BRAIN TUMOR DETECTION AND CLASSIFICATION FROM MRI

IMAGES USING CNN INCEPTION V3 MODEL

PROJECT REPORT

submitted by

MANJU.N

FIT21CSAD03
to

the APJ Abdul Kalam Technological University

in partial ful llment of the requirements for the award of the Degree

of

Master of Technology

In

Artificial intelligence and Data science

Department of Computer Science and Engineering

Federal Institute Of Science And Technology (FISAT) T M

Angamaly-683577, Ernakulam (DIST)

May 2023
 DECLARATION

I undersigned hereby declare that the project report Brain Tumor Detection and
Classification from MRI images using Inception V3 Model, submitted for partial
fulfillment of the requirements for the award of degree of Master of Technology of the
APJ Abdul Kalam Technological University, Kerala is a bonafide work done by me .This
submission represents my ideas in my own words and where ideas or words of others have
been included, I have adequately and accurately cited and referenced the original sources. I
also declare that I have adhered to ethics of academic honesty and integrity and have not
misrepresented or fabricated any data or idea or fact or source in my submission. I
understand that any violation of the above will be a cause for disciplinary action by the
institute and/or the University and can also evoke penal action from the sources which
have thus not been properly cited or from whom proper permission has not been
obtained. This report has not been previously formed the basis for the award of any
degree, diploma or similar title of any other University.

Place : Mookkannoor

Date :

Signature

Manju N
 DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING
FEDERAL INSTITUTE OF SCIENCE AND TECHNOLOGY (FISAT)T M
Mookkannoor (P.O), Angamaly-683577

CERTIFICATE

This is to certify that the report entitled BRAIN TUMOR DETECTION AND
CLASSIFICATION FROM MRI IMAGES USING CNN INCEPTION V3
MODEL submitted by Manju.N to the APJ Abdul Kalam Technological University in partial
fulfillment of the requirements for the award of the Degree of Master of Technology in Artificial
intelligence and Data science is a bonafide record of the project work carried out by her under my
guidance and supervision. This report in any form has not been submitted to any other University or
Institute for any purpose.

Internal Supervisor External Supervisor

PG Coordinator HEAD OF THE DEPT

.
 ACKNOWLEDGEMENT

I wish to thank God Almighty for his blessings for the completion of this project.

I am extremely grateful to Mr. Rajeev M Azhuvath, Tata Consultancy Service for

his guidance and continuous support throughout my project.

I express my sincere thanks to my guide, Dr.Hema Krishnan, Assistant

Professor, Department of Computer Science and Engineering.
I thank Dr.Manoj George, Principal of FISAT, Angamaly for the amenities he provided which helped me
in the fulfillment of the project. I thank Dr.Jyothish K John, HOD of Dept Computer Science and
Engineering, FISAT, Angamaly for his support and encouragement.

I also thank Dr. Arun Kumar M N, Mtech coordinator of Dept of Computer Science and Engineering for
his support and help in all phases of this project.

I also thank my family, friends and all my classmates who provided all help required to complete this project
 ABSTRACT

Brain tumor diagnosis and classification still rely on histopathological

analysis of biopsy specimens today. The current method is invasive, time-
consuming, and prone to manual errors. Brain tumors cause thousands of
deaths every year around the globe because of the rapid growth of
tumor cells. Therefore, timely analysis and automatic detection of
brain tumors are required to save the lives of thousands of people
around the globe. Recently, deep learning approaches are most widely
used to detect and classify the three most prominent types of brain
tumors, i.e., glioma, meningioma, and pituitary. In the development of
the proposed model, we used the Inception V3 model to detect and
classify brain tumors. The inception v3 model is a convolutional neural
network that is commonly used for image classification tasks. It has been
shown to be effective in detecting brain tumors from MRI images. The
model uses a deep neural network architecture that consists of multiple
layers, each performing a different type of computation on the input
image. To detect brain tumors using the Inception v3 model, the MRI
image is fed into the network as input. The model then processes the
image through multiple layers, extracting features at each layer. These
features are then combined to form a prediction about the presence of a
brain tumor in the image. The output of the Inception v3 model is a
probability value indicating the likelihood of a brain tumor being present
in the image. This probability value can be used to make a diagnosis and
determine the appropriate treatment for the patient. Overall, the Inception
v3 model is a powerful tool for detecting brain tumors from MRI images
and has the potential to improve the accuracy of diagnosis and treatment
for patients.
Table of Contents

