Published October 12, 2026 | Version v0.1.7

SpineLocus: YOLO-Pose and HRNet for thoracic DRR landmarks

Authors/Creators

Description

SpineLocus is an independently developed, open-source Python research framework by Luca Serioli for the localization of 11 thoracic vertebral landmarks (T1–T11) in Digitally Reconstructed Radiographs (DRRs).

The framework combines YOLO-Pose for global landmark detection with optional HRNet-based local refinement. It provides coordinate ensembles, test-time augmentation, anatomical identity checks, subject-disjoint dataset splitting, training and evaluation workflows, and reproducibility tools for medical image analysis research.

The repository includes source code, configuration files, documentation and automated tests using synthetic data. Medical images, annotations, individual patient records and trained model weights are not distributed.

Reported performance is based on development validation and does not constitute independent clinical validation. SpineLocus is intended exclusively for research and must not be used for clinical decision-making.

Notes (English)

If you use SpineLocus in your research, please cite this archived software release using its version-specific DOI: https://doi.org/10.5281/zenodo.23325290

Source code and documentation: https://github.com/LucaXTech/spinelocus

Files

LucaXTech/spinelocus-v0.1.7.zip

Files (1.5 MB)

Name Size
md5:f594d0e442e4d429b1508a6471f3d80f
1.5 MB Preview Download

Additional details

Related works

Software

Repository URL
https://github.com/LucaXTech/spinelocus
Programming language
Python
Development Status
Active