The paper aims to investigate the differences in Feature Descriptor
patterns during mental rotation of five different objects for both healthy as
well as brain diseased subjects. Therefore, electroencephalographic (EEG)
activity was measured during mental rotation of objects by various angles with
respect to its present orientation. Source localization using eLORETA inferred
an enhanced activation of pre-frontal and frontal lobe regions during mental
rotation task. Experimental analysis also confirmed maximal activation of lower
alpha frequency band while performing this cognitive task. Differential
Evolutionary (DE) algorithm has been implemented to select the optimal features
which are represented using the Feature Descriptor diagrams. These diagrams
infer that the feature patterns are distinct and vary from object to object.
Moreover, these patterns orient by 450 for 900 mental
rotation and by 750 for 1800 mental rotation of the
presented objects. However, there exists an inconsistency in the Feature
descriptor diagrams for patients suffering from pre-frontal lobe amnesia and
Alzheimer’s disease. It is also found that these diagrams remain unaffected
during mental rotation which infers their incapability to perform such a cognitive
activity. Hence, this work can be effectively utilized to detect people
suffering from memory related disorder.

Index Terms— mental rotation, Differential
Evolutionary (DE) algorithm, power spectral density (PSD), Feature Descriptor.

We Will Write a Custom Essay Specifically
For You For Only $13.90/page!


order now

                                                                                                                                                                
I. Introduction

Extensive research on mental rotation has been performed
for the past few decades including both behavioral and neuronal aspects. Mental
rotation is described as a cognitive process involving spatial transformation
of an object when rotated at a particular orientation from its present state.
Shepard and Metzler 1 originally introduced this concept where a given figure
had to be mentally rotated to determine whether it was equivalent to a target
figure represented in a separate orientation. This study inferred a linear
increase in response time with angular variation in depicted orientation.
Similar behavioral studies 2-3 revealed the enhancement of reaction time with
increasing discrepancy between the upright object and its displayed
orientation. Moreover, other literature studies 4-8 dealt with the execution
speed of mental rotation tasks.

In addition to these evidences, other neuroimaging studies
brought into limelight the role of different brain lobes while performing
mental rotation tasks. Cohen et al. 9 reported the involvement of the frontal
eye field and superior parietal regions during mental rotation. Creem et al.
10 inferred the activation of secondary visual, left posterior parietal, premotor
and frontal regions during mental rotation of scenes. Other survey literature
on fMRI study 11-12 revealed the interactive effects of parietal, infero-temporal
and prefrontal cortex while performing mental rotation task. Moreover,
literature studies on phase synchronization patterns 13-14 revealed increased
synchronization between frontal and parietal lobes during execution of mental
rotation activity.

      Though
there are many literatures for mental rotation using EEG signals, there is a
dearth of literature which can better describe the mental rotation of brain
diseased people. The present paper is an endeavor to satiate this void. The
idea is to make the analysis of extracted EEG features having massive dimension
by statistically abridging the data using differential evolutionary based
feature dimension reduction technique, while apprehending the best features to
help in categorizing various objects encoded in human memory. This technique
also helps us to describe the mental rotation for healthy controls as well as
brain diseased patients.

The rest of the paper is organized as follows- Section II describes
the principles and methods adopted, Section III provides the details of the
experiments and the analysis of results and the conclusions are drawn in
section IV.

                                                                                                                                            
II. Principles
and methodologies

A. Feature Extraction

Temporal features
extracted by EEG analysis contain vital information regarding the cognitive
task being performed by a subject. The occurrence of neuronal excitation due to
execution of a mental task is associated with specific range of frequency
bands. Hence, frequency-domain features like power spectral density (PSD) 15
play a vital role in deciphering brain imagery. Other commonly extracted
features include Hjorth parameters 16, wavelet coefficients 17 and AAR
parameters 18.

B. Feature Selection using Differential
Evolutionary (DE)
Algorithm

Differential evolution (DE) is a stochastic population-based
optimization algorithm. An outline of DE is given below.

Initialisation: DE starts with a population of NP,
D-dimensional target vectors
(representing the candidate solutions of an optimization problem) randomly
initialized within the prescribed bound where