同卵双生子可以最大限度地控制潜在的混杂因素，迄今为止国际上仅见3份孤独症谱系障碍（ASD）双生子脑影像学对照研究的报道。本项目拟在已有ASD临床资源库和脑结构影像学研究的基础上，对60对中国汉族3~10岁同卵双生ASD患儿进行系列研究。首先利用sMRI以及DTI数据对比分析同卵双生共患ASD组、同卵双生非共患ASD组及正常对照组在全脑白质结构网络属性、白质纤维骨架以及重要白质纤维束的特征；然后利用rs-fMRI成像探讨ASD患者的脑默认网络、边缘网络、额叶-纹状体回路等脑功能网络及回路的异常，并进一步验证脑功能网络与脑结构网络之间的关联，明确ASD特定临床表型和脑网络表型之间的内在联系，寻找ASD更为统一的影像学标志。研究还将应用代价敏感半监督支持向量机等机器学习的方法建立ASD诊断分类模型，以数据驱动的方式筛选具有诊断意义的ASD脑影像学标记以及可用于预测疾病严重程度的脑影像学标记。

The monozygotic twins research achieves the utmost control of potential confounding factors. To date there were only three case-control studies studies of ASD twins neuroimaging reported all over the world. Based on existing clinic database and studies on the brain structure of ASD, this study will conduct a serial research on sixty pairs of ASD monozygotic twins aged from 3 to 10. First the technique of sMRI, rs-fMRI and DTI will be combined to analyze the difference of whole-brain white matter network performance, white matter fiber skeleton, and features of important fiber bundles among the monozygotic discordant twins for ASD group, monozygotic twins concordant for ASD group and typical group. Then the rs-fMRI will be utilized to discuss the deficit of functional networks such as default network, limbic network, and the frontal-striatal loop et al, further prove the relationship between the brain functional and structural network, and clarify the inner relations between the specific clinical phenotype and brain network phenotype of ASD, in order to seek the uniform neuroimaging marks of ASD. The machine learning techniques such as Cost-Sensitive Laplacian Support Vector Machines also will be employed to generate the diagnostic model of ASD, to screen out the valuable neuroimaging marks for diagnosing and severity-predicting by the tools of Data-driven.