2026-BoneRep-MLforBoneImaging

Repository accompanying the paper. The analysis in notebooks/MLforBoneImaging.ipynb generates the results/figures.

Associated paper

• Citation: David, M. A., Williams, K. G., Constantine, E. P., Matthias, J., Ferguson, V. L., & Adams, D. J. (2026). Demystifying Machine Learning Approaches in Digital Bone Imaging using MicroCT and HRpQCT. Bone Reports.
• DOI: https://doi.org/10.1016/j.bonr.2026.101911
• RRID: SCR_028186

Start here

• Open the notebook in Colab: https://colab.research.google.com/github/madavid-research/2026-BoneRep-MLforBoneImaging/blob/main/notebooks/MLforBoneImaging.ipynb
• Non-coder guide: docs/START_HERE.md
• FAQ: docs/FAQ.md
• Curated “final” reference table: data/References.csv
• Figure navigation: docs/FIGURE_MAP.md
• View interactive networks: docs/VIEW_NETWORKS.md (start with networks/index.html)
• Optional (advanced): local runs and environment details in docs/FOR_CODERS.md

Stable Colab link (for papers)

The link above always points to the latest main branch.

For a stable link that never changes, use a GitHub release tag or commit SHA in the URL, for example:

• https://colab.research.google.com/github/madavid-research/2026-BoneRep-MLforBoneImaging/blob/<TAG_OR_COMMIT>/notebooks/MLforBoneImaging.ipynb

Jump to figures (in the notebook)

• Figure 7C: Prediction — notebooks/MLforBoneImaging.ipynb
• Figure 7D: Classification — notebooks/MLforBoneImaging.ipynb
• Figure 8: Dimensionality reduction & clustering — notebooks/MLforBoneImaging.ipynb
• Figure 9B/9C/9D: Grad-CAM / SHAP / radar plots — notebooks/MLforBoneImaging.ipynb

See docs/FIGURE_MAP.md for the exact section headings.

Key outputs

• data/References.csv: curated “final” reference table used for downstream analysis/figures
• data/df_pubmed_searches_deduplicated.csv: merged PubMed exports (output of the merge section; can be imported into SciNetX)
• data/PubMedSearchesWords.csv: lookup table listing the PubMed search terms used

What’s in here

• notebooks/ – Jupyter notebook (notebooks/MLforBoneImaging.ipynb)
• pubmed_searches/ – exported PubMed search CSVs (inputs)
• data/ – derived tables (outputs)
• networks/ – interactive HTML network visualizations (outputs)
• images/ – static images used in the notebook/paper

More docs:

• docs/GLOSSARY.md
• docs/TROUBLESHOOTING.md

Viewing interactive networks

Interactive HTML network visualizations are in networks/. Start with networks/index.html, and for viewing tips (including Colab), see docs/VIEW_NETWORKS.md.

If you only want to view the networks (no coding):

1. On GitHub, click Code → Download ZIP
2. Unzip the download
3. Open networks/index.html in a browser

Citation and license

• Cite via CITATION.cff.
• License terms are in LICENSE.

If you need a plain-text citation for the paper, use:

• David, M. A., Williams, K. G., Constantine, E. P., Matthias, J., Ferguson, V. L., & Adams, D. J. (2026). Demystifying Machine Learning Approaches in Digital Bone Imaging using MicroCT and HRpQCT. Bone Reports. DOI: https://doi.org/10.1016/j.bonr.2026.101911. RRID: SCR_028186.

If you need a plain-text repository citation, use:

• David MA, Williams KG, Constantine EP, Matthias J, Ferguson VL, Adams DJ. 2026-BoneRep-MLforBoneImaging (software). GitHub: https://github.com/madavid-research/2026-BoneRep-MLforBoneImaging.

Software used

This work utilized the SciNetX software to generate the network visualizations used in this repository and the paper. If you are interested in the SciNetX software itself, see https://github.com/madavid-research/SciNetX for source code and license terms.