Analysis and 3D representation of visual information from fMRI of ventral temporal cortex using neural network and support vector machine

Abstract

Ventral Temporal Cortex (VTC) is a part of human brain that is responsible for the identification of different types of visual stimuli. The proposed model revolves around the analysis of VTC from fMRI of 6 subjects who were watching certain objects during the scan. Firstly, dataset was processed and an Artificial Neural Network model was trained to predict which object was being viewed by the subject based on the fMRI images. With proper training, approximately 90% accuracy was achieved on an average. Secondly, Support Vector Machine (SVM) algorithm was used to compare the activity of VTC for processing images of 2 different objects and the comparison was mapped into 2D anatomical MRI image of the corresponding brain. Thirdly, a 3D anatomical model was constructed from a series of 2D anatomical MRI slices and the active portion of the VTC was presented in the 3D structure. Finally, all the findings and the potential future works were analyzed and discussed.