Cerebral perfusion computerized tomography: influence of reference vessels, regions of interest and interobserver variability.

INTRODUCTION: There are still no standardized guidelines for perfusion computerized tomography (PCT) analysis.
METHODS: A total of 61 PCT studies were analyzed using either the anterior cerebral artery (ACA) or the middle cerebral artery (MCA) as the arterial reference, and the superior sagittal sinus (SSS) or the vein of Galen (VG) as the venous reference. The sizes of regions of interest (ROI) were investigated comparing PCT results obtained using a hemispheric ROI combined with vascular pixel elimination with those obtained using five smaller ROIs located over the cortex and basal ganglia. In addition, interobserver variations were explored using a standardized protocol.
RESULTS: MCA-based measurements of cerebral blood flow (CBF) and blood volume (CBV) were in accordance with those obtained with the ACA except in 16 patients with ischemic stroke, in whom CBF was overestimated by the ipsilateral MCA. Venous maximal intensity was significantly lower with the VG when compared with the SSS, resulting in overestimation of CBF and CBV. However, in 13.3% of patients the VG ROI yielded higher maximal intensities than the SSS ROI. There was no difference in PCT results between hemispheric ROI and averaged separate ROI when vascular pixel elimination was used. Finally, interobserver variations were as high as 11% for CBF and 12% for CBV.
CONCLUSION: The present results suggest that pathological rather than anatomical considerations should dictate the choice of the arterial ROI. For venous ROI, although SSS seems to be adequate in most instances, deep cerebral veins may occasionally generate higher maximal intensities and should therefore be selected. Importantly, significant user-dependency should be taken into account.