AutoContour (RADAC V5)

AutoContour Model RADAC V5 is a software-only medical device that automatically contours anatomical structures in CT and MR images to assist radiation treatment planners in preparing for radiation therapy. It uses deep-learning-based structure models trained on multi-institutional imaging datasets to generate initial contours that users can review, modify, and export as DICOM-compliant structure sets for import into radiation therapy treatment planning systems.

The device consists of three components: a .NET client application for Windows, a local Windows agent service, and a cloud-based automatic contouring service. It supports CT or MR input for contouring and registration/fusion, with PET/CT input for registration/fusion only. RADAC V5 includes 420 CT-based and 62 MR-based structure models, expanding from the predicate's 260 CT and 35 MR models. Core functionality, data visualization/GUI, and view manipulation capabilities (window/level, pan, zoom, registration) remain equivalent to the predicate.

Not stated in this summary. The document references internal validation protocols and qualitative Likert-scale reviews by clinical experts, but does not cite specific external consensus standards (ISO, IEC, ASTM, etc).

RADAC V5 is substantially equivalent to AutoContour Model RADAC V4 (K242729) because indications for use, patient population, intended users, energy delivery, and core functionality remain unchanged. Verification testing confirmed that features and performance meet predefined acceptance criteria consistent with the predicate. New and updated structure models passed rigorous testing (DSC and Hausdorff Distance metrics) with external clinical validation, and unchanged models passed regression testing. Minor technological differences—primarily the addition of new structure models and integration of models from the reference device Limbus Contour—do not raise significant safety or effectiveness concerns.