MedSAM (Segment Anything in Medical Images)
Created by Jun Ma, PhD, Bo Wang, PhD & UHN Team · University Health Network (UHN) / University of Toronto / Vector Institute (Academic Lab / University)
Universal promptable foundation model for medical image segmentation across 10+ imaging modalities
What This Tool Does
Developed by Bo Wang's lab at the University of Toronto and University Health Network, MedSAM adapts the Segment Anything Model (SAM) on over 1.5 million medical image-mask pairs across CT, MRI, ultrasound, endoscopy, dermoscopy, and fundus photography.
What Makes It Useful or Distinctive
Provides both a Python foundation model and a 3D Slicer plugin allowing researchers to segment arbitrary lesions with a single bounding-box prompt.
How to Get Started Immediately
Clone the MedSAM repository or install the MedSAM plugin inside 3D Slicer to segment 2D and 3D volumes interactively.
Access, Source Licensing & Clinical Intended Use
- Cost & Access Classification
- Completely free to use
- None.
- Source Code & License
- Apache-2.0
- Open-source (OSI-approved license)
- Intended Use & Regulatory Status
- Research & investigational use
- Verify institutional compliance and local regulatory clearance before any patient-facing clinical use.
- Technical Requirements
- Local Installation / Package
- Account required: No · Offline capable: Yes
Connect This Tool to Your Research Workflow on Trace.Science
Find collaborators experienced with MedSAM (Segment Anything in Medical Images), recruit contributors for a study using this tool, or cite it in your manuscript Methods section.
Launch & Source Links
Visit Project SiteView Source RepositoryDocumentation & TutorialsFeatured in Collections
Comparable Verified Tools
Open-source Blender-based 3D human anatomy atlas derived from BodyParts3D
Completely free to use · CC BY-SA 4.0 (3D models) / GPL-3.0 (Blender scripts)Multi-scale 3D spatial reference atlas of healthy human organs and single-cell tissue maps
Completely free to use · MIT License (UI & code) / CC BY 4.0 (Atlas models)Ultra-high-resolution 20-micrometer 3D histological model of the human brain
Completely free to use · Apache-2.0 (Siibra Explorer) / CC BY-NC-SA 4.0 (BigBrain dataset)WebGL-based volumetric viewer for petabyte-scale connectomics and neuroimaging datasets
Completely free to use · Apache-2.0