AI-CVD

AI-CVD® is an AI-powered software tool that automatically analyzes CT scans to provide quantitative measurements of cardiovascular, pulmonary, and metabolic features including coronary artery calcium scores, cardiac chamber volumes, epicardial fat, aortic and valve calcifications, lung density, liver density, bone mineral density, and muscle/fat composition. It is intended to aid healthcare providers in identifying patients at risk for various conditions to support preventive care planning, but does not provide diagnostic interpretation or risk prediction itself.

AI-CVD® uses deep learning-based segmentation built on the open-source TotalSegmentator framework with nnU-Net architecture, trained on contrast-enhanced and non-contrast CT scans. Like the predicate devices, it automatically defines regions of interest, measures volumes and densities within those regions, operates on DICOM CT images, and requires human expert review and approval of segmentation results. The device runs on Linux and performs retrospective measurements from existing CT scans.

Not stated in this summary.

AI-CVD® is substantially equivalent to its predicate device AutoChamber® (K240786) and reference device Auto-BMD (K213760) because it performs the same intended function of providing automated quantitative imaging measurements from CT scans using similar technological approaches (deep learning segmentation, automatic ROI definition, volume and density measurement, human expert review). Clinical validation through retrospective analysis of large population cohorts (MESA and FHS) demonstrated comparable safety and effectiveness to manual expert measurements and gold standard reference tests, with no new safety or effectiveness issues raised.