Large-Scale Spine CT Volumes for Vertebra Segmentation

1,005 CT volumes contain over 11,100 labeled vertebrae from patients with different spinal conditions. The dataset, curated by alexanderdann, provides a large-scale benchmark for medical image analysis tasks. It addresses the previous lack of sizable annotated spine image collections for data-driven deep learning methods.

Use Cases

• Train deep learning models for vertebrae segmentation on 1,005 CT volumes with over 11,100 labeled vertebrae.
• Develop algorithms for 3D spine reconstruction from biplanar radiographs using the provided CT image data.
• Benchmark image super-resolution or enhancement techniques on spine CT volumes belonging to different spinal conditions.
• Conduct quantitative medical image analysis for clinical diagnosis and surgery planning based on segmented vertebrae.

Strengths

• Contains 1,005 CT volumes, a large-scale collection for medical imaging.
• Includes over 11,100 individually labeled vertebrae for segmentation tasks.
• Covers vertebrae belonging to different spinal conditions, adding clinical diversity.

Limitations

• Specific file formats, image resolutions, and detailed annotation protocols are not described.
• The geographic origin and demographic distribution of the patient data are unknown, which may limit generalizability.
• The dataset is curated from multiple sources, which could introduce inconsistencies in imaging protocols or labeling standards.

Provenance

Source

Curated from multiple sources by alexanderdann.

Collection Method

Introduced in a research paper (arXiv:2105.14711) as a benchmark dataset.

Freshness

Last updated on 2025-05-25.