SparseMamba-PCL: Scribble-Supervised Medical Image Segmentation via SAM-Guided Progressive Collaborative Learning

2025-03-03

AI Summary

To address the challenges of generating dense segmentation masks and accurately localizing object boundaries under sparse scribble annotations in medical imaging, this paper proposes a scribble-driven progressive collaborative learning framework. Our key contributions are: (1) a novel scribble-guided boundary estimation module that explicitly models target contours; (2) a cross-architecture feature fusion mechanism integrating Med-SAM and Sparse Mamba; and (3) a skip-sampling Sparse Mamba network for efficient long-range spatial dependency modeling. Leveraging scribble-enhanced propagation, progressive collaborative training, and decoder fine-tuning, our method achieves state-of-the-art performance on ACDC, CHAOS, and MSCMRSeg benchmarks—outperforming nine existing SOTA methods across Dice score, Hausdorff distance (HD95), and average surface distance (ASSD). These results demonstrate the effectiveness of our approach for high-precision segmentation under sparse supervision.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Multi-instance/Multi-view Learning
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs

Abstract

Scribble annotations significantly reduce the cost and labor required for dense labeling in large medical datasets with complex anatomical structures. However, current scribble-supervised learning methods are limited in their ability to effectively propagate sparse annotation labels to dense segmentation masks and accurately segment object boundaries. To address these issues, we propose a Progressive Collaborative Learning framework that leverages novel algorithms and the Med-SAM foundation model to enhance information quality during training. (1) We enrich ground truth scribble segmentation labels through a new algorithm, propagating scribbles to estimate object boundaries. (2) We enhance feature representation by optimizing Med-SAM-guided training through the fusion of feature embeddings from Med-SAM and our proposed Sparse Mamba network. This enriched representation also facilitates the fine-tuning of the Med-SAM decoder with enriched scribbles. (3) For inference, we introduce a Sparse Mamba network, which is highly capable of capturing local and global dependencies by replacing the traditional sequential patch processing method with a skip-sampling procedure. Experiments on the ACDC, CHAOS, and MSCMRSeg datasets validate the effectiveness of our framework, outperforming nine state-of-the-art methods. Our code is available at href{https://github.com/QLYCode/SparseMamba-PCL}{SparseMamba-PCL.git}.

Problem

• Improves scribble-supervised medical image segmentation accuracy.
• Enhances feature representation using Med-SAM and Sparse Mamba network.
• Introduces skip-sampling for better local and global dependency capture.

Innovation

• Propagates scribbles to estimate object boundaries
• Fuses Med-SAM and Sparse Mamba feature embeddings
• Uses skip-sampling for local and global dependencies