This research introduces a novel approach combining the Segment Anything Model (SAM) with preference optimization for medical imaging, dramatically reducing the need for expert annotations.

• Develops an efficient prompting strategy that automatically generates high-quality annotations
• Implements preference optimization that refines segmentation without human intervention
• Achieves superior performance across diverse medical imaging modalities (X-ray, ultrasound, CT)
• Demonstrates effectiveness for lung segmentation, breast tumor detection, and multi-organ segmentation

This advancement matters for healthcare by making accurate medical image segmentation more accessible, reducing costs and expert time requirements while maintaining high diagnostic quality.

Enhancing SAM with Efficient Prompting and Preference Optimization for Semi-supervised Medical Image Segmentation