The scan
never stops.
Isadorae turns the camera a clinician already wears — and an ordinary mouth mirror — into a continuous, real-time 3D reconstruction of every surface. Anterior and posterior. Direct and reflected. No wand, no dedicated pass.
Two industries already proved half of this. Neither proved both halves.
Intraoral scanners solved accuracy. Point-of-view cameras solved ergonomics. Nobody had put continuous, hands-free capture and true 3D reconstruction in the same device — including through a mirror.
Direct line-of-sight wands
Best-in-class accuracy, but the clinician has to stop, pick up the wand, and deliberately scan — an interruption disconnected from the moment of treatment. Posterior surfaces still need direct sight.
Loupe- and headlight-mounted
Already prove the ergonomics: always-on, hands-free, aligned to the clinician's own point of view, no thought required to film. What they capture is a flat 2D feed — never a model.
One rig, both problems solved
Continuous POV — direct and mirror
The same hands-free point of view, but every frame becomes 3D — including posterior teeth seen only through the mirror the clinician already holds. The wand disappears. The record never stops.
A mirror reflection is just a second camera, once you know its plane.
This connects to established photogrammetry and SLAM research on planar mirrors — the standard technique is to treat the reflection as a virtual camera and unfold it into a shared 3D frame.
A forehead camera, an ordinary mirror
The clinician wears the same camera already used with loupes and headlight. The only addition is a small AprilTag fiducial fixed to the mirror handle.
The mirror becomes a virtual camera
A planar mirror reflection is optically equivalent to a camera positioned at the mirror image of the real one, reflected across the mirror plane.
The marker gives away the plane
Every frame that sees the AprilTag yields the mirror's exact 6DOF pose — recovered automatically, with no separate calibration step.
Reflected geometry, unfolded
That pose lets the reflected image of a posterior tooth be unfolded into the same global coordinate frame as everything captured head-on — then standard structure-from-motion applies.
Mirror optics and an unstable rig raise the bar past a stationary scanner.
Intraoral scanners solved reconstruction with a stationary wand and direct line of sight. A head-mounted camera shooting through a hand-held mirror inherits every one of dentistry's real capture conditions on top of that.
Real-time mirror-plane pose
Recovering the mirror's full 6DOF pose from a small fiducial, every frame, while the rig it's mounted to never holds still.
A genuine dual pipeline
Direct views and mirror views are reconstructed separately, each anchored by its own fiducials, then fused into one continuous model.
Sub-100-micron budget
Restorative dentistry's tolerance for margins and contacts, reached from a handheld, continuously moving egocentric rig — not a bench-mounted scanner.
Real capture conditions
Motion blur, saliva and blood, mirror fog, shallow depth of field, patient movement from breathing and swallowing, glare on wet enamel.
Soft tissue in the way
Cheeks and tongue routinely occlude the surface being captured — the reconstruction has to tolerate that, not fail on it.
Fusion at continuous volume
Continuous recording produces far more data than a scan pass. SLAM and Gaussian-splat fusion have to run live, not offline.
A spatial substrate, not a flat chart.
Continuous 3D reconstruction gives every voice note, finding, and X-ray a real 3D coordinate to anchor to — turning the chart into something you can query by location, not just by date.
Notes, anchored to a coordinate — not a line
Voice notes and findings attach to an actual 3D point on the tooth as the clinician speaks, instead of a timestamped line in a flat chart. The record becomes a spatial EHR: query it by location on the arch, not only by visit date.
X-rays registered into the same frame
Radiographs and other imaging register directly onto the continuous 3D reconstruction, so optical and radiographic findings live in one coordinate system instead of separate files a clinician has to mentally align.
Findings, streamed back in real time
Caries, calculus, recession, and cracks stream back to the clinician as they work — spoken aloud, shown on-screen, or overlaid in AR through Ray-Ban Meta Display glasses. The same detection other companies run on a single still image, running continuously on live video.
The system knows what's missing
Guided positioning tells the clinician which surfaces haven't been captured yet, closing the coverage loop instead of leaving it to memory — and catching a bleed or a missed contact nobody was looking at in the moment.
Every visit, comparable to the last
Each procedure becomes a 3D snapshot in the same coordinate frame as the one before it. Auto-generated reports turn continuous capture into documentation for training, liability, and insurance without extra work at the chair.
What each category of instrument actually gives the clinician.
| Instrument | Other companies | Other companies | Isadorae |
|---|---|---|---|
| Capture mode | Deliberate scan pass | Continuous POV video | Continuous POV — direct + mirror |
| Line of sight | Direct only | Direct only | Direct and reflected |
| Hands-free | No — handheld wand | Yes | Yes |
| Output | 3D model | 2D video | Continuous 3D model |
| In-procedure feedback | Post-scan | None | Live, while treating |
| Documentation | Manual, per-scan | Raw footage | Auto-generated, coordinate-anchored |
Isadorae removes it.
A forehead camera the clinician already wears. A mirror already in hand. Every surface, continuously, in 3D.