Deep Learning Methods for Autism Spectrum Disorder Diagnosis Based on fMRI Images

Abstract

Brain injuries are significant disorders affecting human life. Some of these damages can be completely eliminated by methods such as drug therapy. On the other hand, there is no known permanent treatment for damages caused by diseases such as Alzheimer, Autism spectrum disorder (ASD), Multiple sclerosis and Parkinson. Treatments aimed at slowing the progression of the disease are generally applied in these types of disorders. For this reason, essential to diagnose the disease at an early phase before behavioral disorders occur. In this study, a study is presented to detect ASD through resting-state functional magnetic resonance imaging rs-fMRI. However, fMRI data are highly complex data. Within the study's scope, ASD and healthy individuals were distinguished on 871 samples obtained from the ABIDE I data set. The long short-term memory network (LSTM), convolutional neural network (CNN) , and hybrid models are used together for the classification process. The results obtained are promising for the detection of ASD on fMRI.

Keywords

• deep learning
• rnn
• rs-fmri
• autism spectrum disorder