Search papers, labs, and topics across Lattice.
This paper introduces CSWinUNETR, a novel backbone designed for the segmentation of thin anatomical structures in medical images, addressing challenges such as low contrast and class imbalance. The model utilizes cross-shaped stripe self-attention and cyclic shifts to enhance information exchange and preserve fine details, while a sparse-control dynamic snake convolution reconstructs dense curvilinear kernels from sparse predictions. Extensive evaluations across multiple medical imaging benchmarks reveal that CSWinUNETR significantly outperforms existing state-of-the-art methods without relying on task-specific post-processing or topology-aware losses.
CSWinUNETR achieves unprecedented segmentation accuracy for thin anatomical structures, outperforming existing methods without complex post-processing.
Accurate segmentation of thin, tortuous anatomical structures, such as retinal vessels, cerebral vasculature, and facial wrinkles, remains challenging due to low contrast, frequent discontinuities, and severe class imbalance. Although recent convolutional and Transformer-based models have improved performance, they often yield fragmented predictions and fail to recover fine branches. We propose CSWinUNETR, a general-purpose backbone for 2D and 3D thin-structure segmentation. It employs cross-shaped stripe self-attention to model long-range principal-axis context and incorporates cyclic shifts to enhance information exchange across stripes. To better preserve fine-grained details, we further introduce a detail-enhanced multi-scale self-attention module that aggregates contextual features from multi-resolution representations. In addition, we propose sparse-control dynamic snake convolution, which reconstructs reliable dense curvilinear kernels from sparsely predicted control points to better follow tortuous geometry. Extensive experiments on four benchmarks across ophthalmology, neurovascular imaging, and dermatology demonstrate that CSWinUNETR consistently outperforms state-of-the-art methods without task-specific post-processing or topology-aware losses. The code is available at https://github.com/labhai/CSWinUNETR.
