Discrimination of ADHD Based on fMRI Data with Deep Belief Network

摘要

Effective discrimination of attention deficit hyperactivity disorder (ADHD) using imaging and functional biomarkers would have fundamental influence on public health. In this paper, we created a classification model using ADHD-200 dataset focusing on resting state functional magnetic resonance imaging. We predicted ADHD status and subtype by deep belief network (DBN). In the data preprocessing stage, in order to reduce the high dimension of fMRI brain data, brodmann mask, Fast Fourier Transform algorithm (FFT) and max-pooling of frequencies are applied respectively. Experimental results indicate that our method has a good discrimination effect, and outperform the results of the ADHD-200 competition. Meanwhile, our results conform to the biological research that there exists discrepancy in prefrontal cortex and cingulate cortex. As far as we know, it is the first time that the deep learning method has been used for the discrimination of ADHD with fMRI data.