Getting started
Introduction
What MedRun is and how requests, inputs and results fit together.
MedRun is a library of medical-imaging interpretation models. You upload a study, a model reads it, and you get back one structured result: findings, measurements, scores, guidance, evidence files and, if you ask for it, a draft report. Use it from the console or call the same models through the API.
Concepts
| Term | Meaning | Example |
|---|---|---|
| Category | The imaging device family. It sets the input contract and the viewer. | angio (Angiography & Cath Lab) |
| Model | An exam type inside a category: the study a doctor has in hand. | angio/coronary |
| Endpoint | A separate run with its own input, result and price. Every clinical model has interpret; some add tools. |
angio/coronary/interpret, angio/coronary/qca |
| Section | A part of one run's result. Never called or priced on its own. | angio/coronary#stenosis |
| Request | One call to an endpoint, from submit to result. | req_01K6… |
The bare model id is an alias of its main run: angio/coronary = angio/coronary/interpret.
One run, one complete result
A run computes every section its input supports and returns them together, at one price. Each section has a status. When a section is missing, it says why: not_in_input (for example, needs an RCA run), not_assessable (image quality) or not_applicable.
Next steps
- Quickstart: run a model from the console or with one API call.
- Queue API: submit, follow and fetch requests.
- Result format: what comes back.