Nvidia and Dyania Bring AI Chart Review Onto Hospital Servers

Dyania Health deployed its Synapsis AI software across Cleveland Clinic’s internal servers to review the historical records of over 7,800 heart-valve patients.

Much of a patient’s medical history sits buried in unstructured doctor notes and scan reports rather than clean data fields. The Synapsis AI software reads those notes to answer clinical questions while linking every answer directly back to the source record.

For the study, the AI answered 88 specific questions for each patient who underwent an aortic valve replacement between 2005 and 2021.

Researchers used the resulting dataset to build a predictive model estimating the risk of a replacement valve failing at three, five, and ten years post-surgery. Nvidia announced on Friday that the automated chart review processed up to 16 years of records for the cohort entirely on local hospital hardware.

Running the infrastructure locally ensures that sensitive patient records never leave the hospital network, mirroring architecture seen in UK National Health Service triage tools powered by Nvidia technology. Nvidia said that manual chart reviews of this scale would have taken clinical staff years to complete.

Doctors still review every piece of evidence surfaced by the AI and make all final medical decisions. Lars Svensson, head of Cleveland Clinic’s heart institute, called the deployment “a promising step towards enhanced outcomes and more personalized care for patients,” as reported by Nvidia.

The study demonstrates that the tool can successfully construct a dataset and project risk, but it does not prove that it actively improves patient care. Dyania Health said that the software still requires further outside validation before broader adoption.

Cleveland Clinic also holds an equity stake in Dyania Health, an investment disclosed last year when the hospital initially rolled out the tool for clinical trial recruitment. Major technology providers continue accelerating their push into healthcare data infrastructure.

Google Cloud launched Gemini for healthcare, and Bunkerhill secured $55 million to deploy hospital AI agents. Proven clinical gains for patients remain difficult to locate, and physicians have frequently pushed back against bold marketing claims from technology vendors.