Detailed imaging of cerebral glucose metabolism for dementia diagnosis including omission of arterial blood sampling and correction of partial volume effect

基于MAP的脑葡萄糖代谢成像估计算法比普通似然估计算法具有更好的性能。该算法避免了迭代计算和寻找最优解的过程，避免了对初值的依赖性和失收敛。对正常人和Alzheimer病患者的分析结果与ROI分析结果吻合较好，计算时间仅需30 min。同时，提出了一种基于独立成分分析的动脉血漏采算法，并将两种算法结合起来，在较短的测量时间内实现了无动脉采血的葡萄糖代谢成像的实用方案。在校正部分容积效应（PVE），一个算法使用的统计建模和形态学操作。该算法首先根据体素值将MRI图像分割成若干解剖结构，然后通过形态学运算将脑解剖结构的三维信息融合到图像中。最后，使用线性估计来补偿PVE，假设白色物质和灰质中的区域体素值是共同的。将PVE算法应用于外周苯二氮卓受体配体PK11195的PET数据，PVE得到了较好的校正。

MAP-based estimation algorithm could be archived which had better performance than ordinary likelihood approach for imaging of cerebral glucose metabolism. This algorithm could avoid an iterative computation and searching an optimal points, therefore it was free from dependency to an initial value and misconvergence. The results from normal and Alzheimer cases were coincided with ROI-based analysis, and the computation required only 30 minutes.Also, a new algorithm for omission arterial blood sampling based on independent component analysis was established, and using the both algorithm, the practical scheme for glucose metabolism imaging without arterial blood sampling was realized in a shorter measurement duration than usual.On a correction of partial volume effect (PVE), a algorithm using a statistical modeling and morphological operation was developed. The algorithm first segmented a MRI image to some anatomical structures based on voxel values, then a morphological operations were applied to incorporate three-dimensional structure of brain anatomy. Finally, PVE was compensated using a linear estimation assuming regional voxel values in white matter and gray matter were common. When the PVE algorithm was applied to PET data of PK11195, a ligand for peripheral benzodiazepine receptors, PVE could be corrected well.