基于小波变换与SVM的ADHD病人分类

Classification Based Wavelet Translate and SVM in the ADHD

提出基于小波变换的特征提取方法对ADHD病人进行分类研究。采用115名ADHD-200的竞赛静息态功能磁共振数据,首先提取了90个脑区的平均时间序列信号,然后利用小波变换多分辨率分析特性对信号进行3层分解;计算了各个尺度下小波系数的能量值,对能量值进行归一化处理后,将其作为分类特征向量;最后结合SVM分类器采用留一交叉验证法对ADHD病人进行分类。结果表明该方法有助于ADHD病人的分类与诊断。

In this study, we propose an approach to extract features based wavelet transform for the ADHD classification. One hundred and fifteen subjects＇ resting state fMRI data were adopted, which come from ADHD-200 competition. We first extracted the time series of ninety brain areas, and decomposed them into three levels using the wavelet transform for each subject. Secondly, the energy values of any scale were computed and normalized, which construct the classification feature vectors. Finally, we combined the SVM to classification in the ADHD based leave-one-out cross validation. The results demonstrate that the wavelet transform feature extract approach is useful in classification and diagnosis for ADHD.