The automated measurement of CT number linearity using an ACR accreditation phantom

Abstract

Abstract We developed a software to automatically measure the linearity between the CT numbers and densities of objects using an ACR 464 CT phantom, and investigated the CT number linearity of 16 different CT scanners. The software included a segmentation-rotation method. After segmenting five objects within the phantom image, the software computed the mean CT number of each object and plotted a graph between the CT numbers and densities of the objects. Linear regression and coefficients of regression, R2, were automatically calculated. The software was used to investigate the CT number linearity of 16 CT scanners from Toshiba, Siemens, Hitachi, and GE installed at 16 hospitals in Indonesia. The linearity of the CT number obtained on most of the scanners showed a strong linear correlation (R 2 > 0.99) between the CT numbers and densities of the five phantom materials. Two scanners (Siemens Emotion 16) had the strongest linear correlation with R 2 = 0.999, and two Hitachi Eclos scanners had the weakest linear correlation with R 2 < 0.99.