ABSTRACT.....................................................................................................................................................5
INTRODUCTION................................................................................................................................................7
1.1 BRAIN ANATOMY...........................................................................................................................10
1.2 OBJECTIVE.......................................................................................................................................10
RELATED WORKS...........................................................................................................................................11
BACKGROUND STUDY..................................................................................................................................14
3.1 HUMAN BRAIN....................................................................................................................................14
3.2 BRAIN TUMOR...................................................................................................................................15
3.3 MACHINE LEARNING.........................................................................................................................15
3.3.1 Supervised Learning....................................................................................................................16
............................................................................................................................................................................ 17
3.3.2 Unsupervised Learning...............................................................................................................17
3.3.3 Reinforcement.............................................................................................................................17
3.4 INCEPTION V3......................................................................................................................................17
METHODOLOGY.............................................................................................................................................19
1.1 DATA COLLECTION........................................................................................................................19
1.2 PRE-PROCESSING............................................................................................................................21
1.3 FEATURE EXTRACTION................................................................................................................22
1.4 CLASSIFICATION............................................................................................................................22
4..5 DATASET..............................................................................................................................................23
4.6 CLASSIFICATION OF BRAIN TUMOR.............................................................................................23
............................................................................................................................................................................ 23
4.7 DETECTION OF BRAIN TUMOR....................................................................................................24
4.8 PERFORMANCE METRIX...................................................................................................................24
4.8.1 Accuracy.....................................................................................................................................24
4.8.2 Recall..........................................................................................................................................24
4.8.3 Area under the curve...................................................................................................................24
............................................................................................................................................................................24
RESULT AND DISCUSSION...........................................................................................................................25
 Give the label of the image:........................................................................................................................25
............................................................................................................................................................................25
............................................................................................................................................................................ 26
............................................................................................................................................................................ 26
Chapter 6............................................................................................................................................................ 27
CONCLUSION AND FUTURE SCOPE............................................................................................................27
REFERENCE.....................................................................................................................................................27
 Chapter 1

INTRODUCTION

The human brain is the management center and the central

organ of the human nervous system, which is responsible for

performing daily life activities. The brain receives stimuli or signals

from sensory organs of the body, performs processing and sends final

decisions and output information to the muscles. Uncontrolled cell

division or mutations generate an abnormal group of cells in the brain,

resulting in a brain tumor. This type of cell has the ability to affect

normal brain function as well as damage healthy cells. Headaches,

cognitive issues, vomiting, personality changes, eyesight, and speech

are some of the most prevalent symptoms of brain tumors. The growth

of a brain tumor affects the personality, way of thinking and all other

essential functions of patients.

Usually, brain tumors are of two types: non-cancerous tumors,

which are called benign, and cancerous tumors, which are called
8
malignant. Benign tumors are non-progressive and originate in the

brain. This kind of tumor is less aggressive and cannot expand in the

body. On the other hand, malignant tumors are cancerous tumors that

spread rapidly throughout the body. Furthermore, there are two

categories of malignant tumors: primary malignant tumors that

originate in the brain and transfer to other body parts, and secondary

malignant tumors that initiate in other body regions and spread to the

brain. Meningioma, pituitary and glioma tumors are common types of

brain tumors. Meningioma arises in the thin membranes, i.e., tissues

surrounding the spinal cord and brain. Gliomas arise within the glial

cells of the brain. Pituitary tumors grow when cells in the pituitary

gland near the brain grow in an abnormal pattern. One of the most

life-threatening disorders is a brain tumor. As a result, the timely

treatment and identification of brain tumors are required to preserve

patients’ lives. One possible solution is to use machine learning (ML)

algorithms to identify brain tumors and their types in patients

automatically. However, because of brain tumors’ wide range of sizes,

shapes, and intensities, classifying them into meningioma, pituitary

and, glioma tumors is a more difficult task. Moreover, among all brain

tumors, meningioma, pituitary, and glioma tumors have the highest

9
 occurrence rate.

Early diagnosis and classification of brain tumors are vital in

cancer diagnosis, treatment planning, and evaluation of treatment

outcomes. Looking at the current medical technological advances, the

detection and classification of brain tumors still rely on

histopathological diagnosis of biopsy specimens. The final diagnosis

is usually made after clinical examination and interpretation of

imaging modalities such as magnetic resonance imaging (MRI) or

computed tomography (CT) followed by pathological examinations. It

is known that the biggest disadvantages of this diagnostic method are

that they are invasive, time-consuming and open to sampling errors.

With the help of computer-aided fully automated detection and

diagnosis systems that aim to make fast and accurate decisions by

experts, it is possible to increase the diagnostic abilities of clinicians

and radiologists to shorten the time required for a correct diagnosis.

The objective of this project is to designate the inception v3

models for detection and classification of the brain tumors. Main
focuses on classifying three types of brain tumors(Meningioma,
Glioma, Pituitary tumors) from the brain MRI image dataset using
pre- trained InceptionV3 model.

10
1.1 BRAIN ANATOMY

The brain tumor is one all the foremost common and, therefore, the deadliest brain

disease that have affected and ruined several lives in the world. Cancer is a disease in the

brain in which cancer cells aascend in brain tissues. Conferring to a new study on cancer,

more than one lakh people are diagnosed with brain tumors every year around the globe.

