Volumetric dxa (vxa): a new method to extract 3d information from multiple in vivo dxa images

摘要: Three-dimensional geometric and structural measurements of the proximal femur are considerable interest in understanding strength its susceptibility to fracture. Quantitative computed tomography (QCT) with a small voxel size (1mm per side) is current "gold standard" examine macrostructure femur, but it has high effective radiation dose (approximately 2 5 mSv) cost. Volumetric dual-energy X-ray absorptiometry (VXA) uses commercially available DXA system (Hologic Discovery A) reconstruct from four scans statistical atlas, delivering an 0.04 mSv. We present theoretical innovations systems level construction order create VXA volume. First, we generate successful deformations on highly irregular partial anatomy where typical algorithms fail, cover novel mathematical feature-correspondence transformation developed called HPCA, or hierarchical principal component analysis. We then evaluation shape atlas focus generating model correspondences meticulously refined avoid parameterizations which biased irrelevant qualities femur. also rigorously evaluate HPCA algorithm by ability can estimate unlearned shapes, compare another generated competing method. Finally move beyond simply creating volume driving parameterization algebraically use information within images themselves further impose energy constraints narrow neck slice region derive more accurate cortical thickness. Our validated through extensive leave-1-out validations. We QCT 43 elderly women at slices located femoral trochanteric regions interest. For parameters axis length (FNAL), cross-sectional area (SA), bone (CSA), volumetric mineral density (vBMD), engineering minimum, maximum, polar moments inertia (CSMIs) section modulus (Z), demonstrated strong correlations. conclude that capable variety 3D correlated subjects vivo. Moreover, be made device very low dose.