Artificial Intelligence Predicts Fracture Toughness in Human Bones

Strategic Insights -

Researchers at Abdullah Gul University have made significant progress in using artificial intelligence and machine learning to predict fracture toughness in human bones. The study used Raman spectroscopy and machine learning algorithms to analyze ex vivo human femoral cortical bone specimens, providing a new approach to bone research. The findings highlight the potential of AI and ML in advancing bone research, with potential applications in understanding and preventing bone fractures.

Key Takeaways:

  • The study used Raman spectroscopy to analyze ex vivo human femoral cortical bone specimens, feeding spectral features and demographic variables into support vector regression, extreme tree regression, extreme gradient boosting, and ensemble models to predict fracture-toughness metrics.
  • Ensemble models consistently outperformed individual models, with the best performance for crack initiation toughness prediction being achieved using the ensemble approach, yielding a coefficient of determination (R2) of 0.623 and root-mean squared error (RMSE) of 1.320.
  • The XGB model achieved an R2 of 0.737, RMSE of 2.634, MAE of 2.283, and MAPE of 0.240 for prediction of the overall energy to propagate a crack (J-integral).
  • The researchers used demographic variables (age, sex) and structural parameters (cortical porosity, volumetric bone mineral density) to improve prediction accuracy.
  • The study highlights the importance of incorporating mineral quality properties (MP) and organic matrix properties (OMP) for enhanced prediction accuracy.
  • The research has potential applications in understanding and preventing bone fractures, and the authors suggest future studies could focus on larger datasets and more advanced modeling techniques.

Statistics:

  • 118 donors participated in the study, with an age range of 21-101 years.
  • The study used n1Phosphate (PO4)/CH2-wag, n1PO4/Amide I, and other Raman-derived mineral and organic matrix parameters for feature selection.
  • The best performance for crack initiation toughness prediction was achieved with an R2 of 0.623 and RMSE of 1.320.
  • The XGB model achieved an R2 of 0.737, RMSE of 2.634, MAE of 2.283, and MAPE of 0.240 for prediction of the overall energy to propagate a crack (J-integral).

Sources:

  • Prediction of biomechanical properties of ex vivo human femoral cortical bone using Raman spectroscopy and machine learning algorithms. Bone Reports, 2025,26():101870. (Bone Reports - http://www.journals.elsevier.com/bone-reports/)
  • NewsRx. Abdullah Gul University Researchers Report Recent Findings in Machine Learning (Prediction of biomechanical properties of ex vivo human femoral cortical bone using Raman spectroscopy and machine learning algorithms). Medical Imaging Law Weekly. September 16, 2025; p 22.