermEngine brand. ~1-minute full-body capture, regulatory coverage across HIPAA / GDPR / Australia / NZ / Switzerland.

The recognition mattered less for the press hit than for what it confirmed: a four-figure capture stack on consumer hardware was clinically credible enough that a national agency — Innovative Solutions Canada via IndigenousTech.ai — would put it through pilot procurement. That signal is what made the productization decision — go from POC to shipping rig — easy to commit to.

The same pattern recurred on the Vision Pro side: a three-week sprint shipped a native visionOS app on top of the existing DermEngine SDK, validated live at the AAD booth with real dermatologists running the workflow before the show floor opened. Both bets cleared the same bar — credible enough for procurement, fast enough to be cheap, narrow enough to ship — and that’s what unlocked the full integration of capture and review on the same platform.

Outcome

What I’d do differently

• Pull the spatial-review thesis forward by a year. DermVR sat as an internal prototype longer than it should have. If we’d shipped a Quest build to a small pilot cohort earlier, the Vision Pro three-week sprint would have been a port, not an invention.
• Productize earlier, not later. Two government agencies expressing pilot interest at the POC stage was the signal — turning that into a shipping rig should have started in parallel with the AAD demo, not after.
• Bundle capture + review as one buyer story. The two chapters were sold separately at AAD; in retrospect the strongest pitch is the closed loop — a $2M alternative for capture plus an immersive review surface, on the same platform, paid for by the same procurement line.