Classifying brain state in sentence polarity exposure: An ANN model for fMRI data

Abstract

Brain being the most complex organ of human body in which millions of neuron are connected to each other, and pass information in processing of thoughts, emotions, motor activities and linguistic phenomenon. With the advent of non-invasive neuro-anatomical analysis methods like PET scan, fMRI it is now easy to measure neuronal changes in brain. This study analyses the neuronal activity in the brain in sentence polarity detection task using multilayer perceptron classification methodology. The whole brain is divided into almost 5000 three-dimensional volume called voxels from which prominent voxels are selected using symmetrical uncertainty based on entropy for the classification of brain state. The proposed method achieved significantly higher accuracy in classifying brain state in the processing of affirmative and negative sentences. The result obtained also shows that certain brain regions like left dorsolateral prefrontal cortex (LDLPFC) and calcarine sulcus (CALC) are prominent areas which are deterministic in classification of affirmative and negative sentences in brain while right posterior pre-central sulcus (RPPREC) and right supramarginal gyrus (RSGA) are less contributing. © 2020 Lavoisier. All rights reserved.