Technology
AI that serves the report, not one that replaces the radiologist.
What powers Spark today
Open models from the market, hosted and operated in France on HDS-certified infrastructure, with no dependence on a foreign AI provider to process your data.
Speech recognition specialised in radiology, trained on the specialty's vocabulary.
Your report templates as the framework: Spark fills in the structure you defined; it does not invent sections.
Our medical knowledge base
Since it was founded, Doctreen has been formalising the reasoning of expert radiologists as decision trees: 110 trees and 550,000 nodes to date (Doctreen knowledge base, September 2026), built with physicians.
“Having decision trees in areas where we are less expert helps raise the quality of the report.”
Our vision: hybrid AI, by 2027
Marty-CR is Doctreen's hybrid AI engine: it combines this structured knowledge base with generative models. It is not in production. Its release is targeted for 2027, as part of a CE marking process.
The radiologist stays in control
No Doctreen AI signs in place of the physician. Every report is reviewed, edited if needed and validated by the radiologist, in their RIS.
