结合光学脑成像及机器学习分类算法对自闭症大脑活动特征的研究进展

自闭症的诊断一直以来是基于行为层面的观察。近年来脑成像研究表明，自闭症大脑存在结构及功能的改变，这些发现为基于成像的诊断提供了新的途径。准确的诊断要基于对大脑异常特征的准确描述，多个脑生理参量的成像可以为自闭症的大脑特征提供更全面的信息。在现有的无损脑功能成像中，光学脑成像具有较高的时间及可接受的空间分辨率，测量时对头部晃动相对不敏感，适用于对儿童，特别是自闭症儿童的研究。近红外光谱技术提供皮层的血氧代谢，而漫射相关谱技术测量皮层的血流。这两类参量可以提供互补的皮层血液动力学信息，结合更为高效的分类算法，有望取得对自闭症大脑皮层活动特征的更为准确的描述和区分。在广泛文献调研及结合自身研究成果的基础上，对利用光学脑成像研究自闭症的大脑活动特征；结合特征参量利用机器学习分类算法对自闭症的预测；以及利用多模态光学脑成像技术（即结合近红外光谱及漫射相关谱）研究自闭症的前景等问题做了系统的回顾与展望，希望为自闭症相关领域的科研人员提供参考和借鉴。

Research Progress in Characterizing Brain Activity of Autism Spectrum Disorder Patients with Optical Brain Imaging and Machine-Learning Classification Algorithms

Diagnosis for autism spectrum disorder (ASD) has always relied on behavioral observations. However, recent studies with brain imaging show significant alterations in brain structure and function associated with ASD, which opens a new avenue for imaging-based diagnosis. Accurate diagnosis depends on accurate characterization on the alterations. In this regard, multi-contrast imaging can provide more complete and complementary information for characterizing autistic brain. Among various brain imaging modalities, optical brain imaging has good temporal and acceptable spatial resolution. It is also less sensitive to head motion and thus very suitable for studies on children, in particular, children with ASD. Near-infrared spectroscopy (NIRS) probes cerebral blood oxygenation, while diffuse correlation spectroscopy (DCS) measures cerebral blood flow. These hemodynamic variables may provide complementary information on the cerebral hemodynamics. Therefore by using more efficient machine-learning classification algorithm, it is anticipated that more accurate characterization and classification on ASD can be achieved. Based on extensive literature search and our on-going study, we reviewed in this paper the progress on using optical brain imaging to investigate characteristics of autistic brain, and the classification on ASD with various machine-learning algorithms. The prospect of using multimodal optical imaging (combined NIRS with DCS) to study ASD is also discussed. This systematic review and outlook might be beneficial to scientists who are working in ASD-related field.