Classification of physiological diseases using eeg signals and machine learning techniques

IJRET: International Journal of Research in Engineering and Technology eISSN: 2319-1163 | pISSN: 2321-7308
_______________________________________________________________________________________________
Volume: 05 Issue: 08 | Aug-2016, Available @ http://ijret.esatjournals.org 346
CLASSIFICATION OF PHYSIOLOGICAL DISEASES USING EEG
SIGNALS AND MACHINE LEARNING TECHNIQUES
Rashmi Sahu1, Veena N2
1
MTech Student, Department of CSE, BMSIT&M, Bangalore
2
Assistant Professor, Department of CSE, BMSIT&M, Bangalore
Abstract
In rural areas providing advanced diagnostics for various health disorders is not possible in countries like India. With lat est
technological breakthrough,brain signals(EEGsignal) capturing devices are available at rate less 50$. If these brain signalscan
be used to predict any Physiological disorders like heart problem, kidney problems etc., then these EEG devices can be provid ed
to rural health care centre for preliminary investigation and on diagnosis the patient can move to city hospitals for diagnostics
and treatment. In this project, we provide a solution of identifying physiological problems using EEG signals and use machine
learning techniques for diagnosis.
Keywords: EEG Signals, EEG Frame, Feature Extraction
---------------------------------------------------------------------***---------------------------------------------------------------------
1. INTRODUCTION
The PC help is requested in restorative organizations
because of the way that it could enhance the consequences
of people in such a space where the false antagonistic cases
must be at a low rate. It has been demonstrated that twofold
perusing of therapeutic pictures could prompt better Tumor
identification. In any case, the expense suggested in twofold
perusing is high, that is the reason great programming to
help people in restorative foundations is of incredible
intrigue these days. Customary techniques for checking and
diagnosing the infections depend on distinguishing the
nearness of specific components by a human watch.
There are huge number of patients in serious consideration
units and the requirement for persistent perception of such
conditions;a few methods for mechanized analytic
frameworks have been produced as of late to endeavor to
take care of this issue. Such procedures work by changing
the for the most part subjective demonstrative criteria into a
more target quantitative element order issue. Robotized
grouping of Brain signs by utilizing some earlier
information like force and some anatomical components is
proposed. At present there are no strategies broadly
acknowledged along these lines programmed and
dependable techniques for Tumor recognition are of
awesome need and intrigue.
Brain–Computer Interfaces (BCI) is the best plausible
method for giving the correspondence between the human
and the framework by method for mind signals. Using this
BCI the patients can put over their perspectives or
necessities by method for their cerebrum flags just by
deduction process. The sign characterization module is
made out of the acquired EEG signal elements extraction
and the change of these signs into gadget directions. The
EEG order strategy relies on upon the affectation and, in this
way, the response to identify engine symbolism, occasion
related possibilities, moderate cortical possibilities, or
relentless state evoked possibilities. The anticipated EEG
drives the characterization to some exact component
extraction strategies. Doing mind surgery is a confused
undertaking keeping in mind the end goal to that they take
MRI sweep and CT check yet at times the picture may give
reciprocal data and make trouble in finding the tumor from
the white matter of the cerebrum the most recent advances
in PC innovation and diminished expenses have made it
conceivable to grow such frameworks.
Train a machine learning classifier to identify physiological
diseases using EEGsignals and measure the accuracy of this
in terms of classification accuracy.
2. RELATED WORK
We are extending the domain of EEG signal application to
physiological disease diagnosis. EEG signals for unhealthy
individuals are taken. From this elements are separated from
signs utilizing PCA (Principal Component Analysis) system
and after that these elements and elements from solid
subjects are utilized to prepare a Neural Network. In light of
the preparation, the neural system will have the capacity to
anticipate any new flag test components to infected or sound
subject.
Approach
1. First EEG signals for unhealthy and solid subjects are
gained from doctor's facilities and web.
2. PCA examination is done on the signs and components
are extricated.
3. Features are utilized to prepare a neural system.
4. Once neural system preparing is finished, it can be used
for arrangement.
5. To analyze a man for issue, his EEG sign is gathered.
6. PCA is done on gathered sign to concentrate highlights.