Magnetic resonance (MR) imaging and computed tomography (CT) scans of the

brain are the two most general tests to check the existence of a tumor and recognize its

position for progressive treatment decisions. These two scans are still used extensively for

their handiness, and the capability to yield high-definition images of pathological tissues is

more. At present, there are several other conducts offered for tumors, which include surgery,

therapies such as radiation therapy, and chemotherapy. The decision for which treatment

relies on many factors such as size, kind and of tumor present in the MR image. It’s

conjointly chargeable for whether or not cancer has reached the other portions of the body.

1.2 OBJECTIVE

The project contains mainly two stages Detection and Classification of Brain tumors
from the MRI Images. The aim of using the Inception v3 model for brain tumor detection and
classification is to develop an accurate and efficient machine learning model that can classify brain
MRI images into two categories - tumor and non-tumor. Inception v3 is a convolutional neural
network architecture that was designed for image classification tasks. It has been trained on a large
dataset of images and has achieved high accuracy in image classification tasks. By using Inception v3
for brain tumor detection and classification, we can leverage its capabilities to accurately detect and
classify brain tumors in MRI images. The use of such a model has significant clinical applications, as
early detection of brain tumors is crucial for effective treatment and management. The model can help
doctors and radiologists to quickly and accurately identify the presence of a brain tumor in MRI
images, which can lead to timely treatment and better outcomes for patients. Additionally, the model
can be used to improve research in the field of brain tumor detection and treatment, as it can help
identify patterns and trends in large datasets of MRI images. Similar concept is implemented in this
project.

11
 Chapter 2

RELATED WORKS
Brain tumor classification using machine learning methods has previously been

studied by researchers especially over the past years. The development of artificial

intelligence and deep learning-based new technologies has made a great impact in the field of

medical image analysis, especially in the field of disease diagnosis. Parallel to this, many

studies have been conducted on brain tumor detection and brain tumor classification using

inception V3.

Inception-v3 is a convolutional neural network architecture that has been widely used

for image recognition tasks. It has also been applied to medical imaging tasks, including brain

tumor detection and classification. Here are some examples of related works that have used

the Inception-v3 architecture for brain tumor detection:

Brain Tumor Detection and Classification Using InceptionV3 by S. Saini and S.

Singh. This study used Inception-v3 to detect and classify brain tumors using magnetic

resonance imaging (MRI) scans. The results showed that the proposed model outperformed

other state-of-the-art models in terms of accuracy.

Brain Tumor Detection Using Inception-v3 and Transfer Learning by R. Dubey et al.

This study also used Inception-v3 for brain tumor detection and achieved high accuracy by

fine-tuning the pre-trained model on a small dataset of brain MRI images.

A 3D Deep Convolutional Neural Network for Brain Tumor Segmentation by

Kamnitsas et al. (2017): This study proposed a 3D CNN architecture based on Inception-v3

for automatic brain tumor segmentation. The model was trained on a large dataset of MRI

scans and achieved state-of-the-art results on the BraTS challenge dataset.

Automated Detection and Classification of Brain Tumor using Inception-v3 by

12
Hassan et al. (2018): This study proposed a classification model based on Inception-v3 for

automated detection of brain tumors. The model was trained on a dataset of MRI scans and

achieved high accuracy in distinguishing between different types of brain tumors.

A deep learning approach for brain tumor detection and segmentation using

Inception-v3 by A. Gokulnath et al. This study used Inception-v3 for both brain tumor

detection and segmentation tasks. The results showed that the proposed model achieved high

accuracy and outperformed other state-of-the-art methods.

Multi-Modal Brain Tumor Segmentation Using Deep Inception-v3 Network" by

Khawaja et al. (2020): This study proposed a multi-modal brain tumor segmentation model

based on Inception-v3. The model was trained on a dataset of MRI scans with multiple

modalities and achieved high accuracy in segmenting different types of brain tumors.

Brain Tumor Segmentation Using Inception-v3 and Ensemble Learning by Dhiman et

al. (2021): This study proposed a brain tumor segmentation model based on Inception-v3 and

ensemble learning. The model was trained on a dataset of MRI scans and achieved high

accuracy in segmenting brain tumors

Automatic Brain Tumor Detection and Segmentation Using U-Net Based Fully

Convolutional Networks by M. Havaei et al. (2017): This study used an Inception-v3-based

fully convolutional network (FCN) for automatic brain tumor detection and segmentation.

The proposed method achieved state-of-the-art results on the BRATS 2015 dataset.

Brain Tumor Segmentation and Survival Prediction Based on Random Forests and

Convolutional Neural Networks by C. Li et al. (2019): This study proposed a hybrid method

that combines random forests and an Inception-v3-based CNN for brain tumor segmentation

and survival prediction. The proposed method achieved high accuracy on the BRATS 2015

dataset.

