One model holdsthe state of theoperating room.
Fifty seconds of simulated surgery on a phantom, reconstructed from ordinary cameras with no depth sensor in the room.
One signal, five readings
Surgical phase
Sterile technique
Traffic and turnover
5 roles present
Care-bundle events
Metric 3D twin
42,483 points
Fifty seconds of simulated knee surgery on a phantom, from a public dataset, reconstructed by our own model from ordinary cameras. Each path is the floor position of one labelled role, once per frame. Live, rendered in your browser.
The most expensive thing that goes wrong in an operating room is still audited by hand.
00:30of 00:50
What a spot-check never sees
A clipboard audit sees the room once or twice a quarter. Gray-1 reads it continuously. Hold any second of this clip and the room, the people in it and the floor they have covered are all there, read off one state.
Fifty seconds of simulated knee surgery on a phantom, from a public dataset, reconstructed by our own model from ordinary cameras. Each path is the floor position of one labelled role, once per frame. Held, rendered in your browser.
- 3-5%
- Surgical site infection rate after orthopaedic procedures
- $28K
- Average direct medical cost of a single infection
- 68%
- Infections not identified until after discharge
- 1-2
- Compliance spot-checks a quarter, counted by hand on clipboards
A room is not a list of detectors.
The obvious way to build this is one model per thing you want to know. A phase classifier. A hand-hygiene detector. A door counter. A sterile-field monitor. Together they produce a system that contradicts itself, because nothing makes two separate models agree about who is scrubbed.
Gray-1 keeps one description of the room instead, held across the whole procedure: who and what is present, in what role, in what condition, and it does not lose people when they turn away from a camera and come back. Phase, sterility and care-bundle events are all read off that one description. They cannot disagree, because there is only one answer to read.
- Who is in the room
- and in which role
- What they are touching
- and what they are close to
- Who is scrubbed
- and whether the field is still clean
- Where the case is
- prep, surgery, wrap-up
- What just happened
- a breach, a door, a checklist step
Five readouts, one model underneath them.
Each line carries the most specific fact we can point at, and says where the evidence came from.
Surgical phase
Decided once per second while the case runs, and developed against a public dataset of simulated surgery on a phantom. No clinical footage yet.
Sterile technique
A breach and the phase it happened in come from the same description of the room, so the two can never contradict each other.
Traffic and turnover
Every additional entry during a sterile phase degrades operating-room air, which is the documented mechanism linking traffic to infection risk.
Care-bundle events
The checklist comes from our own clinical team and follows the prevention bundle a theatre already works to. No outcome data yet, on any site.
Metric 3D twin
Cameras only, with no depth sensor in the room. Distances are checked against a laser measurement of the same scene.
Each of these is the same model read a different way. The system page walks through how the state is built and what each readout takes from it.
How Gray-1 worksEverything we find, dated, including the results that went against us.
All notesWe are looking for the operating rooms to be wrong in.
The detection engine exists and the clinical judgement does not scale without a partner. If you run an operating room and think compliance should be measurable, we would like to hear where this falls short.