SimSAM: Zero-Shot Medical Image Segmentation via Simulated Interaction
Published in IEEE ISBI, 2024
Recommended citation: SimSAM: Zero-Shot Medical Image Segmentation via Simulated Interaction (Towle, Chen & Zhou, ISBI 2024). https://arxiv.org/abs/2406.00663
We present SimSAM, a method for zero-shot medical image segmentation that simulates user interactions with the Segment Anything Model (SAM) to enable automatic segmentation without manual prompts. By learning to predict effective interaction points, SimSAM achieves competitive performance with fully supervised methods while requiring no target-domain training data.
Recommended citation: SimSAM: Zero-Shot Medical Image Segmentation via Simulated Interaction (Towle, Chen & Zhou, ISBI 2024).