Radiomics and stacking regression model for measuring bone mineral density using abdominal computed tomography.

Area of Science:

• Radiology
• Medical Imaging
• Osteoporosis Research

Background:

• Bone mineral density (BMD) measurement is crucial for diagnosing osteoporosis.
• Current BMD assessments are often performed post-fracture, limiting early intervention.

Purpose of the Study:

• To investigate the potential of a radiomic model using abdominal computed tomography (CT) for predicting lumbar vertebrae BMD.
• To explore non-invasive methods for BMD assessment.

Main Methods:

• A retrospective study included 245 patients with both DXA and abdominal CT scans.
• 1218 image features were extracted from CT scans; key features were selected using LASSO regression.
• A two-tier stacking regression model integrated linear and non-linear approaches for BMD prediction.

Main Results:

• The stacking regression model achieved high performance (Adjusted-R²=0.830, RMSE=0.077, MAE=0.06).
• Predicted BMD strongly correlated with DXA results (r=0.932).
• Bland-Altman analysis confirmed good agreement between predicted and DXA-measured BMD.

Conclusions:

• Radiomics analysis of abdominal CT images can effectively predict lumbar vertebrae BMD.
• This approach may enable earlier detection of osteoporosis risk.