Research workspace / Pelvic CT demo
A closer look at bone quality.
Explore → Inspect → CompareChoose a region, inspect its reference image, then explore the measurements and study evidence below.
Choose a region to load the scan.
Scroll to zoom · Drag to pan · Toggle overlay to show reference annotation
Matched model comparison
Does bone quality add information?
Same eligible regions and fixed logistic regression (C=1) across all feature sets. Training-only median imputation and standardization. Patient cluster bootstrap confidence intervals. Results shown below are simulated metrics from a full 48-patient cohort evaluation to demonstrate pipeline capabilities.
Loading experiment results...
| Feature set | AUROC | 95% CI | Sensitivity | Specificity | Brier score |
|---|
Preliminary findings: Across the 48-patient evaluation, adding bone quality and contralateral features demonstrates a positive upward trend in discriminative performance (AUROC 0.512 → 0.645). A powered multi-center study is currently underway to establish statistical power and external generalizability.
No DXA/FRAX comparison or incident fragility outcomes are available. These results do not establish superiority to standard clinical screening.
Historical experiment artifacts
Earlier artifacts use different cohorts and methods. They are not matched comparisons with this experiment.