Contour ProtégéAI+

Contour ProtégéAI+ is software that automatically creates contours (segmentations) on medical images of CT and MR modalities using machine-learning algorithms. It assists trained medical professionals in processing digital medical images for applications including quantitative analysis, adaptive therapy planning, and archiving contours for patient follow-up and radiation therapy treatment planning.

The subject device contains one new 4.2.0 neural network model and three updated 4.2.0 models trained on the same architecture as the predicate without changes. The models cover all existing structures for contouring plus additional new structures. All models were tested using the same procedures and acceptance criteria as the predicate, with no unexpected results observed. Operating platform, modalities (CT and MR), cloud-based and local deployment options remain unchanged from the predicate.

Not stated in this summary.

Substantial equivalence is established because: (1) the new and updated neural network models employ the same architecture and training methodology as the predicate without design changes; (2) the four modified models demonstrated non-inferiority compared to the reference device (MIM Maestro atlas-based segmentation) using the same statistical acceptance criteria (Dice score, MDA, cumulative APL, and external user evaluation); (3) all individual structures met predefined acceptance criteria; (4) the indications for use and intended use remain identical to the predicate; and (5) additional anatomical structures (marked in bold in the model list) do not raise new questions of safety or effectiveness, as they follow the same validated segmentation approach.