PulseAugur
LIVE 10:14:46
research · [2 sources] ·
0
research

DiffuSAM adapts SAM2 for prompt-free medical image segmentation

Researchers have developed DiffuSAM, a novel approach that adapts the SAM2 segmentation model for medical imaging without requiring user prompts. This method utilizes a diffusion prior to generate segmentation mask-like embeddings, which are then integrated into SAM2's decoder. The system is designed to maintain spatial consistency across medical image volumes by conditioning the diffusion prior on previously segmented slices. Evaluations on BTCV and CHAOS datasets demonstrate competitive performance in few-shot and source-free unsupervised domain adaptation scenarios. AI

Summary written by gemini-2.5-flash-lite from 2 sources. How we write summaries →

IMPACT Enables prompt-free medical image segmentation, potentially reducing the need for expert annotation and fine-tuning.

RANK_REASON Academic paper introducing a new method for medical image segmentation.

Read on arXiv cs.CV →

COVERAGE [2]

  1. arXiv cs.CV TIER_1 · Tal Grossman, Noa Cahan, Lev Ayzenberg, Hayit Greenspan ·

    DiffuSAM: Diffusion-Based Prompt-Free SAM2 for Few-Shot and Source-Free Medical Image Segmentation

    arXiv:2604.24719v1 Announce Type: new Abstract: Segmentation models such as Segment Anything Model (SAM) and SAM2 achieve strong prompt-driven zero-shot performance. However, their training on natural images limits domain transfer to medical data. Consequently, accurate segmentat…

  2. arXiv cs.CV TIER_1 · Hayit Greenspan ·

    DiffuSAM: Diffusion-Based Prompt-Free SAM2 for Few-Shot and Source-Free Medical Image Segmentation

    Segmentation models such as Segment Anything Model (SAM) and SAM2 achieve strong prompt-driven zero-shot performance. However, their training on natural images limits domain transfer to medical data. Consequently, accurate segmentation typically requires extensive fine-tuning and…