Optimization on these three coordinates was performed using a downhill simplex algorithm in order to minimize the area of femoral neck that intersected this plane. This automated algorithm used the NN region defined above as the initial starting location of the plane. Since the algorithm started with the NN region as the initial Sorafenib concentration guess, and this region is between the femur head and greater trochanter, convergence to the plane with the narrowest area was rapid. FNAL was measured perpendicular to this plane through its center of mass from the edge of the femoral head to where

the axis exited the femur distally. To reduce the effects of osteophytes which were prevalent and visible in the QCT dataset, the measurement was repeated eight times along line segments parallel to the neck axis. The eight measurements were Temozolomide concentrically spaced around the neck axis. The final FNAL value was defined as the median of these eight parallel segments and the central measurement. Statistics Parameters calculated from the QCT dataset were considered the gold standard, and the parameters calculated by HSA were compared to QCT by linear regression analysis using GraphPad Prism V 5.03. If the offset (i.e.,

intercept) was not statistically different from zero (p < 0.05), the analysis was repeated with the intercept restricted to zero. In order to test the sensitivity of our results to the

placement of the NN ROI, in addition, the plane through the narrowest part of the femoral neck of the QCT dataset was also used as the basis for an alternate definition of the QCT NN ROI and compared to the HSA NN ROI. Results High linear correlations (r = 0.89–0.95) were found between HSA and QCT for CSA, CSMI, and Z at the NN and IT regions (Figs. 2 and 3). The intercepts of the linear correlation mafosfamide of the parameters were not statistically significant (p < 0.05) at the IT region but were statistically significant at the NN region (Table 1). The slopes of these parameters were all different from unity. Fig. 2 The correlation of HSA with QCT for the narrow neck region Fig. 3 The correlation of HSA with QCT for the trochanter region Table 1 Results of the linear correlation of HSA vs. QCT at the NN and IT regions   NN IT Cross-sectional area (cm2) r 0.95 0.93 Offset 0.32 (0.11) N.S. Slope 2.02 (0.10) 2.00 (0.02) SEE 0.13 0.31 Cross-sectional moment of inertia (cm4) r 0.94 0.93 Offset 0.30 (0.12) N.S. Slope 1.19 (0.06) 1.48 (0.03) SEE 0.22 1.40 Section modulus (cm3) r 0.93 0.89 Offset 0.19 (0.07) N.S. Slope 1.32 (0.08) 1.53 (0.03) SEE 0.10 0.50 Width (cm) r 0.95 0.95 Offset N.S. N.S. Slope 0.979 (0.004) 0.978 (0.003) SEE 0.08 0.10 Femoral neck axis length (cm) r 0.90 – Offset N.S. – Slope 1.003 (0.004) – SEE 0.22 – Numbers in parentheses are standard errors. N.S.