Brainnetome Center, Institute of Automation, Chinese Academy of Sciences, Beijing 100190, China; National Laboratory of Pattern Recognition, Institute of Automation, Chinese Academy of Sciences, Beijing 100190, China.
Department of Nuclear Medicine, the Chinese People's Liberation Army General Hospital, Beijing 100853, China.
Psychiatry Res. 2015 Aug 30;233(2):131-40. doi: 10.1016/j.pscychresns.2015.05.014. Epub 2015 May 30.
Structural magnetic resonance imaging (sMRI) is an established technique for measuring brain atrophy, and dynamic positron emission tomography with (11)C-Pittsburgh compound B ((11)C-PIB PET) has the potential to provide both perfusion and amyloid deposition information. It remains unclear, however, how to better combine perfusion, amyloid deposition and morphological information extracted from dynamic (11)C-PIB PET and sMRI with the goal of improving the diagnosis of Alzheimer's disease (AD) and mild cognitive impairment (MCI). We adopted a linear sparse support vector machine to build classifiers for distinguishing AD and MCI subjects from cognitively normal (CN) subjects based on different combinations of regional measures extracted from imaging data, including perfusion and amyloid deposition information extracted from early and late frames of (11)C-PIB separately, and gray matter volumetric information extracted from sMRI data. The experimental results demonstrated that the classifier built upon the combination of imaging measures extracted from early and late frames of (11)C-PIB as well as sMRI achieved the highest classification accuracy in both classification studies of AD (100%) and MCI (85%), indicating that multimodality information could aid in the diagnosis of AD and MCI.
结构磁共振成像（sMRI）是测量脑萎缩的一种成熟技术，而动态正电子发射断层扫描与（11）C-Pittsburgh 化合物 B（（11）C-PIB PET）结合具有提供灌注和淀粉样蛋白沉积信息的潜力。然而，如何更好地结合从动态（11）C-PIB PET 和 sMRI 提取的灌注、淀粉样蛋白沉积和形态信息，以提高阿尔茨海默病（AD）和轻度认知障碍（MCI）的诊断仍不清楚。我们采用线性稀疏支持向量机基于从成像数据中提取的区域测量的不同组合来构建分类器，以区分 AD 和 MCI 受试者与认知正常（CN）受试者，包括从（11）C-PIB 的早期和晚期帧分别提取的灌注和淀粉样蛋白沉积信息，以及从 sMRI 数据提取的灰质体积信息。实验结果表明，基于（11）C-PIB 的早期和晚期帧以及 sMRI 提取的成像测量组合构建的分类器在 AD（100%）和 MCI（85%）的分类研究中均达到了最高的分类准确性，表明多模态信息可以辅助 AD 和 MCI 的诊断。
