Quantification of the difference in hounsfield units of an electron density phantom between a conventional and standing computed tomography machine

In human CT imaging, opportunistic calculation of bone mineral density (BMD) is used to screen for osteoporosis, where BMD is estimated by Hounsfield Units (HU). There is potential to use HU for estimation of BMD in veterinary imaging. However, HU values can vary between different machines and over time. We undertook a serial measurements study to determine (1) the extent of the difference of the mean HU and noise (HU standard deviation) of eight regions of interest (ROI) of an electron density phantom scanned on two CT machines, a Siemens CT (conventional) machine and the Equina by Asto CT (standing) machine; and (2) to determine if the HU values obtained on each machine differed over time. The mean HU from 30 scans of an electron density phantom were generated from each of the two CT machines (N = 60; conventional and standing). Differences between machines were assessed using a linear mixed effects model. The mean HU of each positive value ROI from the conventional machine was decreased compared with the standing machine (P < 0.001) and the difference was greatest for the denser phantom inserts. A calibration curve equation utilising the four bone-equivalent insert plugs of the phantom, with known calcium hydroxyapatite (CaHA) concentration, was generated using linear regression to represent CaHA, and therefore BMD, for the conventional machine (CaHA = -29.58 + 0.98xROI_mean) and the standing machine (CaHA = -54.53 + 0.95xROI_mean). For opportunistic calculation of BMD from HU in horses, the differences in HU observed between different CT machines warrants calibration of HU using a bone equivalent phantom.