Institute of Biology, Otto-von-Guericke University, Leipziger Str. 44/Haus 91, 39120 Magdeburg, Germany; Center for Behavioral Brain Sciences, Leipziger Str. 44, 39120 Magdeburg, Germany.
Institute of Biology, Otto-von-Guericke University, Leipziger Str. 44/Haus 91, 39120 Magdeburg, Germany; Center for Behavioral Brain Sciences, Leipziger Str. 44, 39120 Magdeburg, Germany.
J Neurosci Methods. 2018 Jan 15;294:15-33. doi: 10.1016/j.jneumeth.2017.10.020. Epub 2017 Nov 1.
Even in resting state, the human brain generates functional signals (fMRI) with complex correlational structure. To simplify this structure, it is common to parcellate a standard brain into coarse chunks. Finer parcellations are considered less reproducible and informative, due to anatomical and functional variability of individual brains.
Grouping signals with similar local correlation profiles, restricted to each anatomical region (Tzourio-Mazoyer et al., 2002), we divide a standard brain into 758 'functional clusters' averaging 1.7cm gray matter volume ('MD758' parcellation). We compare 758 'spatial clusters' of similar size ('S758').
'Functional clusters' are spatially contiguous and cluster quality (integration and segregation of temporal variance) is far superior to 'spatial clusters', comparable to multi-modal parcellations of half the resolution (Craddock et al., 2012; Glasser et al., 2016). Moreover, 'functional clusters' capture many long-range functional correlations, with O(10) reproducibly correlated cluster pairs in different anatomical regions. The pattern of functional correlations closely mirrors long-range anatomical connectivity established by fibre tracking.
MD758 is comparable to coarser parcellations (Craddock et al., 2012; Glasser et al., 2016) in terms of cluster quality, correlational structure (54% relative mutual entropy vs 60% and 61%), and sparseness (35% significant pairwise correlations vs 36% and 44%).
We describe and evaluate a simple path to finer functional parcellations of the human brain. Detailed correlational structure is surprisingly consistent between individuals, opening new possibilities for comparing functional correlations between cognitive conditions, states of health, or pharmacological interventions.
即使在静息状态下，人类大脑也会产生具有复杂相关结构的功能信号（fMRI）。为了简化这种结构，通常将标准大脑分割成粗块。由于个体大脑的解剖结构和功能的可变性，更精细的分割被认为不太具有可重复性和信息量。
我们通过对每个解剖区域（Tzourio-Mazoyer 等人，2002）进行局部相关谱相似的信号分组，将标准大脑划分为 758 个“功能簇”，每个簇平均含有 1.7cm³的灰质体积（“MD758”分割）。我们比较了具有相似大小的 758 个“空间簇”（“S758”）。
“功能簇”在空间上是连续的，簇质量（时间方差的集成和分离）远优于“空间簇”，与分辨率减半的多模态分割相当（Craddock 等人，2012 年；Glasser 等人，2016 年）。此外，“功能簇”还捕获了许多长程功能相关性，在不同的解剖区域中，有 O(10)对可重复相关的簇对。功能相关性的模式与纤维追踪建立的长程解剖连接密切匹配。
在簇质量、相关性结构（相对互信息的 54%对 60%和 61%）和稀疏度（35%的显著成对相关性对 36%和 44%）方面，MD758 与更粗糙的分割（Craddock 等人，2012 年；Glasser 等人，2016 年）相当。
我们描述并评估了一种更精细的人类大脑功能分割的简单方法。个体之间的详细相关性结构惊人地一致，为比较认知条件、健康状态或药物干预下的功能相关性开辟了新的可能性。
