The impact of obesity on the accuracy of DXA BMD for DXA-equivalent BMD estimation

As the radiomics technique using texture features in CT is adopted for accessing DXA-equivalent bone mineral density (BMD), this study aims to compare BMD by DXA and predicted BMD to investigate the impact of obesity and central obesity in general patients.

A total of 710 cases (621 patients) obtained from May 6, 2012, to June 30, 2021, were used in the study. We focused both their abdomen & pelvis CT's first lumbar vertebrae axial cuts to predict estimated BMD and bone mineral content (BMC). In each patient's CT, we extracted the largest trabecular region of the L1 vertebral body as a region of interest (ROI) using the gray-level co-occurrence matrices (GLCM) technique, and linear regression was applied to predict the indices. Cases were divided by central obesity/overall obesity and normal group by body mass index (BMI), waist circumference (WC), or index of central obesity (ICO) standard.

The coefficients were all above 0.73, respectively. P-values from ICO were over 0.05 when the measures were Hip BMD and Hip BMC. In contrast, those from ICO were 0.0131 and 0.0351 when the measures were L1 BMD and L1 BMC, respectively, which show a difference between the two groups.

The CT HU texture analysis method was an effective and economical method for measuring estimated BMD and BMC and evaluating the impact of obesity. We found that central obesity especially exerted an effect on the disturbance of the clinical BMD measurements since groups were significantly different under the ICO standard.