大脑前额叶多通道近红外光谱信号数据的分类鉴别

背景：精神分裂症主要是通过症候学的方法进行诊断，近年来通过神经影像技术与模式识别的结合对精神分裂症患者与正常人进行鉴别的研究已经引起人们的兴趣。目的：利用模式识别的方法对精神分裂症患者和正常人的大脑前额叶多通道近红外光谱信号数据进行分类鉴别，验证其可行性。方法：使用言语流畅性测验作为激活任务，采集精神分裂症患者和正常人的大脑前额叶的近红外光谱信号数据。对采集数据进行预处理后计算各通道均值作为特征，计算接收者操作特征的曲线下方面积对通道特征进行分类性能排序，使用支持向量机按性能排序的特征组合做分类，然后用留一验证法计算分类性能指标，验证分类能力。结果与结论：研究发现特征性能排序前8位的特征组合的准确度最高达到95.24%，并且这8个通道都位于右侧前额叶。推断右侧前额叶区域可能是影响精神分裂症患者的主要脑区，因此根据结果可以推断出近红外光谱数据通过与模式识别方法的结合可以成为辅助诊断精神分裂症病患者的一种手段。

Discriminant analysis of multi-channel near-infrared spectroscopy signal data in the prefrontal lobe

BACKGROUND：Psychiatric disorders such as schizophrenia are largely diagnosed on symptomatology. Recently pattern recognition approaches to the analysis of neuroimaging data such as the classification of patients and healthy controls have attracted people’s interest. OBJECTIVE：To apply pattern recognition approaches to distinguish schizophrenia patients from healthy subjects with multi-channel prefrontal near-infrared spectroscopy signals, and to verify its feasibility. METHODS：The near-infrared spectroscopy data were measured in the bilateral prefrontal areas of schizophrenia patients and healthy subjects during the Verbal Fluency Test task. After preprocessing, we calculated their mean values for each channel, and ranked the channel features based on the area under curve of the Receiver Operator Characteristic. Then, we trained support vector machine on the combinative features and applied Leave-One-Out-Cross-Validation method to verify the classification ability. RESULTS AND CONCLUSION：Our study demonstrated that the combination of the top eight rank channel features could reach the classification accuracy up to 95.24%, and al these channels are located at the right lateral prefrontal cortex. It is inferred that, right lateral prefrontal cortex is the main dominant brain areas in schizophrenia patients;the near-infrared spectroscopy of right lateral prefrontal cortex is a potential means for＆nbsp;assistant diagnosis of schizophrenia patients.