Inside the hospital firewall
Dyania Health said it is collaborating with NVIDIA to run its Synapsis AI clinical reasoning software on health systems' own computing infrastructure, across an entire patient population. Synapsis AI reads a patient's full medical record — years of notes, operative reports, imaging reports and lab results — and answers clinical questions with evidence from the chart, the company said. Because the software runs on NVIDIA accelerated computing inside the health system's private network, patient records are not transferred outside it.
A study of 7,812 patients
The first work from the collaboration is a Cleveland Clinic study of 7,812 patients who received an aortic valve replacement between 2005 and 2021. Running on NVIDIA accelerated computing inside Cleveland Clinic's environment, Synapsis AI identified the patient population and answered 88 clinical questions for each patient, covering the valve and its characteristics, co-existing conditions, laboratory trends, medications, follow-up ultrasound measurements and later clinical events. The study drew on up to 16 years of care per patient, much of it held in narrative notes and echocardiogram reports. It produced a detailed clinical dataset in far less time than a manual review would take, the company said.
A predictive machine-learning model was also trained to forecast valve failure and reintervention risk timelines for every patient. The type of valve implanted was the strongest determinant of how quickly it deteriorated, while age and other medical conditions shaped whether and when a second procedure was performed, according to the release. The model estimates an individual patient's risk at three, five and 10 years and revises the outlook with each new ultrasound; the release said the tool is used to help inform care but requires external validation from clinicians.
What the companies plan
Under the collaboration the companies will work to bring Synapsis AI on NVIDIA infrastructure to health systems in the U.S. and internationally, publish joint case studies and integrate the tool with NVIDIA's healthcare software stack. Areas they are evaluating include NVIDIA Confidential Computing for training and inference, NVIDIA Dynamo for distributed inference serving, Multi-Instance GPU partitioning across concurrent clinical workloads, and scaling to many Blackwell Ultra GPUs as deployments move from batch processing toward near-real-time clinical insight at the point of care.
