在原有2~5岁孤独症谱系障碍（ASD）脑结构影像随访研究基础上，本项目拟入组200例ASD高危同胞为ASD高危组并选取40例同胞正常的新生儿为ASD低危组，前瞻性地从出生随访至24个月确诊分组。通过临床评估与行为观察分析比较ASD高危组与ASD低危组以及最终发展为ASD的病例组行为学的差异。同时在被试12个月及24个月大时利用结构MRI和DTI成像比较ASD高危组与ASD低危组以及ASD组与非ASD组脑结构网络的差异；研究还将在国际上首次利用自然睡眠rs-fMRI探究ASD高危同胞的早期脑功能异常，比较两组被试全脑结构网络属性以及脑功能连接稳定性的差异，并从数据驱动和感兴趣区两个角度筛选具有疾病诊断意义的脑网络特征值。再利用加权深度学习（WDL-RF）、稀疏编码等机器学习的方法建立ASD早期诊断预测模型和疾病严重程度预测模型；最终筛选出适用于2岁以内的ASD高危人群的脑影像学标记。

Based on our previous brain structure imaging follow-up study of children 2 to 5 years old children with autism spectrum disorder (ASD) , this study plans to enroll 200 infants with ASD sibling as the ASD high-risk group and 40 infants with normal sibling as the ASD low-risk group. These two groups will be diagnosed according to the diagnostic criteria of ASD prospectively at 24 months old with follow-up tracking from birth to 24 month. The behavioral differences among the ASD high-risk group, low-risk group and confirmed ASD group were compared by clinical evaluation and behavioral observation analysis. At the same time, structural MRI and DTI imaging were used to compare the differences in brain structure networks between ASD high-risk group and low-risk group, as well as ASD group and non-ASD group at 12 months and 24 months. The study will also use the natural sleep rs-fMRI to explore the early brain dysfunction of ASD high-risk siblings for the first time in the world, and seek the differences in global brain structure network properties and global functional connectivity stability (GFCS) between the two groups. The brain network feature with diagnostic value are screened from both the data-driven and region of interest. Then use the machine learning methods such as Weighted Deep Learning and Random Forest (WDL-RF) and Sparse Coding to establish early prediction model and severity prediction model of ASD. Finally, the brain imaging markers of ASD high-risk populations before 2 years old were screened.