Keywords:
Bones, Musculoskeletal bone, Neuroradiology brain, CT, Nuclear medicine conventional, Diagnostic procedure, Osteoporosis
Authors:
M. Macbain1, M. Kok2, G. Wiggins1, S. T. Lee1, E. Lau1, N. Kutaiba1; 1Heidelberg, VIC/AU, 2Box Hill, VIC/AU
DOI:
10.26044/ranzcr2021/R-0131
Methods and materials
Retrospective cohort study. 432 patients were identified who had undergone a non-contrast CT brain between January 2016 and December 2017 at The Austin Hospital, and underwent a DEXA bone density scan within the following 12 months. 177 patients who had undergone a bone density scan preceding their CT brain were excluded, so as not to bias the cohort to those with known osteoporosis and to exclude patients on treatment for osteoporosis.
The same process was undertaken in a second cohort who underwent non-contrast CT brain at Eastern Health between January 2017 and December 2018. The same exclusion criteria applied to this cohort. A final cohort size of 108 patients was obtained, with two having been excluded for hyperostosis frontalis interna).
Axial 5 mm CT brain slices were viewed on a bone window for the remaining patients. A region of interest tool was used by a radiology registrar and a resident doctor under the supervision of a consultant radiologist to measure the average hounsfield unit density of the right frontal bone inner table at the level of the superior most slice through the lateral ventricles, or within 5 slices towards the vertex (Fig. 1). The viewing radiology registrar and resident were blinded to the DEXA bone density results. Patients were excluded if they had hyperostosis frontalis interna, or if their inner table was so narrow as to not be reliably measured. 9 patients were excluded as their DEXA bone density results were not available, leaving a total of 327 patients for inclusion.
Analysis of variance (ANOVA) tests were used to compare mean skull density values (HU) according to DEXA categories. Mean values were accompanied with standard deviation (SD) and median values with interquartile ranges (IQR). Pearson’s correlation coefficient quantified the relationship between skull density and T-score according to gender. The sensitivity and specificity of skull density was assessed at specified thresholds using empirical receiver operating characteristic (ROC) curve analysis, with area under the curve (AUC) and 95% confidence interval (CI). DEXA category was assigned as the state variable and HU value as the test variable. PPV and NPV calculations were performed using MedCalc Software (Belgium). Remaining statistical calculations and graphics were performed using SPSS, version 26 (IBM SPSS Statistics, IBM Corp). A significance level of p < 0.05 was used in all analyses.