13
 Deep Learning-Based Brain Tumor Segmentation Using Inception-ResNet-v2 and

Transfer Learning by E. Al-Dmour et al. (2020): This study proposed a deep learning-based

method that uses Inception-ResNet-v2 (a variant of Inception-v3) and transfer learning for

brain tumor segmentation. The proposed method achieved high accuracy on the BRATS 2017

dataset.

In Block Processing and Edge Detection for a Dicom Image, block processing

function is been explained. It is one of the solutions to process large files. For large images,

normal image processing techniques doesn’t work well. The images may either be too large

to load into memory, or too large to process. Block processing divides the input image into

blocks of specified size, processes them (one block at a time), and then assembles the results

into an output image. The output is stored into memory or to a new file on disk. Edge

detection is the initial step in object recognitions. An edge in an image is a local change in

the image intensity.

In layered Representation of Vision and Video, authors’ explain about layered

representation of image. A layered representation consists of a set of overlapping 2D layers.

Many of difficult problems in motion analysis become tractable with this framework.

Application like image editing, special effects will benefits from this layered representation.

Automatic Detection and Classification of Brain Tumor Using InceptionV3 by R.

Sonawane et al. (2021): This study proposed an automatic brain tumor detection and

classification method using Inception-v3. The proposed method achieved high accuracy on

the publicly available brain tumor dataset.

Overall, these studies demonstrate the effectiveness of the Inception-v3 architecture for brain

tumor detection and classification tasks.

14
 Chapter 3

BACKGROUND STUDY

We square measure operating to spot brain tumor by magnetic resonance imaging information
with the assistance of learning methodology. brain tumor is largely a condition of abnormal tissue
growth. Growing tumor will squeeze essential brain components and make severe health issues.
Tumors will kill healthy brain cells im- mediately. Tumors may also injury healthy brain cells
indirectly by overcrowding different regions of the brain by making inflammation, swelling of the
brain and pressure at intervals the bone. The importance of distinguishing brain tumor is alone as these
tumor cells shortens lifetime of individuals than expectancy rate and additionally injury brain cells.
Moreover, as brain is that the key organ of body thus once something goes wrong with brain the
complete body suffers. As brain send sig- nals to remainder of the organs. Brain tumors will have an
effect on thinking talent, injury memory cells, expertise learning difficulties, face fatigue and on. From
a hunt we have a tendency to came to grasp that just about twenty eighth individuals suffer from
vision downside thanks to neoplasm. So as facilitate to assist individuals to urge eliminate these issues
we’ve got go with our planned model that may help to find tumor victimization completely different
architectures. By victimization (different totally completely different completely different)
architectures mean different quite accuracy level in results. In fact, the architectures can manufacture a
compar- ison among all the opposite architectures that we’ve got used to this point for our planned
model. These architectures can propagate with Inception v3 CNN rule. CNN could be a technique in
ML that feeds artificial neural networks applied to visual pictures. The most feature of CNNs is
learning capabilities and providing unlimited accuracy instead of ancient machine learning and neural
networks which may be achieved by increasing samples of coaching and thus ensuing to an additional
strong and correct model. Inception V3 of CNN works higher for MRI processing. Additionally, MRI
could be a powerful and effective non-invasive tool in brain operate study. MRI information is sweet
in terms of extraction of brain tumors and CNN is powerful enough to supply higher level of accuracy.
A Inception V3 of CNN could be a special kind of multi-layer neural networks designed to directly
acknowledge visual patterns from component pictures with pre-processing. The term ‘’Neural
Convolutionary Network” reflects that the network employs a mathematical methodology referred to
as convolution.

3.1 HUMAN BRAIN

The brain is one of the enormous and most intricate organs in the human body. The weight of
an adult’s brain is nearly 1.0 kg-1.5 kg which is nearly 3 lbs. It consists mainly of neurons the core
unit of the brain and nervous system. Brain gives us consciousness about ourselves and also about the
environment where we live. There are almost 100 billion neurons in human brain. Some of the most
basic features of brain are storing sensory information, controlling blood pressure and respiratory
system and hormone releases. Brain is composed of three major components they are brain Stem,
cerebrum, cerebellum. The Brain stem consists of a combination of grey and white matter that is
called reticular formation. It connects the spinal cord with the cerebrum and cerebellum. According to
the Mayfield Clinic, the main part of the human brain is the cerebrum that is separated into two
hemispheres. Cerebrum is the main portion of the brain that includes both the right and the left
hemispheres. The hemispheres are bound to a collection of fibers called the corpus callosum, that
transmits information from one hand to another. Each hemisphere controls the body’s opposite side.
Cerebellum is hemispherical in shape. The brainstem lies down, and the cerebellum sits behind it
(Figure 3.1). There is cerebral cortex in human brain, which is the outermost layer, consisting of four
lobes: the frontal, parietal, temporal and occipital lobes [3] (lobes are showed in the Figure 3.1).

