Prediction of osteoporosis from proximal femoral cortical bone thickness and Hounsfield unit value with clinical significance

Abstract

Background: Utilizing dual-energy x-ray absorptiometry (DXA) to assess bone mineral density (BMD) was not routine in many clinical scenarios, leading to missed diagnoses of osteoporosis. The objective of this study is to obtain effective parameters from hip computer tomography (CT) to screen patients with osteoporosis and predict their clinical outcomes.

Methods: A total of 375 patients with hip CT scans for intertrochanteric fracture were included. Among them, 56 patients possessed the data of both hip CT scans and DXA and were settled as a training group. The cortical bone thickness (CTh) and Hounsfield unit (HU) values were abstracted from 31 regions of interest (ROIs) of the proximal femur. In the training group, the correlations between these parameters and BMD were investigated, and their diagnostic efficiency of osteoporosis was assessed. Finally, 375 patients were divided into osteoporotic and nonosteoporotic groups based on the optimal cut-off values, and the clinical difference between subgroups was evaluated.

Results: The CTh value of ROI 21 and the HU value of ROI 14 were moderately correlated with the hip BMD [r = 0.475 and 0.445 (p < 0.001), respectively]. The best diagnostic effect could be obtained by defining osteoporosis as CTh value < 3.19 mm in ROI 21 or HU value < 424.97 HU in ROI 14, with accuracies of 0.821 and 0.883, sensitivities of 84% and 76%, and specificities of 71% and 87%, respectively. The clinical outcome of the nonosteoporotic group was better than that of the osteoporotic group regardless of the division criteria.

Conclusion: The CTh and HU values of specific cortex sites in the proximal femur were positively correlated with BMD of DXA at the hip. Thresholds for osteoporosis based on CTh and HU values could be utilized to screen osteoporosis and predict clinical outcomes.

Keywords: Hounsfield unit; bone mineral density; cortical bone thickness; osteoporosis; proximal femur.

Figure 1

Flow chart of the inclusion and exclusion process.

Figure 2

Area of exploration in proximal femoral region. The femoral neck was divided by axial S1–3 (sections at the subcephalic, middle, and bottom sites of the femoral neck, respectively). Similarly, the shaft region was divided by S4–8 (sections at 20 mm above the upper edge of the femoral lesser trochanter, the upper edge of the femoral lesser trochanter, the vertex of the femoral lesser trochanter, the lower edge of the femoral lesser trochanter, and the 20 mm below the vertex of the femoral lesser trochanter, respectively).

Figure 3

Thirty-one regions of interest were defined in the femoral neck and shaft regions. (A–C) Femoral neck region. Purple line was the longest diameter between the upper and lower walls of femoral neck. Green line was the longest diameter between the anterior and posterior walls of femoral neck. R1–12 were defined as the measurement points of S1–3. (D–H) Femoral shaft region. Red line was the longest diameter between the medial and lateral walls of the femoral shaft. Yellow line was the longest diameter between the anterior and posterior walls of the femoral shaft. R13–31 were defined as the measurement points of S4–8.

Figure 4

Statistical analysis results of the CTh value of the region of interest (ROI) 21 and HU value of the ROI 14. (A) Plot illustrating the correlation between the CTh value of ROI 21 and hip bone mineral density. (B) Diagnostic efficiency of the CTh value of ROI 21 for osteoporosis. (C) Plot illustrating the correlation between the HU value of ROI 14 and hip bone mineral density. (D) Diagnostic efficiency of the HU value of ROI 14 for osteoporosis. CTh, cortical bone thickness; HU, Hounsfield unit; ROI, region of interest.

Figure 5

Rater reliability analysis.