LumbarSR is a paired clinical CT and photon-counting micro-CT dataset for human lumbar vertebrae, designed to support super-resolution, cross-resolution comparison, and trabecular bone analysis in musculoskeletal imaging. The dataset contains 30 paired human dry lumbar vertebral specimens scanned with both clinical helical CT and Micro-PCCT. Clinical CT data were acquired under eight configurations formed by the combination of two in-plane resolutions (195 um and 586 um), two slice thicknesses (500 um and 1000 um), and two reconstruction kernels (bone and soft tissue). The corresponding Micro-PCCT reference volumes are provided at 105 um isotropic resolution. This release includes:- original DICOM data for advanced users who wish to process the data from scratch- registered NIfTI volumes aligned to the Micro-PCCT reference space- released BoneMask ROI volumes aligned to the registered data- dataset documentation and public benchmark resources through the associated repository and project website The released data support research on CT super-resolution, multi-resolution fusion, kernel-aware reconstruction, and downstream trabecular bone morphometry. The public repository also provides registration baselines, super-resolution baselines, evaluation scripts, and benchmark documentation. Recommended resources:- GitHub repository: https://github.com/FrankZhangRp/LumbarSR-Challenge - Project website: https://frankzhangrp.github.io/LumbarSR-Challenge/ Archive contents:- `lumbarSR.zip` together with split parts `lumbarSR.z01`-`lumbarSR.z06` contains the main dataset archive- `BoneMask.zip` contains the released BoneMask ROI package For training and development, the public documentation currently recommends using `Lumbar_01`-`Lumbar_25` for development and `Lumbar_26`-`Lumbar_30` as the released reference test subset. If you use this dataset, please also cite the associated repository and project website listed above.