Deep learning of resting state networks from independent component analysis

Independent component analysis (ICA) is a powerful technique for analyzing functional networks of the brain. It decomposes resting-state functional magnetic resonance imaging data into distinct networks that are temporally correlated but maximally independent in the spatial domain. Manual classification of ICA components is labor intensive and requires expertise; hence, a fully automatic algorithm that can reliably detect various types of functional brain networks is desirable. In this paper, we introduce a deep Convolutional Neural Network (CNN) method, which provides an automatic solution for identifying resting-state networks extracted using ICA. Our results demonstrate that the proposed CNN method achieves over 98% classification accuracy and out-performs template matching methods.