_______________________________________________________________________________________________
7. Neural Network is utilized order the component to be
ailing or ordinary.
Advantage of our Solution
1. The solution cost is very less and can be used in rural
health centers for preliminary investigation.
2. Very little trained man power is only needed as the
systemis very easy to use.
3. ARCHITECTURE
Framework engineering describes applied configuration
which categorizes the conduct and structure of a framework.
A design rendering is a formal depiction of a framework,
arranged in a way that assists thinking about the
fundamental features of a framework. It categorizes the
product structure segments and generates an order out of
which items can be procuredand frameworks built up, that
will help in executing general framework.
Fig-1: System architecture
Feature selection: This module takes training EEG data set
as input. This training data set is helpful for training using
naïve bayes classifier model. Based on the training, the
neural network will be able to predict any new signal sample
features to diseased or healthy subject.
Naïve Bayes classifier:It is a classifier based on Bayes
probability theorem also known as probabilistic classifier.
Naive Bayes classifier is very efficient since it requires a
small amount of training data.
Classifier: Machine learning classifier is used to identify
physiological diseases using EEG signals and measure the
accuracy of this in terms of classification accuracy. The
admin will be responsible for load train data set, train
classifier and measure the accuracy.
Depiction of Classes for the System
Unified Modeling Language (UML) diagram’s class chart is
kind of static pattern graph. It uses framework's classes and
their properties to depict the structure of a framework and
connections between various classes. A class chart is
depicted below:
Fig-2: Class Diagram
The class diagram has the following classes:
Main class: This class has operations load trainset, do train,
invoke classifier and view accuracy.
NB Classifier Class: This class has operations called create
train vector, train and classify.
Performance class: This class has operations called invoke
performance and get accuracy chart.
System’s Operation Sequence Diagram
Unified Modeling Language’s (UML’s) sequence diagramis
a kind of communication diagram which shows how
processes work with each other and in which order.
Sequence diagram is a form of a Message Sequence Chart.
Fig-3: Sequence Diagram of Network Train Flow
Here admin, main, NB classifier are objects. Each object
interacts with other objects in a sequential order through
messages. As shown above.
Feature
Selection
Naïve bayes
Training
Classifier
Feature
Selection
EEG
form part
Diseased /
healthy
Training
EEG
Data set
Label (diseased/ healthy)
+loadtrainset()
+doTrain()
+invokeClasification()
+viewAccuracy()
Main
+createTrainVector()
+train()
+classify()
NB Classifier
+invokePerforamnce()
+getAccuracyChart()
Performance
1
1
1
1
Admin NB ClassifierMain
Load train data
do Train
Create train vector
new
train
Create train
model

_______________________________________________________________________________________________
Sequence Diagram for Test Flow
Fig-4: Sequence Diagram for Test Flow
Here admin, main, NB classifier are objects. Each object
interacts with other objects in a sequential order through
messages. As shown above.
Fig-5: Sequence diagram for performance flow
Here, admin, main, performance and NB classifier are called
objects. Each objects interacts with each other by
exchanging the messages.
4. IMPLEMENTATION
4.1 EEG Framework
The primary reason for the EEG Frame structure is to
concentrate highlight vectors from EEG records and store
them for further information disclosure.
Fig-6: Overview of the EEG framework
The structure accepts that the information in EDF
configuration is accessible, and that it has been separated
(the system does exclude EEG pre-processing techniques. It
then empowers visual review of a solitary record and
highlight extraction from the record, by determining the
extraction parameters.
The yield document is recorded in .csv position that can be
perused by the greater part of the open-source information
revelation stages, for example, Weka or Rapidminer.
Admin
classifier
NB classifier
Diseased/ healthy
invoke classifier
Diseased/ healthy
Main
Admin
Invoke performance
Performance
done
view accuracy
get accuracy chart
Main NB Clasifier
classify
done
Accuracy chart

