Osteoporosis screening using areal bone mineral density estimation from diagnostic CT images.


RATIONALE AND OBJECTIVES A reliable and cost-effective method for osteoporosis screening is important in addressing the increase in osteoporotic fractures due to aging populations. Diagnostic computed tomographic (dCT) images may contain densitometric information useful for osteoporosis screening. The aim of this study was to investigate the relationship between areal bone mineral density (aBMD) and volumetric information on dCT imaging and its suitability for building an osteopenia screening system. The goal of this system is to estimate aBMD and predict bone disease condition on the basis of dCT images of the lumbar spine. MATERIALS AND METHODS Dual-energy x-ray absorptiometry (DXA) aBMD and computed tomographic (CT) images were obtained from 44 male patients (mean age, 60 years). An aBMD from CT images (aBMD(CT)) was computed from the CT volume using established relationships of Hounsfield units to bone density and used to estimate DXA-derived aBMD (aBMD(DxA)). Estimated aBMD(CT) was then applied to diagnose osteopenia of the lumbar spine using statistical methods. RESULTS For the estimation of aBMD(DxA) from aBMD(CT), the proposed approach yielded a high correlation factor of r = 0.852, with a root mean square error of 0.0884 g/cm(2). The correlation was strongest when every slice in the dCT volume and both trabecular and cortical bone components were used. The classifier achieved an overall classification accuracy of 80.1% and an area under the receiver-operating characteristic curve of 0.894. CONCLUSIONS This clinical study demonstrates that aBMD(DxA) can be determined from routine CT data. Estimated aBMD(DxA) can be extended to form a dCT imaging-based opportunistic screening system for the detection and management of osteopenia.

DOI: 10.1016/j.acra.2012.05.017


Citations per Year

1,350 Citations

Semantic Scholar estimates that this publication has received between 52 and 6,643 citations based on the available data.

See our FAQ for additional information.