Purpose
To evaluate the diagnostic performance of Hounsfield unit histogram analysis (HUHA) of precontrast abdominal-pelvic CT scans for predicting osteoporosis.
Materials and methods
The study included 271 patients who had undergone dual X-ray absorptiometry (DXA) and abdominal-pelvic CT within 1 month. HUHA was measured using commercial 3D analysis software (Aquarius iNtuition v4.4.12Ⓡ, TeraRecon) and expressed as a percentage of seven HU range categories related to the ROI: A < 0, 0 ≤ B < 25, 25 ≤ C < 50, 50 ≤ D < 75, 75 ≤ E < 100, 100 ≤ F < 130, and 130 ≤ G. A coronal reformatted precontrast CT image containing the largest Ward’s triangle was selected and then the ROI was drawn over the femoral neck. Correlation (r) and ROC curve analyses were used to assess diagnostic performance in predicting osteoporosis using the femur T-score as the reference standard.
Results
When the femur T-score was used as the reference, the rs of HUHA-A and HUHA-G were 0.74 and -0.57, respectively. Other HUHA values had moderate to weak correlations (r = -0.33 to 0.27). The correlation of HUHA-A was significantly higher than that of HUHA-G (p = 0.03). The area under the curve (0.95) of HUHA-A differed significantly from that of HUHA-G (0.90; p < 0.01). A HUHA-A threshold ≥ 27.7% was shown to predict osteoporosis based on a sensitivity and specificity of 95.6% and 81.7%, respectively.
Conclusion
The HUHA-A value of the femoral neck is closely related to osteoporosis and may help predict osteoporosis.
Key Points
• HUHA correlated strongly with the DXA femur T-score (HUHA-A, r = 0.74).
• The diagnostic performance of HUHA for predicting osteoporosis (AUC = 0.95) was better than that of the average CT HU value (AUC = 0.91; p < 0.05).
• HUHA may help predict osteoporosis and enable semi-quantitative measurement of changes in bone mineral density.