Cleveland Clinic used Dyania Health’s Synapsis AI to review longitudinal records from more than 7,800 people who received aortic valve replacements between 2005 and 2021. The system combined notes, laboratory results and later echocardiograms to support a study of valve deterioration and the risk of another intervention.
Synapsis is designed to extract traceable clinical signals from unstructured records that would otherwise require extensive manual chart review. Each finding links back to its source in the patient record. Dyania says the platform completed work across the full research cohort that would have taken clinicians years to perform by hand.
The system runs on GPU infrastructure inside a health organization’s firewall, so the records do not leave the hospital. Dyania trains and builds models with Nvidia’s NeMo libraries and uses TensorRT tools to optimize inference on local servers. The article is an Nvidia customer case study, so its performance claims come from the vendors and participating institution rather than an independent product comparison.
This is assistive software, not an autonomous clinical decision-maker. Clinicians must confirm the evidence and decide what action, if any, to take. Similar record review could help identify trial candidates or care gaps, but each use would need its own clinical validation and governance.