Voxel selection framework based on meta‐heuristic search and mutual information for brain decoding

Abstract

Visual stimulus decoding is an increasingly important challenge in neuroscience. The goal is to classify the activity patterns from the human brain; during the sighting of visual objects. One of the crucial problems in the brain decoder is the selecting informative voxels. We propose a meta‐heuristic voxel selection framework for brain decoding. It is composed of four phases: preprocessing of fMRI data; filtering insignificant voxels; postprocessing; and meta‐heuristics selection. The main contribution is benefiting a meta‐heuristics search algorithm to guide a wrapper voxel selection. The main criterion to nominate a voxel is based on its mutual information with the provided stimulus label. The results show impressive accuracy rates which are 90.66 ± 3.66 and 91.61 ± 8.24 for DS105 and DS107, respectively. This outperforms the most of existing brain decoders in similar validation conditions. The experimental results are very encouraging which can be successfully used in the brain‐computer interface.

New Insight of Human Brain Connectivity Mapping based on Inter-correlation of Multi-view fMRI Decoder

“…(Algo. 1) illustrates the proposed voxel selection method [26]. For instance, in the Haxby dataset [23], there are eight categories of objects where the duration and time of repetitions, TR, are 24 and 2.5 Seconds, respectively.…”

“…Moreover, voxels usually contain noise and irrelevant information; consequently, many pre-processing routines like normalization, scaling, and voxel selection would be required to conduct activity localization. Some exciting studies on the voxel selection problem can be found in [24]- [26]. Various multiview (MV) learning applications in the real world exist due to several features that describe an object.…”