Model for improved correlation of BMD values between abdominal routine Dual energy CT data and DXA scans

Osteoporosis is a common but underdiagnosed and undertreated disease causing severe morbidity and economic burden. The gold standard for detection of osteoporosis is DXA (dual energy x-ray absorptiometry), which is a dedicated examination for osteoporosis. Dual energy CT (DECT) examinations are increasingly used in daily routine for a wide variety of diagnoses. In the present study, we wanted to examine whether vBMD (volume bone mass density) could be evaluated as a side product in non-contrast as well as contrast phases as well as to evaluate a correction model taking known shortcomings for DXA into account.

A total of 20 patients, i.e. 79 vertebrae (one excluded due to vertebral fracture), mean age 71 years (range 43-85) with a mean BMI (body mass index) of 26 (range 17-33) were examined with both abdominal/pelvic DECT as well as DXA. Furthermore, aortic calcium was measured as well as the presence of osteoarthritis of the spine (OAS) and osteoarthritis in facet joints (OAF) with a 5-grade scaling system.

A significant correlation was found between DXA BMD and vBMD from DECT with no contrast (WNC) (r = 0.424, p = 0.001), and with venous contrast (WVC) (r = 0.402, p < 0.001), but no significant correlation was found with arterial contrast (WAC). Using multivariate linear regression with DXA BMD as dependent, two models were created combining DECT WNC, aortic calciumscore (ACS), OAS and BMI yielding an R² = 0.616 (model 1) and replacement of WNC to WVC a R² = 0.612 (model 2). The Pearson correlation between DXA and predictive DXA BMD value of model 1 was r = 0.785 (p < 0.001) and model 2 r = 0.782 (p < 0.001).

There is a correlation between DXA BMD and DECT in non-contrast and venous contrast scans but not in arterial scans. The correlation is further improved by quantifying the degree of different confounding factors (osteoarthritis of the spine, body mass index and aortic calcium score) and taking these into account in an explanatory model. Future software solutions with DECT data as input data might be able to automatically measure the BMD in the trabecular bone as well as measuring the confounding factors automatically in order to obtain spinal DXA comparable BMD values.