pyaar · dental annotation
Operationalizing clinical annotation for dental AI.
Dental AI companies need validated, adjudicated datasets from real clinicians. The secure viewer and the storage already exist. What is missing between “platform” and “submission-ready dataset” is the operated layer: recruiting the clinicians, running the multi-rater reads, adjudicating disagreement, and packaging the result.
This is the read-and-annotate loop that layer runs on. It opens dental CBCT in a stock open-source viewer, captures structured reads from several dentists, and lets a clinical lead review them for a validation dataset. Dental first because it is the lowest-friction wedge. Radiology is the same machine, pointed at a harder market later.
The full-service stack, one accountable partner
- PyaarRight-size the dataset. Power / sample-size analysis for the submission endpoint, so you gather exactly the annotations your validation study needs.
- KitwareThe imaging platform. VolView and Girder for secure viewing, permissions, and self-hosted governance. The data plane.
- PyaarThe operated annotation layer. A credentialed contract-clinician network, multi-rater campaigns, and a clinical lead who adjudicates and signs off.
- InnoliticsCompliant engineering. Medical software practices and validation design so the work holds up for regulatory review.
- CyberCrestSecurity. Cybersecurity and compliance gap analysis across the pipeline.
The sandbox runs on dental cone-beam CT from the 3D Slicer / Kitware sample data (pre and post orthognathic surgery). In a real deployment, PHI stays on the customer’s own host.