MedSAM

MedSAM is a foundational model for medical image segmentation, fine-tuned from Meta’s Segment Anything Model (SAM) on an unprecedentedly large dataset of over 1.5 million medical image-mask pairs across 10 imaging modalities.

Key Capabilities

• Universal Segmentation: Generalizes robustly across 11 different modalities including CT, MRI, X-Ray, ultrasound, pathology, and endoscopy without requiring task-specific retraining.
• Zero-Shot Transfer: Demonstrates strong zero-shot segmentation performance on unseen medical tasks when provided with a simple bounding box prompt.
• Foundation for Downstream Tasks: Serves as a powerful backbone that can be efficiently fine-tuned for specialized, high-accuracy clinical segmentation applications.