_______________________________________________________________________________________________
4.2 Testing
Table-1: Module Testing
Classes Implemented Functions used in classes Tests Executed Observation
Main
Load Train set()
Do train()
Invoke Classifier()
view Accuracy()
Functionality is working correctly Pass
Naïve Bayes Classifier
Train vector()
Train()
Classify()
Module are functioning properly Pass
Performance
Invoke performance()
get accuracy()
Measured Performance Pass
5. RESULTS
Fig-7: Selecting Signals for Feature Extraction
Fig-8: Feature Extraction
Fig-
9: Selection of Linear Time Domain Features
Fig-10: Selection of Approximate Entropy and Sample
Entropy

_______________________________________________________________________________________________
Fig-11: Train the Classifier by Browsing the Training Data
Set Text File
Fig-12: Training Completed
Fig-13: Classification
Fig-14: Log Information about Records and no. of Disease
Prone
Fig-15: Performance Measurement Process
Fig-16: Accuracy Graph for Data Set
6. CONCLUSION
In order to improve the system performance, the classifier
has trained with the features of principal components. The
reproduced framework gave that the better order exactness
of information tests and similarity in this determination.
Robotized cerebrum ailments determination framework with
directed back engendering with food forward neural system
to distinguish variation from the norm, for example, tumor
and epileptic case.

_______________________________________________________________________________________________
ACKNOWLEDGEMENT
The satisfaction and euphoria that accompany the successful
completion of any task would be incomplete without the
mention of the people who made it possible, whose constant
guidance and encouragement crowned my effort with
success.
I extend my deep sense of sincere gratitude to Dr. MOHAN
BABU. G N, Principal,BMS Institute of Technology &
Management for providing me an opportunity to continue
my higher studies.
I express my sincere gratitude to Dr. THIPPESWAMY. G,
HOD, Dept. of Computer ScienceandEngineering, BMS
Institute of Technology & Management for his valuable
suggestions and expert advice.
I deeply express my sincere gratitude to my internal guide,
Mrs. Veena N., Assistant Professor, Department
ofComputer Science and Engineering, for her
encouragement and advice throughout this Project work.
Special thanks to all the staff members of Computer Science
Department for their help and kind co-operation.
Last, but not least, I would like to thank my parents and
friends for their encouragement and support given to me in
order to finish this precious work.
REFERENCES
[1]. EEG Brain Signal Classification for Diseases Diagnosis,
Durai Kumar.D1, Balaji S2, ShaziyaParveen B*3,
IJESR/May 2015/ Vol-5/Issue-5/204-211
[2]. Temporal and Spatial Features of Single-Trial EEG for
Brain-Computer Interface Qibin Zhao and Liqing Zhang,
Department of Computer Science and Engineering,
Shanghai Jiao Tong University, Shanghai 200240, China.
[3]. Large-Scale Image Database Triage via EEG Evoked
Responses, Yonghong Huang1, Deniz Erdogmus1, Santosh
Mathan2 and Misha Pavel1, 1 Oregon Health & Science
University, Portland, OR, USA, 2 Human Centered Systems
Laboratory, Honeywell, Redmond, WA, USA.
[4]. Brain Tumor Classification using Probabilistic, Neural
Network, S.U.Sumathi1, S. Geetha*2, International Journal
of Innovative Research in Science, Engineering and
Technology, (An ISO 3297: 2007 Certified Organization),
Vol. 4, Issue 9, September 2015.
[5]. Texture Classification, Using Weighted Probabilistic
Neural Networks, M. Kiranmayee& M. Subbarao, Dept. of
Information Technology, AITS, Rajampet, A.P.
BIOGRAPHIES
Rashmi Sahu, BE (Information
Technology) M.Tech (Computer Science &
Engineering) from BMSIT&M, Bangalore.
Mrs. Veena N., Assistant Professor,
Department of CSE, BMSIT&M, Bangalore

Classification of physiological diseases using eeg signals and machine learning techniques

More Related Content

What's hot (19)

Viewers also liked (20)

Similar to Classification of physiological diseases using eeg signals and machine learning techniques (20)

More from eSAT Journals (20)

Recently uploaded (20)

Classification of physiological diseases using eeg signals and machine learning techniques