Mechanistic-Data Synergy for Predicting Fracture Load in Canine Bone with Appendicular Osteosarcoma.

XGBoost-based ML surrogates can rapidly predict fracture-initiation loads in canine appendicular osteosarcoma to support stabilization decisions.

Canine appendicular osteosarcoma significantly elevates pathological fracture risk, and while Stereotactic Body Radiation Therapy (SBRT) can improve quality of life in large-breed dogs, post-SBRT fracture rates remain high and prophylactic stabilization carries notable complications. Patient-specific Finite Element Analysis (FEA) can estimate fracture-initiation thresholds, but its computational demands limit routine clinical use. This study developed Machine Learning (ML) surrogate models trained on FEA simulation data derived from CT scans of 16 dogs to rapidly predict fracture loads under 25 loading scenarios, with and without intramedullary stabilization. Five ML models were evaluated; XGBoost achieved the best performance, with R≈0.90 and a normalized mean absolute error of ~8% on held-out cases. Key predictors included lesion burden and body weight, with anatomical parameters contributing secondary effects; age and sex were negligible predictors. Patient-level generalization was more consistent for humeral cases than tibial or radial cases, highlighting the need for larger, bone-type-balanced datasets. Retrospective clinical outcomes from six treated dogs provided descriptive context. Findings suggest intramedullary stabilization can delay fracture initiation, though effectiveness varies by patient. This proof-of-concept framework offers a scalable, data-driven tool for clinical decision support in fracture risk assessment, with accuracy expected to improve as larger outcome datasets become available.

This summary was distilled by AI and may occasionally misinterpret data. Confirm critical details with the primary literature before clinical application.