Ling Li ^1* Jiahui Li ¹ Hui Wu ^2* Yanping Zhao ^1* Qinmei Liu ^2* Hairong Zhang ^1* Wei Xu ^2*

• ¹ Jilin University, Changchun, China
• ² Department of Neonatology of The First Hospital of Jilin University, Changchun, Hebei Province, China

Approximately 15 million premature infants are born every year, who are suffering the risks of neurological impairments. Accurate assessment of brain maturity in preterm infants is crucial for timely intervention and treatment planning. Electroencephalography (EEG) is a noninvasive method used for this purpose. However, the assessment of brain maturity utilizing all channels and entire features may cause the problems of the high computational burden and overfitting, reducing the performance of the prediction system. In this study we propose an automatic prediction framework based on EEG to predict the functional brain age (FBA) for assessing brain maturity in preterm infants, in which we focus on optimizing channels and features selection. For channel selection, we present a method based on the combination of binary particle swarm optimization (BPSO), forward addition (FA) and backward elimination (BE) methods. For feature selection, we combine the Pearson correlation coefficient (PCC), recursive feature elimination (RFE) method and support vector machine (SVR) model. The experimental results show that the proposed automatic prediction framework based on SVR model is superior to other comparative methods, achieving prediction accuracy of the FBA within ±1 week is 76.71%, and that within ±2 weeks is 94.52%. Effective channels and features selection for the prediction of the FBA based on EEG in preterm infants can enhance model's performance while reducing computational costs.