FastSAM-3DSlicer: A 3D-Slicer Extension for 3D Volumetric Segment Anything Model with Uncertainty Quantification

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computational Anatomy

Background:

• Accurate segmentation of medical images is vital for clinical decision-making.
• Segment Anything Model (SAM) shows potential but lacks user-friendly 3D integration.
• Existing SAM applications are limited in volumetric medical imaging workflows.

Purpose of the Study:

• To develop a user-friendly 3D Slicer extension integrating 2D and 3D SAM models for medical image segmentation.
• To enable efficient, real-time interactive segmentation of 3D volumetric medical images.
• To streamline medical image analysis workflows with automated segmentation handling.

Main Methods:

• Developed FastSAM-3DSlicer, a 3D Slicer extension incorporating SAM-Med2D, MedSAM, SAM-Med3D, and FastSAM-3D.
• Integrated automated handling of raw image data, user prompts, and segmented masks.
• Implemented an uncertainty quantification scheme for enhanced reliability.

Main Results:

• FastSAM-3DSlicer provides seamless interaction and visualization for 3D medical images.
• FastSAM-3D achieved low inference times: 1.09s (CPU) and 0.73s (GPU) per volume.
• The extension is suitable for real-time interactive segmentation and clinical integration.

Conclusions:

• FastSAM-3DSlicer offers an efficient, precise, and user-friendly platform for 2D and 3D interactive volumetric medical image segmentation.
• The tool enhances the practical application of SAMs in medical image analysis.
• Publicly available code facilitates adoption and further development.