There has been increasing evidence for atypical white matter (WM) microstructure in autistic people, but findings have been divergent. The development of autistic people in early childhood is clouded by the concurrently rapid brain growth, which might lead to the inconsistent findings of atypical WM microstructure in autism. Here, we aimed to reveal the developmental nature of autistic children and delineate atypical WM microstructure throughout early childhood while taking developmental considerations into account. In this study, diffusion tensor imaging was acquired from two independent cohorts, containing 91 autistic children and 100 typically developing children (TDC), aged 4–7 years. Developmental prediction modeling using support vector regression based on TDC participants was conducted to estimate the WM atypical development index of autistic children. Then, subgroups of autistic children were identified by using the k-means clustering method and were compared to each other on the basis of demographic information, WM atypical development index, and autistic trait by using two-sample t-test. Relationship of the WM atypical development index with age was estimated by using partial correlation. Furthermore, we performed threshold-free cluster enhancement-based two-sample t-test for the group comparison in WM microstructures of each subgroup of autistic children with the rematched subsets of TDC. We clustered autistic children into two subgroups according to WM atypical development index. The two subgroups exhibited distinct developmental stages and age-dependent diversity. WM atypical development index was found negatively associated with age. Moreover, an inverse pattern of atypical WM microstructures and different clinical manifestations in the two stages, with subgroup 1 showing overgrowth with low level of autistic traits and subgroup 2 exhibiting delayed maturation with high level of autistic traits, were revealed. This study illustrated age-dependent heterogeneity in early childhood autistic children and delineated developmental stage-specific difference that ranged from an overgrowth pattern to a delayed pattern. Trial registration This study has been registered at ClinicalTrials.gov (Identifier: NCT02807766) on June 21, 2016 ( https://clinicaltrials.gov/ct2/show/NCT02807766 ).

中文翻译：

---

基于扩散张量成像的发育预测模型揭示了幼儿自闭症大脑中年龄依赖性的异质性

越来越多的证据表明自闭症患者的白质（WM）微观结构不典型，但研究结果存在分歧。自闭症患者在儿童早期的发育因大脑同时快速生长而受到影响，这可能导致自闭症患者非典型 WM 微观结构的结果不一致。在这里，我们的目的是揭示自闭症儿童的发育本质，并在考虑发育因素的同时描绘整个幼儿期的非典型 WM 微观结构。在这项研究中，弥散张量成像是从两个独立队列中获取的，其中包括 91 名自闭症儿童和 100 名年龄为 4-7 岁的正常发育儿童 (TDC)。采用基于 TDC 参与者的支持向量回归进行发育预测建模，以估计自闭症儿童的 WM 非典型发育指数。然后，使用k均值聚类方法识别自闭症儿童的亚组，并使用双样本t检验根据人口统计信息、WM非典型发展指数和自闭症特征进行相互比较。通过使用偏相关估计WM非典型发育指数与年龄的关系。此外，我们进行了基于无阈值聚类增强的双样本 t 检验，对自闭症儿童每个亚组的 WM 微观结构与 TDC 重新匹配子集进行了组间比较。我们根据 WM 非典型发展指数将自闭症儿童分为两个亚组。这两个亚群表现出不同的发育阶段和年龄依赖性的多样性。研究发现 WM 非典型发育指数与年龄呈负相关。此外，非典型 WM 微观结构的相反模式和两个阶段的不同临床表现也被揭示，亚组 1 表现出过度生长，具有低水平的自闭症特征，亚组 2 表现出成熟延迟，具有高水平的自闭症特征。这项研究说明了幼儿自闭症儿童的年龄依赖性异质性，并描绘了从过度生长模式到延迟模式的特定发育阶段差异。试验注册 本研究已于 2016 年 6 月 21 日在 ClinicalTrials.gov 上注册（标识符：NCT02807766）（https://clinicaltrials.gov/ct2/show/NCT02807766）。