PPT on BRAIN TUMOR detection in MRI images based on IMAGE SEGMENTATION

www.company.com
EFFICIENT SEGMENTATION
METHODS FOR TUMOR
DETECTION IN MRI IMAGES
BY:
S.Md. NOOR ZEBA KHANAM
S.SAI SOWMYA
G.PREETHI
K.SRAVANTHI

www.company.com
ABSTRACT
 Brain tumor extraction and its analysis are challenging tasks
in Medical image processing because brain image is
complicated.
 Segmentation plays a very important role in the medical image
processing.
 In that way MRI (magnetic resonance imaging) has become a
useful medical diagnostic tool for the diagnosis of brain & other
medical images.
 In this project, we are presenting a comparative study of Three
segmentation methods implemented for tumor detection.
 The methods include k-means clustering using watershed
algorithm, optimized k-means and optimized c-means using
genetic algorithm.

www.company.com
INTRODUCTION
• The BRAIN is the most important part of central nervous system.
• The main task of the doctors is to detect the tumor which is a
time consuming for which they feel burden.
• Brain tumor is an intracranial solid neoplasm.
• The only optimal solution for this problem is the use of ‘Image
Segmentation’.
Figure : Example of an MRI showing the
presence of tumor in brain

www.company.com
IMAGE SEGMENTATION
• The purpose of image segmentation is to partition an
image into meaningful regions with respect to a particular
application.
• The segmentation might be grey level, colour, texture,
depth or motion.
• Example:
……

www.company.com
EXISTING METHODS
Fusion based : Overlapping the train image of the victim over a
test image of same age group, thereby detecting the
tumor.
Demerits :
 The overlapping creates complexity due to different
dimensions of both images.
 Time consuming process.
Canny Based : To overcome the problem of detecting the edges,
the better way is the use of Canny based edge detection.
Demerits :
 Not support color images.
 This leads to increase in time to reach the optimal solution.

www.company.com
PROPOSED METHOD
 The method include
‘k-means clustering +watershed,
optimized k-means +genetic algorithm
and
optimized C- means +genetic algorithm’.
 At the end of process the tumor is extracted from the MRI
image and also its exact position and shape are determined in
colour.

www.company.com
THEME OF PROPOSED METHOD
K-means
+
watershed
Optimized
K-means
+
GA
Optimized
C-means
+
GA
Successful
detection
+
high
accuracy
+
color.

www.company.com
Clustering
• Clustering is a process of collection of objects which are
similar between them while dissimilar objects belong to
other clusters.
• A clustering technique is used to obtain a partition of N
objects using a suitable measure such as resemblance
function as a distance measure ‘d’.

www.company.com
Region of
interest
Center of
mass
CLUSTERING PROCESS
10

www.company.com
Region of
interest
Center of
mass
CLUSTERING PROCESS
11

www.company.com
Region of
interest
Center of
mass
CLUSTERING PROCESS
12

www.company.com
Figure : Clustering Technique
Final Clusters

www.company.com
K-means clustering
k-means clustering aims to partition n observations into ‘K’
clusters in which each observation belongs to the cluster
with the nearest mean.
(a) original image (b) expert selection (c) K-means selection

www.company.com
WATERSHED ALGORITHM
• Watershed algorithm is used in image process primarily
for segmentation purposes.
• This algorithm can be used if the foreground and
background of the image can be identified.
MERITS:
 It works best to capture the weak edges.
 Watershed algorithm improves the primary results of
segmentation of tumour done by k-means.

www.company.com
K-means clustering with watershed
Merits:
 If variables are huge, then K-Means most of the times
computationally faster than, if we keep k small.
 Watershed algorithm improves the primary results of
segmentation of tumour done by k-means.
Demerits:
 Difficult to predict K-Value & k-means cannot find non-
convex clusters.
 Different initial partitions can result in different final
clusters.
 This method does not work well with clusters of different
size and different density.

www.company.com
C-means clustering
• It is well known that the output of K-Means algorithm
depends hardly on the initial seeds number as well as the
final clusters number.
• Therefore to avoid such obstacle FCM is suggested.
• The fuzzy C-means relax the condition by allowing the
feature vector to have multiple membership grades to
multiple cluster.
Figure: Result of Fuzzy C-means

www.company.com
GENETIC ALGORITHM
• The term genetic is derived from Greek word ‘genesis’
which means ‘to grow ‘or ‘to become’.
• The implementation of Genetic algorithm begins with an
initial population of chromosomes which are randomly
selected.
MERIT:
 It is the best optimizing tool.
 It gives best result when used with Fuzzy c-means
clustering…

www.company.com
C-means clustering with Genetic
algorithm
MERITS:
 This method considers only image intensity.
 Unlike k-means where data point must exclusively belong
to one cluster center here data point is assigned to 2 or
more clusters.
DEMERITS:
 Aprior specification of the number of clusters.
 We get the better result but at the expense of more
number of iteration.

www.company.com
MAIN STRATEGY OF PROPOSED
METHOD
Proposed Method
Tumor is detected with high
accuracy
Effectively detects the tumor
area & internal Structure
We get the resultant image in
color
C-means clustering + Genetic algorithm
Here data point is assigned to
2 or more clusters
GA gives best result in little
time
Best when works with C-
means
K-means Clustering + Watershed algorithm
It is computationally faster, if we take K value
small
WA used to capture weak edges

www.company.com
FUTURE SCOPE
In terms of the near-future
 As Medical image segmentation plays a very important
role in the field of image guided surgeries.
 By creating Three dimensional (3D) anatomical models
from individual patients, training, planning, and computer
guidance during surgery is improved.

www.company.com
RESULTS:
Fig.1.Results for first stage as K-means
clustering.
Fig.2.Results of Watershed algorithm

www.company.com
RESULTS:
Fig: Result of K-means and Watershed algorithm for one test image

www.company.com
RESULTS:
FIG: Resultant Image of
C-means Clustering for
cluster-1, cluster-2, cluster-3

www.company.com
RESULT :
FIG: Final MRI image for One Test image

PPT on BRAIN TUMOR detection in MRI images based on IMAGE SEGMENTATION

More Related Content

What's hot

Viewers also liked

Similar to PPT on BRAIN TUMOR detection in MRI images based on IMAGE SEGMENTATION

Recently uploaded

PPT on BRAIN TUMOR detection in MRI images based on IMAGE SEGMENTATION