TeraRecon Aorta.CT (1.1.0)

TeraRecon Aorta.CT is a software-as-a-medical-device that automatically segments the aorta and identifies 22 anatomical landmarks from contrast-enhanced CT angiography images. It outputs DICOM segmentation files for use by trained professionals in clinical decision-making and procedure planning, without providing a diagnosis. The device is indicated for adult patients but excludes those with pre-existing aortic devices, bicuspid aortic valve anomaly, aortic dissection, aortic rupture, or abdominal metallic devices.

TeraRecon Aorta.CT uses supervised deep learning-based algorithms to process CT DICOM images and output DICOM segmentation and center-of-flow results. The predicate (AutoSeg-H) uses deep learning-based algorithms for similar CT image processing. Both are deployed as containerized applications on standard computers, use the same modality (CT), and produce DICOM-compliant segmentation outputs. The key difference is that the predicate offers additional coronary artery and 4-chamber cardiac analysis tools, whereas the subject device focuses narrowly on aortic segmentation and landmarking.

Not stated in this summary. The submission references DICE (Dice Similarity Coefficient) scoring for segmentation accuracy and Euclidean distance metrics for landmark location validation, but does not cite specific ISO, IEC, or ASTM consensus standards.

Substantial equivalence is supported by identical modality (CT) and algorithm technology class (deep learning), identical intended use of DICOM image processing for clinician-aided segmentation, identical patient population (adults), identical installation method (computer application), and identical output format (DICOM-compliant files). Differences in scope (aortic-only vs. multi-anatomy analysis) and exclusion criteria do not raise different safety or effectiveness questions because the core segmentation and validation methodology is the same. Empirical testing demonstrated the device achieved mean DICE scores of 88–90% and landmark accuracy ≥80%, meeting pre-specified acceptance criteria and matching predicate performance.