Evaluation of diagnostic accuracy in CT perfusion analysis in moyamoya disease

The purpose of the present study was to determine optimal threshold of vascular pixel elimination (VPE) for CT perfusion (CTP) and to assess diagnostic accuracy of CTP by comparing with xenon enhanced CT (XeCT) in moyamoya disease.

Twenty-three patients underwent XeCT and CTP. Cerebral blood flow (CBF) images were generated for XeCT and CTP using nine types of software. Region of interest (ROI) measurement was performed on XeCT-CBF and CTP-CBF. Linear regression analysis was performed between XeCT-CBF and CTP-CBF in all software, without and with VPE. The Pearson correlation coefficient was calculated, and an optimal threshold was determined based on maximum correlation coefficients. Correlation coefficients at various VPE thresholds including data of no-VPE were compared with each other. The maximum correlation coefficient at the optimal threshold was also compared.

Optimal thresholds varied among software types (0.8-2.2 and 7-14 ml/100 g in relative and absolute VPE, respectively). There were significant differences between correlation coefficients at a range of VPE thresholds compared to no-VPE in most software types. There were significant differences in maximum correlation coefficient at optimal threshold among various software types.

Optimal threshold of VPE for CTP could be determined and diagnostic accuracy of CTP varied among software types in moyamoya disease.