CT calibration for micro structural parameters on multi-site studies

FELIX THOMSEN; CLAUS-C. GLÜER; CLAUDIO DELRIEUX

Bad Staffelstein

Workshop; 21th International Bone Densitometry Workshop; 2017

In typical multi-site or post-hocstudies which use different scanners or scan protocols, the assessment of microstructural parameters can be highly biased when using standard densitycalibration. We present (1) a globalhistogram-based and (2) a local convolution-based normalization technique tocompensate systematic differences between sites or scan protocols. The globalmicro structural calibration (MSC) rescales the difference between the localdensity voxel and the BMD of the VOI; the local MSC computes the sum of thescaled high- and unscaled low-pass, and is based on a Gaussian convolution in3D. The density-calibration resemblesa fixed threshold in mg/cc, the global MSC a threshold in terms of BV/TV and thelocal MSC resembles a concurrently fixed threshold in mg/cc and BV/TV amonggroups. Repeated scans of twenty VOIswere generated. In PMMA embedded vertebral specimens were five times repeatedlyscanned under a HRQCT (360mAs) and standard QCT setting (140mAs), both with120kVp, 188x188x300µm³ resolution, involving a calibration phantom and two bodyphantoms for simulating in-vivo conditions. BMD, TMD, BV/TV, BS/BV, Tb.N,Tb.Th. and Tb.Sp (model independent and direct-secant-method*) were computed. Short-term- (SD) andlong-term-precision (SEE) were normalized with the range of the medians of theVOIs. SEE was obtained by comparing with related HRpQCT scans (XtremeCT,82x82x82µm³). P-values of SD and SEE were obtained from ANOVA and respectivelyregression analyses. Compared to density calibration, localMSC improved reproducibility (SD) of all parameters (p<0.01) except for BMD(p=0.97) and Tb.Th (p=0.13). The global MSC improved SD of TMD, BV/TV, Tb.N,Tb.Th, Tb.Sp and BS/BV (p<0.05). Significant changes of accuracy (SEE)were not observed with either of both micro-structural calibrations.The proposedpre-processing methods are applicable for all micro-structural parameters and mightbe in particular useful if relative few patients oppose a large number of scansettings or sites.