Analysis and Classification of Discriminative Region in Cognitive Functional MRI Data

摘要

Feature engineering techniques such as feature selection and extraction dominate the process of cognitive state learning. The extraction of relevant features from high-dimensional multi-way functional MRI (fMRI) data is essential for the classification of a cognitive task. The dimensionality of fMRI influences the analysis of brain data. fMRI data is arranged as a number of voxels, region of interests (ROI) and snapshots. The extraction of a specific pattern of interest within the noisy components is a challenging task. In this paper, a tensor gradient-based feature extraction technique decomposes the multi-way fMRI data into a number of components. Voxel time series data from different ROIs has been used to find the region of discrimination. Clustering-based maximum margin feature selection method has been proposed to select the minimum number of voxels as attributes. The proposed techniques provide a better learning accuracy for the StarPlus fMRI data.