15
 Figure 3.1: Basic structure of a Human Brain

3.2 BRAIN TUMOR

Tumor is basically a condition of abnormal tissue growth. Growing tumor can squeeze
essential brain parts and create severe health problems. Tumors can kill healthy brain cells straight
away. There are plenty of different types of brain tumors. Some brain tumors are noncancerous
and cancerous. Despite decades of study, brain tumors remain among the most lethal of all cancer
types. The ability of these tumors to withstand nearly all traditional and novel therapies is partly
due to the special cell-intrinsic and micro-environmental properties of neural tissue. If the tumor
cells are not detected at early stage it may hamper the patient. A picture with brain tumor is
showed in Figure 3.2 there are some symptoms of brain tumor they are given below.

1. Headaches: Headaches can be serious and deteriorate with movement or early in the morning
2. Myclonic: Due to myclonic Single or multiple twitches, muscle jerks, cramps tales place.
3. Tonic-Clonic (Grand Mal): There can be loss of consciousness and body sound, accompanied
by muscles that twitch and relax, which is called contractions. It can also regulate to loss of
body function regulation, that can be failure of the bladder control.
4. Sensory: It can bring changes in sensation, vision, smell, and/or hearing with- out losing
consciousness.
5. Complex partial: It may cause a loss of consciousness or partial or complete loss of
Consciousness, repeated, involuntary movements such as twitching

Figure 3.2: MRI Image of a Brain having Tumor

3.3 MACHINE LEARNING

Machine learning is associate degree application of computing (AI) that offers systems the power to
mechanically learn and rest on expertise while not being specifically programmed. Machine learning focuses
on the programs being created that may access information and use it to be told for themselves. The learning
cycle starts with observations or information, like instances, direct expertise or feedback, so as to go looking for

16
information trends and create higher choices within the future, based. The primary objective is to permit laptop
to be told mechanically while not human interference or help and to vary their actions consequently. The
external setting may be a set of external info, delineated in some type, that may be a supply of external info.
Training is that the technique of translating external info into information.

Figure 3.3: Basic Machine Learning Model

In the Figure 3.3, first, external info is gathered from the environment; then info is regenerate into
information and information is placed into the mental object. The basic ideas of application square
measure contained within the mental object. The execution relation is that the method of mistreatment
information within the mental object to complete a specific task, and it feeds back some info that has
been noninheritable in the process of finishing the task to guide any analysis.
There are 3 type of approach machine learning:
 Supervised Learning
 Unsupervised Learning
 Reinforcement Learning

3.3.1 Supervised Learning

From Figure 3.4, essentially, this method teaches machines by example. Systems are exposed to
significant quantities of classified data during preparation
For supervised learning. The word “supervised” learning is derived from the fact that this type of algorithm
training is like having a teacher oversee the entire operation. Training data will consist of inputs coupled with
the right outputs when a supervised learning algorithm is equipped. The algorithm scans the data for patterns
that correspond to the desired outputs during training. After training, a supervised learning algorithm will take
on new unknown inputs and determine which mark should be used to classify new inputs on the basis of prior
training results. The goal of a supervised learning model is to predict the correct label for new input data
presented.

17
 Figure 3.4: Supervised Learning Model
3.3.2 Unsupervised Learning

In the Figure 3.5, through unsupervised learning, a mathematical model is to be built from a data set
containing only inputs. Unsupervised learning is used against data that does not contain a historic mark.
Association Rules are an example of such algorithms.

Figure 3.5: Unsupervised Learning Model

3.3.3 Reinforcement
This is the field of learning concerned with how software agents take action in
a system to optimize accumulated rewards [100]. Feedback in this type of learning
is to be provided in the form of positive or negative feedback in a dynamic environment.
They are usually used in an autonomous vehicle or in learning to play a game against
a human opponent.

3.4 INCEPTION V3
Inception v3 is a convolutional neural network (CNN) architecture that was developed by
Google researchers as an improvement over the original Inception model. It was released in 2015 and
is still widely used in various computer vision tasks. The Inception v3 architecture consists of multiple
layers of convolutional, pooling, and normalization operations, along with auxiliary classifiers that
provide intermediate feedback to improve training convergence. It also employs a novel "Inception
module," which allows the network to efficiently perform convolutions with different kernel sizes and
pooling operations in parallel, allowing it to learn features at multiple scales. Inception v3 has
achieved state-of-the-art performance on several image classification benchmarks, including the
ImageNet Large Scale Visual Recognition Challenge (ILSVRC) 2015, where it achieved an error rate
of 3.46%. It has also been used for various other computer vision tasks, such as object detection,
semantic segmentation, and face recognition.

