Predictive efficacy of phantom-less calibration of computed tomography for cage subsidence after anterior cervical discectomy and fusion
摘要
Cage subsidence is a common complication after anterior cervical discectomy and fusion (ACDF). However, the predictive accuracy of conventional Hounsfield units (HU) for subsidence varies across studies. We aimed to develop a simple, phantom-less calibrated method for measuring bone mineral density (BMD) using preoperative cervical computed tomography (CT) and compared its predictive efficacy for cage subsidence with HU values.
MethodsWe retrospectively reviewed ACDF procedures performed by the corresponding author at our institution. Preoperative cervical spine CT scans were used to measure HU values at the surgical level. Additionally, HU values for calibration tissues—including cortical bone (CB), blood (B), skeletal muscle (SM), and adipose tissue (A)—were also measured. A linear regression equation was established from the HU measurements of these tissues and their standardized density values. The HU value at the surgical level was substituted into this equation to calculate volumetric BMD (vBMD). The area under the curve (AUC) was then calculated to compare the predictive efficacy of HU values and vBMD measurements for subsidence.
ResultsEighty-one patients were included in the study. Both HU values and calibrated vBMD were found to be risk factors for cage subsidence. Compared to HU values (AUC 0.703), calibrated vBMD (AUC 0.738–0.802) showed a stronger correlation with subsidence and higher predictive efficacy. Among the vBMD measurements, the combination of skeletal muscle and adipose (SM + A) provided the highest predictive efficacy (AUC 0.802), with an optimal cut-off value of 197.5 mg/cm³.
ConclusionBoth HU values and phantom-less calibrated vBMD are reliable predictors of cage subsidence after ACDF. The phantom-less calibrated vBMD can improve predictive efficacy for cage subsidence compared to HU values. We propose using SM + A calibrated vBMD to predict cage subsidence.