2026-05-06T12:18:07Z https://api.figshare.com/v2/oai
oai:figshare.com:article/30640043 2025-11-21T12:53:05Z category_28864 category_29035 category_24553 category_27082 category_28879 portal_549 item_type_3 month_year_11_2025
10.5522/04/30640043.v1 https://figshare.com/articles/dataset/C3VD-Raycasting-10k_A_Clinical_Point_Cloud_Registration_Dataset_for_Image-Guided_Colonoscopy/30640043 https://ndownloader.figshare.com/files/59624549 https://doi.org/10.48550/arXiv.2511.00260 Jiang, Linzhe Linzhe Jiang Huang, Jiayuan Jiayuan Huang Bano, Sophia Sophia Bano Clarkson, Matt Matt Clarkson Mao, Zhehua Zhehua Mao Hoque, Mobarak Mobarak Hoque 0000-0002-7162-2822 <![CDATA[C3VD-Raycasting-10k: A Clinical Point Cloud Registration Dataset for Image-Guided Colonoscopy]]> Artificial intelligence not elsewhere classified Graphics, augmented reality and games not elsewhere classified Surgery Digital health Computer vision Point cloud registration Surgical navigation State Space Models (SSM) Deep Learning Colonoscopy 2025-11-21 2025-11-21 Dataset 2025 University College London C3VD-Raycasting-10k is a clinically grounded benchmark dataset for 3D point cloud registration in image-guided colonoscopy. It contains 10,014 geometrically aligned point-cloud pairs that simulate the cross-modal alignment problem between preoperative CT anatomy and intraoperative endoscopic observations.

The dataset is derived from clinical CT and endoscopy data provided by the Colonoscopy 3D Video Dataset (C3VD) [Bobrow et al., MedIA 2023]. Starting from complete CT-based colon meshes and recorded endoscope trajectories, we use physics-based ray casting to generate realistic intraoperative viewpoints. For each recorded camera pose, we cast rays from the endoscopic viewpoint onto the CT-derived surface to obtain a partial target point cloud that mimics what is observable during colonoscopy. The corresponding source point cloud is sampled from the dense CT mesh representing the underlying preoperative anatomy.

Each sample in C3VD-Raycasting-10k therefore consists of:

  • A dense source point cloud derived from the preoperative CT colon mesh.
  • A partial target point cloud generated by ray casting from an endoscopic viewpoint, with occlusions and visibility constraints that reflect realistic intraoperative conditions.

By construction, the dataset emphasizes challenging but clinically relevant cases, including:

  • Partial-to-partial alignment with varying field-of-view, coverage, and missing regions.
  • Locally homogeneous geometry and repetitive structures that cause feature degeneracy on tubular organ surfaces.
  • Cross-modal variability between CT-derived anatomy and endoscopic appearance, while still providing precise geometric ground truth.

C3VD-Raycasting-10k is designed to support rigorous and reproducible benchmarking of 3D registration algorithms for image-guided colonoscopy and related minimally invasive procedures.

Citing the Dataset

Cite [Linzhe:arXiv2025] whenever research making use of this dataset is reported in any academic publication or research report.

Declaration

This point cloud dataset is derived from the Colonoscopy 3D Video Dataset (C3VD) (https://durrlab.github.io/C3VD/).

Original data: Bobrow et al., "Colonoscopy 3D video dataset with paired depth from 2D-3D registration", Medical Image Analysis, 2023.

In accordance with the original C3VD dataset license, our derived point cloud dataset is also released under the CC BY-NC-SA 4.0 license and may only be used for non-commercial purposes.

]]>