FastSAM-3DSlicer

FastSAM-3DSlicer integrates 2D and 3D Segment Anything Model (SAM) variants for interactive volumetric medical image segmentation with uncertainty quantification.

Key Features:

• Multi-model integration: Incorporates SAM-Med2D, MedSAM, SAM-Med3D, and FastSAM-3D models.
• Real-time volumetric segmentation: Achieves inference times of 1.09 s per volume on CPU and 0.73 s per volume on GPU.
• Uncertainty quantification: Implements an uncertainty quantification scheme leveraging the rapid inference capabilities of FastSAM-3D.
• Automated data handling: Automates processing of raw image data, user prompts, and segmented masks.

Scientific Applications:

• Anatomical structure segmentation: Segments anatomical structures in 3D volumetric medical images for diagnosis and treatment planning.
• Pathological region segmentation: Identifies pathological regions in medical images for disease monitoring.
• Interactive medical image analysis: Enables real-time interactive segmentation without domain adaptation or retraining on unseen image types.

Methodology:

FastSAM-3DSlicer is implemented as an extension on the open-source 3D Slicer platform, integrating 2D and 3D SAM models for efficient, real-time segmentation of 3D volumetric medical images with seamless interaction and visualization.