随着学龄期儿童（6-12岁）认知能力的快速发展，其脑白质也呈现出稳定的发育进程。弥散张量成像（Diffusion tensor imaging, DTI）因其对被试和硬件较低的扫描要求，被广泛应用于儿童脑白质发育研究。在DTI数据分析中，空间标准化是进行被试间分析或将个体空间与标准空间映射的必要步骤。配准算法是其核心过程。相较于基于标量的配准，基于张量的配准已被证明具有更好的表现，且适用于其他弥散加权成像。然而，现有的弥散张量模板大多基于成人数据建立，并不一定适用于大脑持续发育的学龄儿童。因此，本研究基于大规模学龄期儿童弥散磁共振数据，利用DTI-TK和ANTs在标准空间中建立了的学龄儿童弥散张量模板(School-age children diffusion tensor template, SACT)，并在中国和高加索儿童的数据集中分别验证。

具体地，针对不同年龄段被试数目分布不一致的情况，为减少特定年龄段被试对构建模板的影响，本研究采取了随机抽样的策略，从每个年龄段（6-12岁，共6个年龄段）内随机抽取20名被试，即每次抽样后将得到120名被试。基于随机抽取的120名被试，首先利用DTI-TK构建120人组水平张量模板，然后利用与DTI数据对应的T1w数据，通过ANTs，将组水平的张量模板映射到标准空间（CHN-PD的T1w模板）。上述抽样和模板构建过程将重复100次，每次抽样后均进行模板构建。通过在12，000例抽样数据上的验证，本研究发现基于在不同抽样数据构建的张量模板之间具有极高的相似度。进而将基于100次抽样构建的模板进行平均得到最终的SACT模板。

进一步，本研究在中国和高加索儿童的数据集上分别验证了SACT模板的配准表现。相较于成人张量模板（IIT模板），基于SACT模板的配准表现出更高的组间一致性和更少的形变。同时，本研究以学龄儿童脑白质发育为例，探索了SACT模板对统计分析的影响。相较于IIT模板，尽管基于SACT模板的统计结果表现出相似的空间分布模式，但其统计值的分布呈现出与成人模板的显著差异，并直接影响到基于置换检验的显著性分析结果。因此，这提示在后续的学龄儿童研究中不应忽视配准模板的影响。目前，SACT模板已经公开发布，有望在儿童脑发育和临床问题的研究中起到积极作用。

With the rapid cognitive development of school-age children (6-12 years old), their white matter also shows a steady developmental progression. Diffusion tensor imaging (DTI) has been widely used in children's white matter development studies because of its low scanning requirements for subjects and hardware. In DTI data analysis, spatial normalization is a necessary step for inter-subject analysis or mapping individual space to standard space. The alignment algorithm is the core process. Compared to scalar-based alignment, tensor-based alignment has been shown to have better performance and is suitable for other diffusion-weighted imaging. However, most of the existing diffusion tensor templates are built based on adult data and are not necessarily applicable to school-aged children with continuously developing brains. Therefore, in this study, a School-age children diffusion tensor template (SACT) built in standard space using DTI-TK and ANTs based on large-scale school-age children diffusion MRI data was validated separately in Chinese and Caucasian children datasets.

Specifically, in order to reduce the influence of age-specific subjects on the construction of the template, a random sampling strategy was adopted to randomly select 20 subjects from each age group (6-12 years old, 6 age groups in total), i.e., 120 subjects would be obtained after each sampling. Based on the 120 randomly selected subjects, the 120-person group-level tensor template was first constructed using DTI-TK, and then the group-level tensor template was mapped to the standard space (T1w template of CHN-PD) by ANTs using T1w data corresponding to the DTI data. The above sampling and template construction process will be repeated 100 times, with template construction performed after each sampling. Through validation on 12,000 cases of sampling data, this study found very high similarity between the tensor templates constructed based on different sampling data. The final SACT template was obtained by averaging the templates constructed based on 100 samples.

Further, the alignment performance of the SACT template was verified on the Chinese and Caucasian children datasets, respectively. Compared with the adult tensor template (IIT template), the alignment based on the SACT template showed higher inter-group agreement and less deformation. Meanwhile, this study explored the impact of SACT template on statistical analysis using the example of white matter development in school-aged children. Compared to the IIT template, although the statistical results based on the SACT template showed similar spatial distribution patterns, the distribution of statistical values showed significant differences from the adult template and directly affected the results of the significance analysis based on the permutation test. Therefore, this suggests that the influence of the alignment template should not be ignored in subsequent studies with school-aged children. The SACT template is now publicly available and is expected to play a positive role in the study of brain development and clinical issues in children.