DHANALAKSHMI SRINIVASAN

COLLEGE OF ENGINEERING AND TECHNOLOGY

DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING

PROJECT PRESENTATION

NAME OF THE STUDENTS: REG.NO.

ANANT PRAKASH 310520104016
MD NIYAMUL ALAM 310520104064
MD MINHAZ AHMED 310520104063
DEEPAK KUMAR 310520104124
GUIDED BY-
Dr. K. John Peter(HOD CSE)

5-10-2024
DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING 1
 DHANALAKSHMI SRINIVASAN
COLLEGE OF ENGINEERING AND TECHNOLOGY

DEEP LEARNING
DOMAIN:
TITLE: BRAIN TUMOR DETECTION AND ANALYSIS USING DEEP LEARNING

10/02/2024 DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING 2

 ABSTRACT
• The human brain, a vital controller of the humanoid system, can be affected by abnormal cell
growth leading to tumors and, ultimately, brain cancer.
• In the realm of healthcare, Computer Vision enhances accuracy, particularly in imaging methods
like CT scans, X-rays, and MRI.
• Our paper focuses on utilizing various techniques for brain cancer detection through MRI.
• - We employ bilateral filtering for pre-processing to eliminate noise.
• - Followed by binary thresholding and Convolutional Neural Network (CNN) segmentation for precise tumor detection.
• - Utilizing training, testing, and validation datasets, our machine predicts the presence of a brain tumor.
• Performance metrics, including accuracy, sensitivity, and specificity, assess the outcomes.
• We anticipate our proposed approach to outperform existing methods.
• Our system provides comprehensive analysis, covering probability assessment, tumor size, location,
type, and growth rate estimation, empowering clinicians with precise diagnostic insights.

10/02/2024 DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING 3

 OBJECTIVE
The objective of our study is to develop an advanced approach for the
detection of brain tumor and analysis using Deep Learning. This involves:

 Advance the capabilities of Convolutional Neural Networks(CNNs) for more accurate

and efficient brain tumor detection and analysis.
 Utilize extensive datasets encompassing diverse imaging modalities and patient
demographics to train the model.
 Implement advancement in image processing techniques to improve the quality
and clarity of medical imaging data.
 Develop algorithms to extract and analyze additional tumor features, such as
probability assessment, tumor size, location, type, and growth rate estimation to enhance diagnostic
accuracy.
 Integrate machine learning techniques for predictive modeling of tumor growth rate and treatment
response.

10/02/2024 DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING 4

 EXISTING SYSTEM
The existing system for brain cancer detection utilizes similar techniques but often
exhibits lower accuracy and less features due to several limitations:
Limited image processing : The existing system may employ basic image
processing techniques with fewer layers, resulting in reduced ability to extract relevant
features from MRI scans accurately.

Lower accuracy rates: Due to the simplistic nature of the image processing
pipeline and lower dataset, the existing system may suffer from lower accuracy rates
in tumor detection. This can lead to false positives or false negatives, potentially
impacting patient outcomes and treatment decisions.

Reliance on manual interpretation: The reliance on manual interpretation by

radiologists in the existing system introduces subjectivity and variability in diagnoses.
This manual approach is prone to human error and may not fully leverage the potential
of advanced computational methods.
10/02/2024 DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING 5
 EXISTING SYSTEM
Lack of advanced machine learning algorithms: The existing system may
lack the integration of advanced machine learning algorithms, such as Convolutional
Neural Networks (CNNs), which have shown significant promise in improving the
accuracy of brain tumor detection by automatically learning discriminative features
from MRI scans.

Lack of comprehensive features: Existing systems may lack the ability to

provide detailed information about tumor characteristics such as size, location, type,
and growth rate, limiting clinicians’ understanding of the patients’ condition.
Overall, while the existing system shares similarities with our proposed approach, its
limitations in terms of image processing, accuracy and features underscore the
need for more sophisticated and advanced techniques to enhance the reliability and
effectiveness of brain tumor detection and analysis.

10/02/2024 DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING 6

 PROPOSED SYSTEM
Setting Ourselves Apart: Advancements in Brain Cancer Detection And Analysis Compared to
Current Methods

 Advanced Image Processing: Our project employs a sophisticated image processing pipeline with
multiple layers, enhancing feature extraction from MRI scans compared to the existing system's simpler
techniques.
 Higher Accuracy: Utilizing binary thresholding and CNN segmentation, our project achieves superior
accuracy in brain tumor detection, minimizing false positives and negatives, unlike the existing system.
 Tumor Analysis: Our system provides comprehensive analysis, covering probability assessment, tumor
size, location, type, and growth rate estimation, empowering clinicians with precise diagnostic insights.

10/02/2024 DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING 7

 PROPOSED SYSTEM
 Comprehensive Dataset Utilization: Our project utilizes extensive datasets for robust
performance evaluation, enabling accurate prediction of brain tumors, a feature lacking in the
existing system.

 Anticipated Performance Improvement: With advanced techniques, our project aims to

outperform the existing system in accuracy and sensitivity, promising enhanced brain cancer
diagnosis and patient care.

