Automated Pathological Primitive Segmentation Using a Zero-Shot Mask Generation Approach with a Visual Foundation Model

A novel approach that adapts a pre-trained natural image encoder for detection-based region proposals, enabling end-to-end pathological primitive segmentation without the need for additional training or fine-tuning.