Analysis of fMRI data based on prediction of neural response using MVPA

Abstract

fMRI data is an emerging approach that shows all the information of the brain that is represented in the subject of the brain at a particular point in time. Multi-voxel pattern analysis (MVPA) is gaining interest in the neuro imaging because it allows Cognitive states to be modeled as distributed patterns of neural activity. The MVPA approach allows to several cognitive state of brain reading. In order to relate the neural actions to cognition in fMRI information Multi-voxel sample examination is used. The intend of this project is to construct a classification model. When the quantity of features(voxels) exceeds the total quantity of data scrutiny, it creates replica over fitting. For this reason it is significant to choose informative voxels previous to constructing a classification model. This project mostly deals with two methods that are MI and partial least square regression (PLS). By using these two techniques effective feature assortment is done. Based on the degree of association to the stimulus circumstances informative voxel index must be created. The proposed work is evaluated by determining the performance of standard classification algorithms. The results obtained from the proposed work which is based on PLS and MI method improves the classification accuracy.