Today we're thrilled to announce that Dinmar Labs has signed a research and development agreement with a major U.S. university and its academic medical center — the first of what we believe will be several deep clinical partnerships.
What the partnership involves
Under the agreement, our teams will collaborate on the validation and further development of Dinmar's AI models using de-identified imaging data and clinical expertise from the university's radiology department. Specifically, we will:
- Evaluate our lung nodule detection model on a large, independent retrospective dataset
- Compare AI-assisted reads against unaided radiologist reads on the same studies
- Measure the impact of AI triage on time-to-review for urgent findings
- Co-develop new detection capabilities guided by clinical need
"Partnering with academic medicine keeps us honest. Independent data and expert scrutiny are the only way to know whether a model is truly ready."
— Roger Dinmar, Founder & CEOWhy independent validation matters
Models can perform impressively on the data they were developed with and then falter in a new environment. Testing on data from a different institution — different scanners, protocols and patients — is the true test of generalization. Our partner's radiologists will also help us understand how AI findings fit into real reading workflows, which is just as important as raw accuracy.
Building toward clinical trials
This retrospective work lays the foundation for prospective clinical trials, in which the AI runs alongside clinicians on new cases as they happen. We expect to share initial findings from the collaboration later this year.
Building a Data Pipeline Clinicians Can Trust