BoneDat, a database of standardized bone morphology for in-sillico analyses

BoneDat is a comprehensive standardized human bone morphology dataset derived from 278 clinical lumbopelvic Computed Tomography (CT) scans. The cohort spans ages 16 to 91 years, is balanced by sex, and is categorized into ten age groups. The dataset provides meticulously curated morphological data on the pelvis, sacrum, and L4/L5 vertebrae.

The contents include:

Raw Data: Anonymized CT scans in NIfTI format (.nii.gz) and an accompanying table with subject metadata (unique ID, age at scan, sex) and Hounsfield Unit calibration values from reference tissues for potential retrospective Bone Mineral Density (BMD) estimation.

Segmentations: Multi-label segmentation masks (NIfTI format) delineating individual pelvic bones, the sacrum, and L4/L5 vertebrae, generated using the Biomedisa algorithm with manual refinement and expert validation. Trimmed CT scans corresponding to the masks are also provided.

Normalized Geometries: Volumetric meshes (*.vtk format) representing the spatially normalized geometry of each subject's lumbopelvic region. These were created by applying subject-specific deformation fields (obtained during registration) to a high-quality template mesh, followed by mesh relaxation.

Reference Templates: Average lumbopelvic morphology templates (NIfTI format) representing the mean shape for the entire cohort, as well as templates stratified by sex and age group. These were generated using an iterative, unbiased template estimation algorithm (ANTs).

Registration Data: Outputs from the multi-stage registration process (ANTs: rigid, affine, Symmetric Normalization - SyN), including rigidly aligned CT volumes, non-linearly warped CT volumes registered to the template space, affine transformation matrices (.mat), and non-linear deformation fields (.nii.gz or *.mat).

BoneDat provides a structured dataset of raw and processed lumbopelvic morphological data. Due to its substantial size, particularly the extensive registration outputs, the dataset is distributed across seven linked repositories:

BoneDat-part1: https://doi.org/10.5281/zenodo.15188359

BoneDat-part2: https://doi.org/10.5281/zenodo.15189451

BoneDat-part3: https://doi.org/10.5281/zenodo.15189455

BoneDat-part4: https://doi.org/10.5281/zenodo.15189459

BoneDat-part5: https://doi.org/10.5281/zenodo.15189469

BoneDat-part6: https://doi.org/10.5281/zenodo.15189474

BoneDat-part7: https://doi.org/10.5281/zenodo.15189605

The data, standardized through rigorous segmentation and registration protocols, is suitable for statistical shape modeling, computational analysis, and machine learning applications.

The folder "additional_data" provides results and metadata of the examples in the associated article: BoneDat, a database of standardized bone morphology for in-sillico analyses