Multi-Class Brain Age Discrimination Using Machine Learning Algorithm

Abstract

Resting-state functional connectivity analyses have revealed a significant effect on the inter-regional interactions in brain. The brain age prediction based on resting-state functional magnetic resonance imaging has been proved as biomarkers to characterize the typical brain development and neuropsychiatric disorders. The brain age prediction model based on functional connectivity measurements derived from resting-state functional magnetic resonance imaging has received a lots of interest in recent years due to its great success in age prediction. However, some of the recent studies rely on experienced neuroscientist experts to select appropriate connectivity features in order to build a robust model for prediction while the others just selected the features based on trial-and-error test. Besides, the subjects used in this studies omitted some subjects that can be divided into two groups with less similarity which may confused the prediction model. In this study, we proposed a multi-class age categories discrimination method with the connectivity features selected via K-means clustering with no prior knowledge provided. The experimental results show that with K-means selected features the proposed model better discriminate multi-class age categories.