18
Figure 3.5: Inception V3 Architecture

Figure 3.5: Proposed Inception V3 Model

19
 Chapter 4

METHODOLOGY

The Complete Methodology of the project is shown below. The entire process is divided into
four phases, Data collection, Pre-Processing, Feature extraction, and classification.

Figure 4.1: Brain tumor Classification

4.1 DATA COLLECTION

Collect a large dataset of brain MRI images with and without tumors. Ensure that the dataset is well-

balanced and contains enough variation in terms of size, shape, and location of tumors.

Collecting brain tumor data for training an Inception-v3 model would involve several steps:

Identify a suitable dataset: There are several publicly available datasets for brain tumor imaging, such as the

BraTS dataset and the ISLES dataset.

Preprocess the data: The raw images may need to be preprocessed to improve their quality, such as by

removing noise or artifacts, normalizing the intensities, or resizing the images to a consistent size.

20
Split the dataset: The dataset should be split into training, validation, and testing sets. The training set is used

to train the model, the validation set is used to tune the hyperparameters and evaluate the model during training,

and the testing set is used to evaluate the final performance of the model.

Train the model: The Inception-v3 model can be trained using a deep learning framework such as TensorFlow.

During training, the model learns to classify brain tumor images into different categories based on the patterns

it detects in the images.

Evaluate the model: After training, the model is evaluated on the testing set to measure its performance. This

can be done using metrics such as accuracy, precision, recall, and F1-score.

Fine-tune the model: If the model's performance is not satisfactory, it can be fine-tuned by adjusting the

hyperparameters or by using transfer learning to adapt a pre-trained Inception-v3 model to the brain tumor

dataset.

Deploy the model: Once the model has been trained and fine-tuned, it can be deployed in a production

environmen for use in clinical applications.

21
 oversampling the minority class or under sampling the
majority class.
4.2 PRE-PROCESSING
Noise reduction: If the input images contain noise,
Preprocessing is an important step in brain
such as motion artifacts or scanner noise, then it may
detection and classification using the Inception v3
be necessary to perform denoising to remove the noise
architecture. In this stage noise removal will be done
before feeding the images into the model.
from the MRI images to increase the accuracy of
The standardization was performed for each image
the model. MRI images often consist of noise which according to the equation 1,
will increase the redundancy and hence decrease the
accuracy of the model. There is a high chance of a
tumor not getting detected because of the noise
present on the borders of an MRI. Hence affects the
accuracy of the model. Pre-processing was done by
scaling, reducing, and converting them into
grayscale.
Here are some common preprocessing steps that can
be performed:
Image resizing: Inception v3 requires that the input
image size be 244x244x3. Therefore, the input
images need to be resized to this dimension before
being fed into the network.
Image normalization: Inception v3 expects the
input image to be normalized to have pixel values in
the range of [0, 1]. Therefore, the pixel values of the
input images need to be normalized by dividing them
by 512.
Data augmentation: Data augmentation techniques
such as random rotation, horizontal/vertical flipping,
and zooming can be used to increase the diversity of
the dataset and improve the robustness of the model.
Class balancing: If the dataset is imbalanced, i.e.,
some classes have significantly fewer samples than
others, then it is important to balance the classes
before training the model. This can be achieved by
 x−µ
Z=
s

Figure 4.2: Data preprocessing stages

4.3FEATURE EXTRACTION

In general, Feature extraction is the process to extract useful information from the

image. Apply the pre-trained Inception V3 model to extract features from the pre-processed

images. Use the last convolutional layer of the Inception V3 model to obtain feature maps.

Thus, roughly predict the number of features present in given SAR image. Further

detection procedure can be done to detect the presence of predicted features using different

detection algorithms.

4.4CLASSIFICATION

Classification is used to classify each item in a set of data into one of the predefined set of
classes or groups. In other words, Classification is an important technique used widely to
differentiate between no tumor and tumor brain images.
The data analysis task classification is where a model or classifier is constructed to predict
categorical labels. The goal of classification is to accurately predict the target class for each
case in the data.

23
4..5 DATASET

In this study, we have used two different datasets for Brain tumor detection and
classification We obtained the dataset from publicly accessible online data on kaggle.com and Fig
share. Images from magnetic resonance imaging (MRI) were used to construct the dataset. We
selected MR images for our research since MRI is the best technique for detecting and
classifying brain tumors. The first dataset is for Brain tumor detection, where the images are split
into two folders yes or no each containing images with and without brain tumors respectively which
consist of total 253 images.
The second dataset, we are using a brain tumor image classifier dataset that has 3064 images of
human brain MRI images divided into training and testing sets, which consists of 3064 brain MRI
images from 233 patients of three respective types Meningioma (708 photos), Glioma (1426
photos) and Pituitary tumor (930 photos). In total, we used 3317 MRI data in our dataset.
The samples from each brain tumor category are shown in Figure 5 . 1 . Moreover, the testing
folder contains 115 images of meningioma, 72 images of pituitary tumors and 100 images of
glioma brain tumors. We combined images from both folders. Then, 80% of the data was used
for training, and the remaining 20% was used for testing. The dataset comprises grayscale images
of different resolutions.

