SlicerNNInteractive brings interactive, AI-assisted segmentation of 3D medical images into 3D Slicer. It connects Slicer's image viewing and annotation tools with nnInteractive, the promptable segmentation framework developed by Fabian Isensee and colleagues. You indicate a structure, inspect the proposed segmentation, and refine it with further prompts. This makes it easier to create detailed annotations without outlining every slice by hand.

The plugin supports points, scribbles, bounding boxes, and lassos. These interactions let you identify the target and correct areas that need attention while working directly with the volume. The result is a 3D segmentation that you can inspect in Slicer's slice views and 3D viewer. Keyboard shortcuts help keep the annotation process moving when working through multiple structures.

Watch the interactive workflow

The demonstration below shows the prompting workflow in Slicer. Each interaction gives the model more information about the intended structure, so you can build up a segmentation and correct its boundaries as you go.

The extension separates the annotation interface from model inference through a client-server architecture. This allows you to use Slicer on your workstation while running the model on a GPU server, or run both on the same machine. See the repository for installation instructions and the current feature set.

Publication update: the JOSS paper below was published in July 2026, after the original date of this post.

Publications

Please use this when referring to the post or publication: