Predicting Body Composition from 3D Optical Images and Biomarkers

847 patient records from Pennington Biomedical Research Center enable prediction of Appendicular Lean Mass, Body Fat Percentage, and Bone Mineral Density using 3D optical imaging biomarkers. The work compares supervised models like Support Vector Regression to a semi-supervised p-Laplacian model for data-constrained environments. The dataset supports research into non-invasive alternatives to costly DXA scans.

Use Cases

• Predict Appendicular Lean Mass (ALM) based on 3D optical image biomarkers like height and volume.
• Estimate Body Fat Percentage (BFP) using features such as left calf circumference.
• Forecast Bone Mineral Density (BMD) as an alternative to DXA scans.
• Train semi-supervised learning models for regression in data-constrained medical settings.

Strengths

• Contains data from 847 patients, providing a foundation for model training.
• Specific model performance metrics are reported, such as ~8% error for ALM/BMD with SVR.
• Explicitly addresses the challenge of limited labeled data with a p-Laplacian model.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The dataset is small in scale at 110.6 KB, indicating limited scope.

Provenance

Source

Pennington Biomedical Research Center

Collection Method

Biomarkers obtained from 3D optical images.

Freshness

Last updated 2026-05-13 09:28:38; freshness should be verified.