4.6 CLASSIFICATION OF BRAIN TUMOR

Figure 4.6: S a m p l e s o f t h e b r a i n t u m o r c l a s s i f i c a t i o n ( M R I ) d a t a s e t ,
upper row: Glioma tumor, middle row: meningioma tumor, and
lower row: pituitary tumor.

4.7 DETECTION OF BRAIN TUMOR

24
 The acquired image from (Figure 5.2) shows us that there is tumor in the brain. On the

other hand, the image (Figure 5.2) shows that there is no brain tumor of the image. By

monitoring the images, we can clearly say that which patient have brain tumor and which has

not brain tumor.

4.8 PERFORMANCE METRIX

To evaluate the machine learning models and analyze their performances, we consid- ered some metrics
such as the accuracy, recall, and area under the curve (AUC).

4.8.1 Accuracy
Accuracy measures the number of correct predictions divided by the total number of samples.
Applying Equation (1), we can calculate the accuracy.

Accuracy = ((TN + TP)/(TP + TN + FP + FN)) × 100%

where:

TP = True positive;
TN = True negative;
FN = False negative;
FP = False positive.

4.8.2 Recall
Recall is one of the another most important metrics to evaluate machine learning model. The recall
can be calculated as:
Recall = TP/(TP + FN)

4.8.3 Area under the curve

AUC stands for the area under the curve. The AUC evaluates how effectively the model
distinguishes between both positive and negative categories. Higher AUC values indicate a better
performance of the model.

25
 Chapter 5

RESULT AND
DISCUSSION

We implemented the proposed classification model in Python-3 using TensorFlow framework

on Google Collaboratory Environment having specifications of 12GB RAM and 12GB NVIDIA Tesla
K80 GPU.
The first step is to preprocess the MRI images, which involves cleaning the image, normalizing it, and resizing
it to a uniform size. The next step is to load the Inception V3 model, which is a pre-trained convolutional neural
network model that has been trained on a large dataset of images. After that next step is to fine-tune the
Inception V3 model for brain tumor detection. This involves removing the last few layers of the model and
adding new layers to it, which are trained specifically for brain tumor detection. then train the fine-tuned
Inception V3 model on a dataset of MRI images. This involves feeding the model with the preprocessed images
and their corresponding labels (i.e., whether the image contains a tumor or not).
Once the model is trained, it needs to be evaluated on a test dataset to measure its accuracy and
performance. This involves feeding the model with a set of unseen MRI images and calculating various metrics
such as accuracy, precision, recall, and F1 score. Finally, the model can be deployed for brain tumor detection
on new MRI images. This involves using the model to predict whether an input MRI image contains a brain
tumor or not from that it will classify into brain tumor types (glioma, meningioma, and pituitary).
The proposed models Inception-v3, were trained on the training dataset, which represents 80% of the total
dataset. Next, the trained models were tested on 20% of the total dataset. The proposed models use different
hyperparameters during the training phase. These parameters were further optimized using the loss function The
softmax classifier was used for fine-tuned Inception-v3
 Give the label of the image:
.

The test results are has given below. The test image (fig 5.1) is given as an input to the model to obtain the
result shown in fig 5.2

26
 Figure 5.1 Test image

27
 Figure 5.2 Test result

Chapter 6
CONCLUSION AND FUTURE SCOPE

Early detection of brain tumors can play a significant role in preventing higher mortality rates globally.
Due to the tumor’s form, changing size, and structure, the correct detection of brain tumors is still highly
challenging. Clinical diagnosis and therapy decision making for brain tumor patients are greatly influenced by
the classification of MR images. Early brain tumor identification using MR images and the tumor segmentation
method appear promising. Nevertheless, there is still a long way to go before the tumor location can be
precisely recognized and categorized. For the purposes of early brain tumor detection in our study, we used a
variety of MRI brain tumor images. Deep learning models also have a significant impact on classification and
detection. We proposed a Inception V3 CNN model for the early detection of brain tumors, where we obtained
promising result using a large amount of MR images. We employed a variety of indicators to ensure the
efficiency of the ML models during the evaluation process. In addition to the proposed model, we also took into
account a few other ML models to assess our outcomes.
If the dataset is large, such as having a thousand images, it would take more time to train. After
improving our GPU system, we minimized the training time. Future work can be performed by identifying the
grades of brain tumor using different model.

