RefineSeg: Dual Coarse-to-Fine Learning for Medical Image Segmentation
🤖 AI Summary
Medical image segmentation suffers from a scarcity of high-quality pixel-level annotations. To address this, we propose a two-stage coarse-to-fine weakly supervised framework that achieves precise segmentation using only noisy and incomplete coarse-grained annotations—specifically, labels indicating target structures alongside supplementary anatomical regions. Our key innovation is a learnable transition matrix that explicitly models region-level mislabeling and missing annotations within coarse labels. By jointly training on multiple coarse annotation sets, the framework progressively refines network predictions to converge toward the true segmentation distribution. Crucially, we integrate this matrix-based modeling seamlessly into an end-to-end deep neural network. Evaluated on ACDC, MSCMRseg, and UK Biobank datasets, our method significantly outperforms existing weakly supervised approaches and closely matches fully supervised performance. This work establishes a novel paradigm for cost-effective, high-accuracy medical image segmentation.
Application Category
📝 Abstract
High-quality pixel-level annotations of medical images are essential for supervised segmentation tasks, but obtaining such annotations is costly and requires medical expertise. To address this challenge, we propose a novel coarse-to-fine segmentation framework that relies entirely on coarse-level annotations, encompassing both target and complementary drawings, despite their inherent noise. The framework works by introducing transition matrices in order to model the inaccurate and incomplete regions in the coarse annotations. By jointly training on multiple sets of coarse annotations, it progressively refines the network's outputs and infers the true segmentation distribution, achieving a robust approximation of precise labels through matrix-based modeling. To validate the flexibility and effectiveness of the proposed method, we demonstrate the results on two public cardiac imaging datasets, ACDC and MSCMRseg, and further evaluate its performance on the UK Biobank dataset. Experimental results indicate that our approach surpasses the state-of-the-art weakly supervised methods and closely matches the fully supervised approach.
Problem
Research questions and friction points this paper is trying to address.
Reduces reliance on costly expert medical annotations
Models noisy coarse annotations via transition matrices
Achieves precise segmentation with weak supervision
Innovation
Methods, ideas, or system contributions that make the work stand out.
Coarse-to-fine segmentation with noisy annotations
Transition matrices model inaccurate annotation regions
Joint training refines segmentation via matrix modeling
🔎 Similar Papers
No similar papers found.
