基于定量磁化率成像人脑深层灰质核团概率性图谱的构建

Construction of the probability map of the human deep brain nuclei by quantitative susceptibility mapping

目的基于定量磁化率成像(quantitative susceptibility mapping,QSM)技术制作可用于自动分割大脑深层灰质核团的概率图谱。材料与方法 15名健康受试者参与研究,所有受试者扫描均在3.0 T磁共振成像设备系统上完成。在随机选取的10名受试者得到的标准空间QSM图上,手动勾画出六个双侧脑深部灰质核团,之后采用相应的图谱评价方法选择最优概率阈值的图谱作为最终的概率图谱。在其余5名受试者得到的标准空间QSM图上,分别使用三种图谱(概率图谱、AAL图谱和Johns Hopkins图谱)自动分割和由2名研究者手动勾画出六个双侧脑深部灰质核团感兴趣区,并分别计算自动分割与手动勾画得到的区域的相似度Dice系数和磁化率值,以评价概率图谱的准确性。结果在基底节区域,概率图谱分割结果的Dice系数明显高于AAL图谱,但和Johns Hopkins图谱区别不大;在颅底和小脑区域,概率图谱分割结果的Dice系数明显高于Johns Hopkins图谱。与其他两种图谱相比,概率图谱自动分割深部核团后测量得到的磁化率值,更接近于手动勾画核团测量得到的磁化率值,其差别更小。结论基于多名受试者QSM图像构建的脑深部灰质核团概率图谱,对大脑灰质核团分割效果更加可靠,可有效提高图像分析工作的效率。

Objective： Based on quantitative susceptibility mapping （QSM） technique, an auto-segment probabilistic atlas for the gray matter nuclei in deep brain was established in the present study. Materials and Methods： The QSM data from 15 healthy subjects were acquired on a clinical 3.0 T MRI scanner with a 12 channel matrix head coil. Ten subjects were randomly selected to create a gray matter nuclei atlas of the deep brain, and the remained five subjects were used to evaluate the effectiveness of the atlas. Specifically, the regions of interest （ROI） in six bilateral structures drawn manually by two raters were used as the gold standard, meanwhile, these corresponding ROIs were automatically segmented by three kinds of atlas. To assess the accuracy of proposed segment approach, the probabilistic atlas was compared with both AAL and Johns Hopkins atlas by calculating the Dice coefficient and the susceptibility values in the auto-segment and manual-segment ROls, respectively. Results： The Dice coefficient in our probability atlas was significantly higher than the AAL in the basal ganglia region and the Johns Hopkins atlas in the skull base and cerebellum, respectively. Moreover, the susceptibility values in our probability atlas were more closer to that of manual segment region compared with the other two atlases. Conclusions： The probability atlas based on the QSM images is more reliable than both AAL and Johns Hopkins atlas in the segment of gray matter nuclei of deep brain. This atlas may be effective to improve the efficiency of image analysis in the clinical research.