Predicting 3D Bone Mineral Density Distribution Using Machine Learning

Strategic Insights -

Researchers at Ewha Womans University have developed a novel method for predicting the 3D distribution of bone mineral density (BMD) using artificial intelligence. By employing machine learning algorithms, the researchers aimed to bridge the gap between dual-energy X-ray absorptiometry (DXA) scans and computed tomography (CT) in terms of volumetric bone assessment. The study's findings suggest that a low-cost and low-radiation alternative for bone health evaluation is possible.

Key Takeaways:

  • The study used data from 34 participants, including aligned DXA and CT scans for the proximal femur, to train two machine learning models: Extreme Gradient Boosting (XGB) and Gradient-Enhanced Neural Network (GENN).
  • The GENN model outperformed XGB, achieving mean absolute percentage errors (MAPE) of 12.98 ± 1.70%, 13.28 ± 2.01%, and 9.63 ± 1.66% for minimum, maximum, and the number of nonzero pixel intensities, respectively.
  • The proposed GENN framework offers a method for predicting 3D BMD distributions from a 2D-DXA scan, rivaling CT-based assessments, while reducing costs and radiation exposure.
  • The study demonstrated the potential of machine learning in personalized bone health evaluation and early osteoporosis diagnosis.
  • The results of this research were published in the journal Osteoporosis and Sarcopenia, with the article "Gradient-enhanced neural network and extreme gradient boosting modeling for the prediction of the 3D bone mineral density distribution from 2D-DXA scans" (2025, 11(3): 98-106).

Statistics:

  • 34 participants were included in the study.
  • The average registration accuracy of the GENN model was 89% ± 4% (p < 0.01).
  • The GENN model achieved a mean absolute percentage error (MAPE) of 12.98% ± 1.70% for the minimum BMD value.
  • The GENN model achieved a MAPE of 9.63% ± 1.66% for the number of nonzero pixel intensities.

Sources:

  • Gradient-enhanced neural network and extreme gradient boosting modeling for the prediction of the 3D bone mineral density distribution from 2D-DXA scans, Osteoporosis and Sarcopenia, 2025,11(3):98-106.
  • DOI: 10.1016/j.afos.2025.09.002.