CADS: 22,022 CT Volumes with 167 Anatomical Segmentations

CADS is a dataset of 22,022 Computed Tomography (CT) volumes with complete annotations for 167 anatomical structures, spanning from head to knee. It was created by huggingface as a framework for whole-body anatomical segmentation in medical imaging.

Use Cases

• Train a segmentation model to identify 167 anatomical structures from head to knee in CT volumes.
• Develop a fully automated pipeline for segmenting diverse anatomical systems in whole-body CT scans.
• Benchmark the performance of computer vision models on the extensive annotation of 167 anatomical structures.

Strengths

• Contains 22,022 CT volumes, providing a substantial data foundation.
• Features annotations for 167 anatomical structures, covering a wide anatomical scope.
• Described as the most extensive whole-body CT dataset for segmentation.

Limitations

• Specific details on data provenance, patient demographics, and geographic origin are not provided.
• The quality and consistency of the automated annotations across all 22,022 volumes are not independently verified.
• The dataset's size in terms of storage and specific file formats is unknown.

Provenance

Source

huggingface

Collection Method

Described as a fully automated framework for segmentation; specific gathering method not detailed.

Freshness

Last updated on 2025-12-17.