40,000–80,000 patients die annually in the U.S. from diagnostic errors.
Artificial intelligence is enabling extraordinary advances in healthcare. It promises to reduce costs and improve healthcare outcomes. But artificial intelligence is only as accurate—and therefore as effective—as the data it is trained on.
Centaur Labs is accelerating medical AI through trusted, accurate medical data labeling. Instead of relying on medical professionals exclusively based on their credentials, we use collective intelligence and performance incentives to achieve an unprecedented level of accuracy. We’re rethinking how medical decisions are made.
The Centaur approach reimagines the medical second opinion. Patients and healthcare companies will often trust a single doctor’s advice, even though doctors disagree with each other at alarming rates.
This disagreement is pervasive across clinical areas and proven in research studies. Misdiagnoses and errors in medical decisions cost lives.
Born out of research at MIT’s Center for Collective Intelligence, our approach is based on three core principles:
We gather multiple opinions per case, which is directly correlated to an improvement in accuracy.
Our network of medical experts—over 100,000 labelers and counting—includes diverse skill sets, with breadth and depth of expertise.
Opinions are not created equal. We weigh opinions based on each labeler's performance record for each task.
To gather these opinions, one tool we use is our mobile app, DiagnosUs. DiagnosUs gamifies medical annotation to solicit the highest quality labels and brings together a diverse network of labelers from around the world.
We know the stakes are high and medical AI is hard, so we’re relentless about delivering high quality data labels. We collaborate with and continue to invest in our network of experts. We listen to our customers’ challenges and customize solutions to improve their life-saving technology.