@misc{oai:repo.qst.go.jp:00069406,
 author = {Kimura, Yuichi and Naganawa, Mika and et.al and 木村 裕一 and 長縄 美香},
 month = {Jul},
 note = {Introduction This study aims at evaluating the algorithm for k2 imaging using clustering-based kinetic approach [1] (CAKS) to 11C-verapamil (VER) which is a substrate of P-glycoprotein (P-gp) and is popular to measure P-gp functions. In CAKS,
voxel-based tTACs are clustered based on such quantity that is related to the kinetics of VER in the brain. The behavior of VER can be described with one-tissue-two-compartment model [2]. Thus, the ratio of R = Integral t alpha C(t) dt / Integral C(t) dt has a direct relationship with k2 where the integral interval is from 0 to the last frame [1]. tTAC in a voxels are clustered based on each R, and
then an averaged tTAC in a cluster is utilized for compartment model estimation. This averaging improves a noise level in tTAC; the number of voxels in a cluster is 1000 in an implemented CAKS. Also, a computational time is reduced because estimates are not performed for each voxel but for each cluster. Note that K1 can be computed for every voxels although k2 is obtained for a cluster.
Methods Five normal volunteers involved who were 57 to 76 yrs, and two women and three men. A set of dynamic scan was conducted using Headtome-V (Shimadzu Co., Kyoto, Japan) with arterial blood samplings and metabolite correction. ROIs were placed on the frontal cortex, cerebellum, choroid plexus and centrum semiovale as the white matter, and the averaged tTAC in the ROI was inputted to a usual nonlinear estimation algorithm to derive the estimates of K1 and k2 in ROI. Also CAKS was applied to the data for K1 and k2 imaging, and the ROI-averaged estimates in the ROI were compared with the ROI-based kinetic analysis.
Results and Conclusion The comparison is shown in Fig. 1 in which the ROI-based usual kinetic analysis and CAKS are assigned on x- and y-axis, respectively. The regressions were y=1.07x+0.00 (r2=0.99), y=1.06x+0.00 (r2=0.94) and y=1.02x-0.04 (r2 = 0.99) for K1, k2 and total distribution volume (VT ), respectively. Therefore, the estimates using CAKS are almost identical with a usual ROI-based kinetics. The typical parametric images of VER are demonstrated in Fig. 2. In K1, a structure between gray and white matters could be seen. The choroid plexsus had a higher K1 than other regions, but the k2 was homogeneous. The clustering scheme based on kinetics could suppress the bad noise statistics in voxel-based tTAC so that reliable parametric images could be expected. The computational time was less than 5 minutes for the image of 125-by- 125 voxels and 40 slices, therfore, CAKS is practical for the parametric imaging on P-gp., Neuroreceptor Mapping 2008},
 title = {Evaluation of k2 imaging algorithm with 11C-verapamil using clustering-based kinetic approach},
 year = {2008}
}