Quantitative cerebral blood flow measurement with dynamic perfusion CT using the vascular-pixel elimination method:...

Blood vessels are usually conspicuous on dynamic CT perfusion images. The presence of large vessels may lead to overestimation of the quantitative value of cerebral blood flow (CBF). We evaluated the efficacy of the vascular-pixel elimination (VPE) method in quantitative CT perfusion imaging, in comparison with positron emission tomography (PET).

Five healthy volunteers underwent CT perfusion and PET studies. A four-channel multi-detector row CT scanner was used. Dynamic cine scanning was performed after bolus injection of an intravenous contrast agent. CT-CBF was calculated by the central volume principle and deconvolution method. PET was performed after infusion of (15)O-labeled water. PET-CBF was calculated by using a nonlinear least squares method. Average CBF values of the whole section, gray matter, and white matter with both CT and PET were compared after image registration. The comparison was performed with and without VPE. In the VPE method, the vascular pixels were defined by the cerebral blood volume value of the pixel. The threshold of VPE was changed from 5 to 20 mL/100 g. Pixel-by-pixel correlation between CT-CBF and PET-CBF and linear regression analysis were also performed.

Without VPE, CT-CBF was overestimated in all subjects. As the VPE threshold decreased, CT-CBF decreased and the correlation coefficient increased. The best correlation was observed at a VPE threshold of 8 mL/100 g in four of the five subjects. Average CT-CBF values, without VPE, of the whole section, gray matter, and white matter were 59.01, 66.73, and 42.53 mL/100 g/min, respectively. With VPE (threshold, 8 mL/100 g), average CT-CBF values of the whole section, gray matter, and white matter were 45.56, 52.75, and 30.38, respectively. The corresponding PET-CBF values were 46.86, 50.89, and 38.20 mL/100 g/min, respectively.

Vascular pixels should be excluded from the calculation of CT-CBF to avoid overestimation of the CBF values. If vascular pixels are excluded, CBF calculation with CT perfusion imaging is considerably accurate.