seg: import nibabel as nib seg = nib.load(file name).get_fdata() # seg is a np array / volume of (512,512,N) with rib labels

cl: import numpy as np cl = np.load(file name)['cl'] # cl is a np array of (24,500,3), each rib contains 500 points

Jiancheng Yang, Shixuan Gu, Donglai Wei, Hanspeter Pfister, Bingbing Ni

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≈95%) with significant efficiency (10∼40× faster than prior arts).

The RibSeg Dataset contains annotations for both rib segmentation and centerline.

Rib Segmentation	Rib Centerline

Overview of RibSeg dataset:

For training data, please download the source CT scans from RibFrac Dataset to ./data/ribfrac directory:

For the source CT scans, please refer to the MICCAI 2020 RibFrac Challenge on grand-challenge.org (click Join first).

For annotations, please download RibSeg dataset to ./data/ribseg directory on Zenodo.

run data_prepare.py to create data for training.

Based on RibFrac dataset and RibSeg dataset, we binarized the CT scans and the annotations for rib segmentation to ./data/pn for the convenience of training PointNet++.

You can train your model through the following command line:

python train_ribseg.py --model pointnet2_part_seg_msg --log_dir <model_directory>

You can test your model through the following command line:

python test_ribseg.py --log_dir <model_directory>

You can conduct inference through the following command line:

python inference.py --log_dir <model_directory>

You can run our model through the following command line:

python inference.py --log_dir c2_a

You can get the volume-wise test result through the following command line:

python post_proc.py

If you find this project useful, please cite our paper as:

Jiancheng Yang, Shixuan Gu, Donglai Wei, Hanspeter Pfister, Bingbing Ni. "RibSeg Dataset and Strong Point Cloud Baselines for Rib Segmentation from CT Scans". International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI), 2021.

or using bibtex:

@inproceedings{yang2021ribseg,
  title={RibSeg Dataset and Strong Point Cloud Baselines for Rib Segmentation from CT Scans},
  author={Yang, Jiancheng and Gu, Shixuan and Wei, Donglai and Pfister, Hanspeter and Ni, Bingbing},
  booktitle={International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI)},
  pages={611--621},
  year={2021},
  organization={Springer}
}