Help improve this record. Suggest a correction
Use casesCLINICAL RESEARCH / BIOMEDICAL RESEARCH INSTITUTE · UNIVERSITY-HOSPITAL NETWORK
Rare-disease diagnosis support from dental photographs and radiographs, planned on an on-site appliance.
A six-month pilot of a vision-plus-retrieval pipeline that turns intra-oral images into standardised phenotypes and weighted, source-cited rare-disease hypotheses. Production is designed for a DGX Spark-class appliance on site, air-gap capable, for health-data sovereignty.
- Story
- Stack
- Load
- Measurements
THE SETTING
Who uses it, and where.
Public biomedical research with a university-hospital clinical network and rare-disease expert centres. Environment: clinical research.
Users. Up to 5 administrators and 20 end users (clinicians, geneticists, researchers, doctoral students); 100 to 300 annotated cases for validation.
THE STACK
What runs on the machine.
- Vision model for per-tooth anomaly extraction in standard terminologies.
- Retrieval-augmented reasoning over a structured rare-disease corpus.
- Clinician report and plain-language patient summary generator.
- Optional anonymised research registry.
- Model IDs: not yet published.
Network posture. Pilot on the Understand Tech platform. Production designed with no public-cloud dependency and air-gap capability.
THE WORKLOADS
What it does, day to day.
- Secure upload, automatic quality check and anonymisation of intra-oral photographs and radiographs.
- Per-tooth phenotype extraction with confidence scores.
- Weighted rare-disease hypotheses with suggested gene panels and cited sources.
- Clinician report, patient summary, optional registry entry for retrospective research.
OUTCOME
What changed.
Planned scopePlanned scope: reproducible, traceable phenotyping; faster triage of complex cases to expert centres; a standardised phenotype registry for translational research. The system is explicitly non-binding: hypotheses, never a diagnosis; the clinician validates every output.
EVIDENCE & LIMITS
How to read this record.
The institute is not named. This is a planned deployment described in a statement of work; no result exists yet.
LESSONS & FAILURES
What the team learned.
RELATED RECORDS