SNS PROJECT PROPOSAL

PROJECT TITLE:
BRAIN TUMOR DETECTION by Using
MATLAB.

GROUP MEMBERS:

DUA ZAHRA CPEN221101020

HAMZA ARSHAD CPEN221101031

MUJTABA AHMAD CPEN221101023

 Abstract :
Image segmentation is one of the most challenging techniques in
the field of medical image processing. Brain tumor segmentation is
emerging technique in this field. This paper refers to the brain
tumor segmentation from MR images of patients taken from ‘Brain
web’. Hybrid methodology is used to extract and detect tumor from
MR images using basic concepts of digital image processing.
MATLAB is used for this purpose in this paper for applying the
proposed algorithm.

Introduction:
Brain tumor is abnormal growth of tissues which grow
uncontrollably and unchecked by the check points which
control the growth of cells normally. Brain tumor can be either
primary or metastatic.
Tumor that spread from other parts of body to brain is said to
be metastatic. Brain tumor is one of the likely causes of
mortality among children and adult. In children brain tumors
are the cause of one quarter of all cancer deaths. There are
about 200 different types of tumors diagnosed in UK each year.
Medical resonance imaging is an advance imaging technique
which is used to obtain high quality images of different parts of
body. MRI is used for anatomical analysis of brain development
and brain abnormalities (Logeswari and Karnan 2009). These
MRI images are pre-processed so that further morphological
operations can be performed on these images for the detection
of size, shape and location of tumor.
Pre-processing of MRI images is done for noise removal and
image enhancement while morphological operations are
performed using MATLA B algorithms to separate and detect
tumor in brain(Murugavalli and Rajamani 2007).
The ultimate goal of segmentation is to extract important
features from the image data. The segmentation of tumor from
MRI images of brain is time consuming process(Toure, Beiji et
al. 2010). MATLAB is fast algorithm used for detection of tumor
from MRI images in a very short time. At the end tumor is
mapped onto original greyscale image with 255 intensity to
make tumor visible in the image.
Methodology:
These are the steps performed on MRI image of brain tumor
using MATLAB algorithms.

median filter is
highpass filter is
MRI imageof brain convert it into applied to
applied for noise
is given as input greyscale image enhance image
removal
quality

threshold
watershed morphological
segmentation is
segmentation is operation is
computed using
computed computed
MATLAB

SYSTEM BLOCK DIAGRAM:

Grayscale image:
Grayscale image is preferred format for image processing
because gray scale images are much less complex and one can
talk in detail about contrast, brightness, and edges without
considering colors(Roy and Bandyopadhyay 2012). Colors are
much more complex since they are composed of three
channels that are why image is converted into grayscale before
further processing in image segmentation.

Morphological operations:
Top hat filter is used to highlight the sharp peaks and gradients
in image. Apply morphological operations in which ‘strel’ i.e.
structuring element of disk shaped is taken and applied on the
grayscale image for extracting the tumor from MR image of
patients(Deng, Xiao et al. 2009). The tumor obtained in the
above step has no clear edges or boundaries but it has
somehow sharpened the tumor in the image.

Watershed segmentation:
Watershed image is superimposed over the image after high
pass filter has been applied. Watershed transformation is done
to segments the cells in an image. It is applied on gray scale
image using MATLAB watershed algorithms.
For this work, real world patient data is taken for analysis from
‘brain web’. As the tumor in MR images has more intensity
than the background that’s why it becomes easy to locate and
extract the tumor from MR image.
Previously segmentation was done manually but that process is time
consuming and demand intensive labor(Bezdek, Hall et al., 1993).
That’s why process was automated but still results were not
satisfactory.
Some of the brain tumor MR images taken for testing the proposed
methodology using the hybrid approach for image segmentation that
involves the combination of top hat filter and watershed
algorithm(Patil and Bhalchandra 2012). However, their example images
are simple case of MRI scans. However, on more complicated images
their proposed algorithm failed to generate good results. In contrast,
our proposed method shows promising results on more complicated
cases. Figure shows the results of both methods

Results :
Pre-processing of MR images:
This is gray scale image of Figure 2. which is converted into gray using
MATLAB algorithms.(Kekre, Sarode et al., 2010) this is done once image
is resized so that morphological operations can be performed later.
Morphological operation: Image acquired after applying top hat filter
kernel transformation which is used to highlight the sharp gradients at
peaks. This is what ideal watershed image for this MR image since we
used different algorithm that’s why we didn’t get the ideal result.
Watershed should show cells in the tumor areas of MR but in our case
it has shown cells of the whole brain image(Mustaqeem, Javed et al.,
2012).
Tumor can be detected in the image. By applying further MATLAB
algorithms we can detect the size of this tumor. Since the intensity of
tumor is much more than background of image that’s the only reason
tumor can be detected in MRI(Mancas, Gosselin et al. 2005).For now
we can locate the tumor in image precisely.

Future Work :
The above results have been obtained using different algorithms in
MATLAB as stated in methodology section. These results have been
obtained just to detect the exact location of tumor in the MRI images
of real data set of patients. There are number of alternate techniques
which can also be used for segmentation. All these techniques use
different algorithms in MATLAB for tumor detection(Padmavathi,
Muthukumar et al. 2010). There are number of future prospects in this
field. This work if elaborated further it can be used to detect tumor
size, stage and type using image processing techniques.

Core Stages:
Data Acquisition:
Acquire a dataset of MRI brain images, ideally including both healthy
and tumor-affected scans. This data might be available from public
sources or through collaborations with medical institutions (**Ensure
adherence to ethical guidelines and data privacy regulations).
Pre-processing:
Enhance the image quality for better analysis. This may involve: Noise
reduction techniques to remove artifacts. Contrast enhancement to
improve the distinction between brain tissues. Skull stripping to isolate
the brain region.
Segmentation:
This crucial step separates the brain tumor from the surrounding
healthy tissue. Common techniques in MATLAB include:
Thresholding: Classifies pixels based on intensity levels.
Edge detection: Identifies boundaries between regions.
Region-based segmentation: Groups pixels with similar characteristics.
Feature Extraction:
Extract quantitative information from the segmented tumor region.
This could involve:
Shape features (area, perimeter) Textural features (intensity
distribution)
Classification (Optional):
If your goal is to differentiate between tumor types (malignant vs.
benign), you can employ machine learning algorithms like:
Support Vector Machines (SVM): Classifies data points based on
hyperplanes.
Artificial Neural Networks (ANN): Learn complex patterns from data for
prediction.
CONCLUSIONS:
MRI images are best suitable for brain tumor detection. In this study
Digital Image Processing Techniques are important for brain tumor
detection MRI images. The pre-processing techniques include different
methods like Filtering, Contrast enhancement, Edge detection is used
for image smoothing. The pre-processed images are used for post
processing operation like threshold and Chen features are extracted
using GLCM and classification is done by Artificial Neural Network
where we have obtained 76.5% of accuracy when run on a dataset of
300 images in classifying tumor and non tumor images. This work help
in detection of tumor which in turn save the precious time of doctor
and pathologist to diagnose the tumor automatically in short span of
time.
Future Scope:
• In future, this technique can be developed to classify the tumors
based on feature extractions.
• This technique can be applied for ovarian, breast, lungs,skin, tumors.

DEADLINE:
The project willl be submit in 16weeks of this semester.