AUTHOR=Dou Ruhai , Gao Weijia , Meng Qingmin , Zhang Xiaotong , Cao Weifang , Kuang Liangfeng , Niu Jinpeng , Guo Yongxin , Cui Dong , Jiao Qing , Qiu Jianfeng , Su Linyan , Lu Guangming 

TITLE=Machine learning algorithm performance evaluation in structural magnetic resonance imaging-based classification of pediatric bipolar disorders type I patients

JOURNAL=Frontiers in Computational Neuroscience

VOLUME=Volume 16 - 2022

YEAR=2022

URL=https://www.frontiersin.org/journals/computational-neuroscience/articles/10.3389/fncom.2022.915477

DOI=10.3389/fncom.2022.915477

ISSN=1662-5188

ABSTRACT=The diagnosis based on clinical assessment of pediatric bipolar disorder (PBD) may sometimes lead to misdiagnosis in clinical practice. For the past several years, machine learn (ML) methods was introduced into the classification of BD, which was helpful to the diagnosis of BD. In the present study, brain cortical thickness and subcortical volume of 36 PBD-Ⅰ patients and 19 age-sex matched healthy controls (HCs) were extracted from magnetic resonance imaging (MRI) data and set as features for classification. The dimensionality-reduced feature subset which filtered by Lasso or f_classif was sent to the 6 classifiers (logistic regression, support vector machine, random forest classifier, naïve bayes, k-nearest neighbor, AdaBoot algorithm), and the classifiers were trained and tested. Among all the classifier, the top 2 classifiers with the highest accuracy were logistic regression (84.19%) and support vector machine (82.80%). Feature selection was performed in the 6 algorithms to obtain the most important variables including right middle temporal gyrus and bilateral pallidum, which is consistent with structural and functional anomalous changes of these brain regions in PBD patients. These findings take computer-aided diagnosis of BD a step forward.