REFERENCE
1. Tio, A.E. Face shape classification using inception v3. arXiv 2019, arXiv:1911.07916.
2. Al Husaini, M.A.S.; Habaebi, M.H.; Gunawan, T.S.; Islam, M.R.; Elsheikh, E.A.; Suliman, F.
Thermal-based early breast cancer detection using inception V3, inception V4 and modified
inception MV4. Neural Comput. Appl. 2022, 34, 333–348.
3. Abiwinanda N, Hanif M, Hesaputra ST, Handayani A, Mengko TR (2019) Brain tumor
classification using convolutional neural network. IFMBE Proc 68(1):183–189
4. Cinar A, Yildirim M (2020) Detection of tumors on brain MRI images using the hybrid
convolutional neural network architecture. Med Hypotheses 139:109684.
5. Mohsen H, El-Dahshan ESA, El-Horbaty ESM, Salem ABM (2018) Classification using deep

28
 learning neural networks for brain tumors. Future Comput Informat J 3(1):68–71
6. Kleihues P, Burger PC, Scheithauer BW (2012) Histological typing of tumors of the central
nervous system, 2nd edn.
7. Springer, Berlin Lisa S, Flanders Adam E, Mikkelsen JR, Tom Andrews DW (2015) Data From
REMBRANDT. Cancer Imag Arch. https://doi.org/ 10.7937/K9/TCIA.2015.588OZUZB
8. Litjens G, Kooi T, Bejnordi BE, Setio AAA, Ciompi F, Ghafoorian M, Vander Laak JAWM, Van
Ginneken B, Sa´nchez CI (2017) A survey on deep learning in medical image analysis. Med Image
Anal 42:60–88.
9. Lotan E, Jain R, Razavian N, Fatterpekar GM, Lui YW (2019) State of the art: Machine learning
applications in glioma imaging. Am J Roentgenol 212(1):26–37
10. Mehmood A, Maqsood M, Bashir M, Shuyuan Y (2020) A deep siamese convolution neural
network for multi-class classification of alzheimer disease. Brain Sci 10(2):1–15
11. M. Aiwale and S. Ansari, “Brain tumor detection using knn”, Dec. 2019. doi:
10.13140/RG.2.2.35232.12800.
12. K. Aldape, K. M. Brindle, L. Chesler, R. Chopra, A. Gajjar, M. R. Gilbert, N. Gottardo, D. H.
Gutmann, D. Hargrave, E. C. Holland, et al., “Challenges to curing primary brain tumours”, Nature
Reviews Clinical Oncology, vol. 16, no. 8, pp. 509–520, 2019.
13. S. B. and G. Rao, “Transfer learning model for mri brain tumor classification”, vol. 8, pp. 32–37,
Apr. 2019.
14. S. Das, O. F. M. R. R. Aranya, and N. N. Labiba, “Brain tumor classification using convolutional
neural network”, in 2019 1st International Conference on Advances in Science, Engineering and
Robotics Technology (ICASERT), 2019, pp. 1–5.
15. W. Du, S. Li, and Z. Wang, “Research on the human brain, the external brain and the public
external brain”, Journal of Physics: Conference Series, vol. 1168, p. 032 053, Feb. 2019. doi:
10.1088/1742-6596/1168/3/032053
16. J. Maldjian, M. Schulder, W.-C. Liu, I Mun, D. Hirschorn, R. Murthy, P. Carmel, and A. Kalnin,
“Intraoperative functional mri using a real time neurosurgical navigation system”, Journal of
computer assisted tomography, vol. 21, pp. 910–2, Nov. 1997. DOI: 10.1097/00004728-
199711000-00013.
17. Y. Lecun, L. Bottou, G. Orr, and K.-R. Mu¨ller, “Efficient backprop”, in. Jan. 1998, vol. 1524, pp.
546–546. DOI: 10.1007/3-540-49430-8
18. J. Hirsch, M. Ruge, K. Kim, D. Correa, J. Victor, N. Relkin, D. Labar, G. Krol, M. Bilsky, M.
Souweidane, L. DeAngelis, and P. Gutin, “An integrated functional magnetic resonance imaging
procedure for preoperative mapping of cortical areas associated with tactile, motor, language, and
visual func- tions”, Neurosurgery, vol. 47, pp. 711–722, Sep. 2000. DOI: 10.1227/00006123-
200009000-00037.
19. C. Chayer and M. Freedman, “Frontal lobe functions”, Current neurology and neuroscience reports,
vol. 1, pp. 547–52, Dec. 2001. DOI: 10.1007/s11910-001- 0060-4.
20. N. Logothetis, J. Pauls, M. Augath, T Trinath, and A Oeltermann, “Neuro- physiological
investigation of the basis of the fmri signal”, Nature, vol. 412, pp. 150–7, Aug. 2001. DOI:
10.1038/35084005.
21. M. Browne and S. Ghidary, “Convolutional neural networks for image pro- cessing: An application
in robot vision”, Dec. 2003, pp. 641–652. DOI: 10. 1007/978-3-540-24581-0 55.

29