Cognitive State Classification using Genetic Algorithm based Linear Collaborative Discriminant Regression

Abstract

Functional Magnetic Resonance imaging (fMRI) provides sequence of 3D images which contains large number of voxels as information. There are many statistical methods evolved in last few years to analyze this information. Main concern of all these techniques is huge dimensions of the data produced by these images. This paper proposes an efficient hybrid method for feature selection and classification. This method combine entropy based genetic algorithm (EGA) with Linear Collaborative Discriminant Regression Classification (LCDRC) to form feature based classification method. Entropy based genetic algorithm is applied to find maximum significance between the input and output and also it radically reduces the redundancy within the input features. Experiments’ using Star-Plus dataset to classify fMRI images shows that EGA-LCDRC reduces up to 60% features and produces 96.73% accuracy.