Deep learning approaches for analysing visual stimuli from human fMRI

Decoding brain activities corresponding to an external stimulus is an excellent challenge because of the complexity of brain activities data and the understood of the activity of the brain is not yet complete. This research paper focuses on the functional magnetic resonance imaging (fMRI) data of different people corresponding to external visual stimulus images. The images consist of three main types: letters, artificial shapes, and natural images. The proposed model provides an analysis and classification of the three main types of images using Support Vector Machine and Convolutional Neural Networks by proving that the brain is affected differently by each type. In addition, the identification of 200 natural image categories was analyzed by calculating the similarity between the features for each category by using the features of visual images using CNN and fMRI using the regression model. Also, we used deep convolution generative adversarial networks (DCGANs) for image reconstruction.