top of page
Can AI read a patient record in any language?
Multinational trials recruit from records in Dutch, French, German, English... Language shouldn't change whether someone qualifies for a trial.
Â
We translated 100 patient profiles and screened each version against the same trials. The results were closer than we expected, and the differences we did find came from somewhere surprising.
Monsana Research Papers
New
Our studies on how AI pre-screening performs across languages and AI models.
OpenAI or Mistral: does the model matter?
A US model or a European one? For hospitals weighing GDPR and data sovereignty, it's a key question.
Â
We ran the same patients through both. Two runs of the same model disagreed more than the two model families did. So what should drive the choice?
bottom of page