Circle Cardiovascular Imaging’s AI models offer smart imaging with secure data
Circle Cardiovascular Imaging’s cvi42 incorporates advanced AI models to automate key tasks such as segmentation, contouring and quantitative measurement across cardiac MRI and CT studies. However, the company says that beyond functionality, it is equally important to understand how this AI operates in practice, particularly in terms of data handling, model behaviour and clinical integration.
A defining feature of cvi42’s AI framework is that all processing occurs within the customer’s secure environment. Imaging data and derived outputs remain on local infrastructure, whether on a work-station or a customer-managed server. This fully local architecture aligns with hospital IT policies and data governance requirements. Clinicians can therefore benefit from AI-driven analysis without introducing additional risks to patient privacy or network security.
The AI models embedded in cvi42 are developed within Circle CVI’s controlled research and development environment. Using diverse, representative
datasets, Circle CVI’s data science and clinical research teams train and validate algorithms through supervised learning approaches. Models learn to identify anatomical structures such as the left ven-tricle, myocardium and aortic root by comparing predictions against expert-annotated reference data.
Once an algorithm meets defined clinical and regulatory performance standards, it is finalised, encrypted and integrated into the software. The model is effectively frozen, meaning its behaviour is fixed and does not evolve after deployment. This ensures that every installation of cvi42 delivers consistent, reproducible results based on a validated model, the company says. Importantly, cvi42’s AI does not learn from data processed at the customer site. Each analysis applies pre-trained model para-meters without storing patient data, transmitting information externally, or adapting based on user edits or prior cases.
Read this report on page 6 of the August 2026 issue of RAD Magazine.
