RibSeg Dataset and Strong Point Cloud Baselines for Rib Segmentation from CT Scans

Manual rib inspections in computed tomography (CT) scans are clinically critical but labor-intensive, as 24 ribs are typically elongated and oblique in 3D volumes. Automatic rib segmentation methods can speed up the process through rib measurement and visualization. However, prior arts mostly use in-house labeled datasets that are publicly unavailable and work on dense 3D volumes that are computationally inefficient. To address these issues, we develop a labeled rib segmentation benchmark, named RibSeg, including 490 CT scans (11,719 individual ribs) from a public dataset. For ground truth generation, we used existing morphology-based algorithms and manually refined its results. Then, considering the sparsity of ribs in 3D volumes, we thresholded and sampled sparse voxels from the input and designed a point cloud-based baseline method for rib segmentation. The proposed method achieves state-of-the-art segmentation performance (Dice\(\approx95\%\)) with significant efficiency (\(10\sim40\times\) faster than prior arts). The RibSeg dataset, code, and model in PyTorch are available at https://github.com/M3DV/RibSeg. <strong>Note:</strong> This repository provides rib segmentation ("RibFrac31-rib-seg.nii.gz") and centerline ("RibFrac31-rib-cl.nii.gz") <em>annotations</em> for 490 cases in RibFrac dataset. Please download the corresponding CT <em>images </em>("RibFrac31-image.nii.gz") at https://ribfrac.grand-challenge.org/ (1-click registration is needed via <em>"Join"</em>).计算机断层扫描（CT）影像中的手动肋骨检视在临床中至关重要，但却耗时费力——由于3D体素内的24根肋骨通常呈细长且倾斜分布。自动化肋骨分割方法可通过肋骨测量与可视化流程，大幅提升该任务的效率。然而，现有研究大多采用无法公开获取的内部标注数据集，且基于稠密3D体素开展处理，计算效率较低。为解决上述问题，我们构建了一款标注化肋骨分割基准数据集RibSeg，该数据集源自公开数据集，包含490例CT扫描影像（共计11719根独立肋骨）。在真值标签（ground truth）生成环节，我们采用了现有基于形态学的算法，并对其输出结果进行了人工修正。随后，考虑到3D体素内肋骨的分布稀疏性，我们对输入数据进行阈值化处理并采样稀疏体素，同时设计了一款基于点云的肋骨分割基准方法。所提方法实现了当前最优（state-of-the-art）的分割性能，Dice系数约为95%，同时计算效率显著提升，比现有研究快10~40倍。RibSeg数据集、代码以及PyTorch框架下的模型可在https://github.com/M3DV/RibSeg获取。<strong>注：</strong>本仓库为RibFrac数据集中的490例样本提供了肋骨分割标注文件（"RibFrac31-rib-seg.nii.gz"）与中心线标注文件（"RibFrac31-rib-cl.nii.gz"）。请前往https://ribfrac.grand-challenge.org/下载对应的CT影像文件（"RibFrac31-image.nii.gz"），需通过"Join"按钮完成一键注册。