Geng Bowen, Gao Ming, Piao Ruiqing, Liu Chengxiang, Xu Ke, Zhang Shuming, Zeng Xiao, Liu Peng, Wang Yanzhu
Life Science Research Center, School of Life Sciences and Technology, Xidian University, Xi'an, China.
Engineering Research Center of Molecular and Neuro Imaging Ministry of Education, School of Life Sciences and Technology, Xidian University, Xi'an, China.
Front Psychiatry. 2022 Jul 25;13:906404. doi: 10.3389/fpsyt.2022.906404. eCollection 2022.
This study aimed to develop an effective support vector machine (SVM) classifier based on the multi-modal data for detecting the main brain networks involved in group separation of premature ejaculation (PE).
A total of fifty-two patients with lifelong PE and 36 matched healthy controls were enrolled in this study. Structural MRI data, functional MRI data, and diffusion tensor imaging (DTI) data were used to process SPM12, DPABI4.5, and PANDA, respectively. A total of 12,735 features were reduced by the Mann-Whitney U test. The resilience nets method was further used to select features.
Finally, 36 features (3 structural MRI, 7 functional MRI, and 26 DTI) were chosen in the training dataset. We got the best SVM model with an accuracy of 97.5% and an area under the curve (AUC) of 0.986 in the training dataset as well as an accuracy of 91.4% and an AUC of 0.966 in the testing dataset.
Our findings showed that the majority of the brain abnormalities for the classification was located within or across several networks. This study may contribute to the neural mechanisms of PE and provide new insights into the pathophysiology of patients with lifelong PE.
本研究旨在基于多模态数据开发一种有效的支持向量机（SVM）分类器，以检测参与早泄（PE）组间分离的主要脑网络。
本研究共纳入52例终身性早泄患者和36例匹配的健康对照。分别使用结构MRI数据、功能MRI数据和扩散张量成像（DTI）数据通过SPM12、DPABI4.5和PANDA进行处理。通过曼-惠特尼U检验减少了总共12735个特征。进一步使用弹性网络方法选择特征。
最终，在训练数据集中选择了36个特征（3个结构MRI、7个功能MRI和26个DTI）。我们在训练数据集中获得了最佳的SVM模型，准确率为97.5%，曲线下面积（AUC）为0.986，在测试数据集中准确率为91.4%，AUC为0.966。
我们的研究结果表明，用于分类的大多数脑异常位于几个网络内或跨几个网络。本研究可能有助于了解PE的神经机制，并为终身性PE患者的病理生理学提供新的见解。
