HUR-MACL: High-Uncertainty Region-Guided Multi-Architecture Collaborative Learning for Head and Neck Multi-Organ Segmentation
This work addresses a critical need in medical imaging for radiation therapy by enhancing segmentation accuracy for head and neck organs, though it appears incremental as it builds on existing hybrid architecture approaches.
The paper tackles the problem of inaccurate segmentation of small, complex organs in head and neck radiation therapy by proposing a model that identifies high-uncertainty regions and uses multiple architectures to improve accuracy, achieving state-of-the-art results on three datasets.
Accurate segmentation of organs at risk in the head and neck is essential for radiation therapy, yet deep learning models often fail on small, complexly shaped organs. While hybrid architectures that combine different models show promise, they typically just concatenate features without exploiting the unique strengths of each component. This results in functional overlap and limited segmentation accuracy. To address these issues, we propose a high uncertainty region-guided multi-architecture collaborative learning (HUR-MACL) model for multi-organ segmentation in the head and neck. This model adaptively identifies high uncertainty regions using a convolutional neural network, and for these regions, Vision Mamba as well as Deformable CNN are utilized to jointly improve their segmentation accuracy. Additionally, a heterogeneous feature distillation loss was proposed to promote collaborative learning between the two architectures in high uncertainty regions to further enhance performance. Our method achieves SOTA results on two public datasets and one private dataset.