Liu Qingyuan, Wei Yongbin, Liu Dongxu, Qi Ting, Zhao Kun, Zhang Ya-Hong, Cui Long-Biao, Liu Yong, van den Heuvel Martijn P
Center for Artificial Intelligence in Medical Imaging, School of Artificial Intelligence, Beijing University of Posts and Telecommunications, Beijing, China.
Department of Psychiatry, Xi'an Gaoxin Hospital, Xi'an, China.
Commun Biol. 2025 Jun 19;8(1):943. doi: 10.1038/s42003-025-08341-z.
Incorporating summary statistics across neuroimaging studies is important for enhancing translatability but poses challenges to connectomic analyses due to diverse methodological pipelines and brain atlases. We present TACOS (Transform brAin COnnectomes across atlaSes), a novel tool that translates network-based statistics across different atlases without requiring individual raw data. TACOS employs linear models based on anatomical information from brain parcellations and white matter fibers. Testing across 17 atlases, we show TACOS-transformed t-statistics to correlate well to the ground truth for both structural (r = 0.32-0.95) and functional networks (r = 0.57-0.95) using HCP surrogate statistics. These correlations remain consistent when tested with independent data from populations of different ancestries. Furthermore, TACOS effectively harmonizes connectomic results across multi-site schizophrenia data cohorts (r = 0.57-0.94 and 0.75-0.95 for structural and functional networks, respectively). This tool enables cross-atlas transformations of network-based statistics, showing great potential for downstream applications that share and combine multi-site connectomic data.
整合神经影像学研究中的汇总统计数据对于提高可翻译性很重要，但由于方法流程和脑图谱的多样性，给连接组分析带来了挑战。我们提出了TACOS（跨图谱转换脑连接组），这是一种新颖的工具，无需个体原始数据即可跨不同图谱转换基于网络的统计数据。TACOS采用基于脑部分割和白质纤维的解剖学信息的线性模型。通过对17个图谱进行测试，我们发现使用HCP替代统计数据时，TACOS转换后的t统计量与结构网络（r = 0.32 - 0.95）和功能网络（r = 0.57 - 0.95）的真实情况具有良好的相关性。当使用来自不同祖先群体的独立数据进行测试时，这些相关性仍然保持一致。此外，TACOS有效地协调了多站点精神分裂症数据队列中的连接组结果（结构网络和功能网络的r分别为0.57 - 0.94和0.75 - 0.95）。该工具能够对基于网络的统计数据进行跨图谱转换，在共享和组合多站点连接组数据的下游应用中显示出巨大潜力。
