New framework enhances medical image segmentation with SAM and active learning

Researchers have developed SUGFW+, a novel framework designed to improve medical image segmentation models, particularly in scenarios with limited annotated data. This approach leverages the Segment Anything Model (SAM) for feature extraction and uncertainty calculation, integrating these elements into a Patch-based Global Distinct Representation module. The framework then employs a Greedy Selection with Cluster and Uncertainty strategy to select diverse and informative samples for annotation, which are subsequently used in an Uncertainty-Prompted Fine-Tuning process for SAM. Experiments across four datasets show SUGFW+ outperforming existing methods in cold-start active learning for medical image segmentation. AI

IMPACT This framework could significantly reduce the cost and time required for training accurate medical image segmentation models by optimizing the use of limited annotated data.

RANK_REASON The cluster contains a research paper detailing a new framework for medical image segmentation. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• Greedy Selection with Cluster and Uncertainty (GSCU)
• Medical Image Segmentation
• Patch-based Global Distinct Representation (PGDR)
• Patch-level Feature and Uncertainty Calculation (PFUC)
• Sam
• Segment Anything Model (SAM)
• SUGFW+
• Uncertainty-Prompted Fine-Tuning (UPFT)

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →