Multi-Region Neural Representation: A novel model for decoding visual stimuli in human brains

Abstract

Multivariate Pattern (MVP) classification holds enormous potential for decoding visual stimuli in the human brain by employing task-based fMRI data sets. There is a wide range of challenges in the MVP techniques, i.e. decreasing noise and sparsity, defining effective regions of interest (ROIs), visualizing results, and the cost of brain studies. In overcoming these challenges, this paper proposes a novel model of neural representation, which can automatically detect the active regions for each visual stimulus and then utilize these anatomical regions for visualizing and analyzing the functional activities. Therefore, this model provides an opportunity for neuroscientists to ask this question: what is the effect of a stimulus on each of the detected regions instead of just study the fluctuation of voxels in the manually selected ROIs. Moreover, our method introduces analyzing snapshots of brain image for decreasing sparsity rather than using the whole of fMRI time series. Further, a new Gaussian smoothing method is proposed for removing noise of voxels in the level of ROIs. The proposed method enables us to combine different fMRI data sets for reducing the cost of brain studies. Experimental studies on 4 visual categories (words, consonants, objects and nonsense photos) confirm that the proposed method achieves superior performance to state-of-the-art methods.

Multi-Objective Cognitive Model: a Supervised Approach for Multi-subject fMRI Analysis

“…Mohr et al analyzed different classification techniques, i.e., the first norm regularized SVM [2], the second norm regularized SVM [9], the Elastic Net [44], and the Graph Net [15], to predict distinctive neural activities in the human brain [29]. They figured out the first norm regularized SVM can rapidly improve the classification performance in fMRI analysis [29,42]. Osher et al developed a network-based method by employing the human brain's anatomical features in order to classify distinctive neural responses [31].…”

“…Osher et al developed a network-based method by employing the human brain's anatomical features in order to classify distinctive neural responses [31]. Yousefnezhad et al proposed two new ensemble learning approaches by utilizing weighted AdaBoost [40], and Bagging [42].…”

“…Cognitive models generated by MVP methods must be validated by employing multi-subject fMRI images [5,7,26,41]. There are mainly two steps in an MVP standard pipeline, which must be applied to preprocessed fMRI images to generate the cognitive models, i.e., functional alignment, classification analysis [42]. Further, each of these main steps can include some subtasks, such as applying feature selection before classification analysis [5,42].…”

“…There are mainly two steps in an MVP standard pipeline, which must be applied to preprocessed fMRI images to generate the cognitive models, i.e., functional alignment, classification analysis [42]. Further, each of these main steps can include some subtasks, such as applying feature selection before classification analysis [5,42]. Different human brains naturally generate distinctive patterns [41,20].…”

“…As discussed before, there are different steps (including the steps and subtasks) for MVP analysis. The main problem is that these steps employ disjoint objective functions, which are separately run to generate a cognitive model [42] and there is no unique solution for each of these objective functions [5]. Therefore, an optimal result in one of these steps may not be optimum for the next steps.…”