 Simple Output: a Our system produces straightforward results, categorizing findings as

either “Yes” or “No” indicating the presence or absence of the condition in a manner easily
understandable by any individual.

10/02/2024 8
 LITERATURE REVIEW
Author’s Title Purpose Method Conclusion
name
1. Gumaei A, Hybrid Feature Develop a hybrid Utilized The hybrid
Hassan MM, Extraction feature extraction regularized feature extraction
Hassan MR, Method with method for extreme learning method coupled
Alelaiwi A, Regularized accurate brain machine for brain with regularized
Fortino G. Extreme Learning tumor tumor extreme learning
Machine for Brain classification. classification machine shows
Tumor promise in
Classification improving brain
tumor
classification
accuracy,
potentially
enhancing
diagnostic
10/02/2024 DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING capabilities 9
 LITERATURE
REVIEW
Author’s Title Purpose Method Conclusion
name
2. Malathi M, Brain Tumour Develop a Utilized The CNN-based
segmentation model
Sinthia P. Segmentation convolutional TensorFlow
demonstrates
Using neural network framework for
promise in
Convolutional (CNN) for implementing the accurately
Neural Network accurate brain CNN-based brain segmenting brain
with TensorFlow tumor tumor tumors, offering
segmentation. segmentation potential
model. improvements in
diagnostic accuracy
3. Meng Y, Tang and treatment
C, Yu J, Meng S, Exposure to Lead planning.
Increases the Risk Conduct a meta- Analyzed existing
Zhang W.
of Meningioma analysis to studies to quantify The meta-analysis
the relationship suggests that
and Brain Cancer: investigate the
between lead exposure to lead is
A Meta-Analysis. association exposure and the associated with an
between lead incidence of increased risk of
10/02/2024 DEPARTMENT OF COMPUTER 10
exposure SCIENCE
and the& ENGINEERING
meningioma and meningioma and brain
 LITERATURE
REVIEW
Author’s Title Purpose Method Conclusion
name
4. Mittal M, Goyal Deep Learning Develop an Employed deep The deep learning-
LM, Kaur S, Kaur I, advanced tumor learning based segmentation
Based approach
Verma A, Jude HD Enhanced segmentation techniques for
demonstrates
approach for MR enhanced tumor
Tumor improved accuracy in
brain images segmentation in segmenting tumors
Segmentation using deep MR brain images from MR brain
Approach for learning images, offering
MR Brain potential
advancements in
Images medical imaging and
diagnosis
5. Muhannad Faleh Brain Tumor/Mass Develop a Utilized isolated The classification
Alanazi, Muhammad Classification classification framework employing
and developed isolated and developed
Umair Ali, Shaik Javeed Framework Using framework for brain
Hussain, Amad Zafar,
transfer deep- transfer deep-learning
Magnetic-Resonance- tumor/mass
Mohammed Mohatram, Imaging-Based detectionSCIENCE
using learning models models based on MRI
10/02/2024 DEPARTMENT OF COMPUTER & ENGINEERING data shows promise 11 in
Muhammad Irfan, Raed Isolated and isolated and trained on MRI
 ARCHITECTURE DIAGRAM
USER INTERFACE MODULE
USER INPUT

IMAGE PROCESSING MODULE

PROCESSED IMAGE

CONVOLUTIONAL NUERAL NETWORK(CNN)

MODEL
CLASSIFICATION

TUMOR CLASSIFICATION MODULE

CLASSIFIACATION RESULT

TUMOR SIZE ESTIMATION

MODULE
TUMOR SIZE

DISPLAY RESULT

10/02/2024 DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING 12

 ALGORITHM
• Algorithms Used:
• Convolutional Neural Networks (CNNs)
•
•
Adam Optimizer
Binary Cross-Entropy Loss
Libraires Used:
• Techniques Employed: • Os
•
•
Image Preprocessing (resizing, grayscale conversion)
Normalization of Pixel Values
• Random
• Data Augmentation • Tensorflow and
• Batch Normalization
tensorflow.keras
Model Training:
• Splitting data into Training and Validation Sets • Numpy
• Compilation with Adam Optimizer and Binary Cross-Entropy Loss • Cv2
• Evaluation Metrics: Accuracy
• Model Deployment: • Sklearn.model_selection
• Saving Trained Model • Tkinter
• Integration with Tkinter for GUI
• Visualization using Matplotlib • Mtplotlib.pyplot

10/02/2024 DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING 13

 MODULES

 Convolutional Neural Networks (CNNs):

 Utilized for automated brain tumor detection and classification in MRI images.
 Extracts relevant features from images using convolutional and max-pooling layers.
 Trained on labeled datasets with optimization using algorithms like Adam optimizer and
binary cross-entropy loss.

 Image Processing and Analysis:

 Applies preprocessing techniques (resizing, grayscale conversion) to standardize input
images.
 Employs image processing algorithms (binary thresholding, contour detection) for tumor
size estimation and region delineation.
 Enhances data quality and facilitates accurate tumor analysis, contributing to improved
diagnostic outcomes.
5/10/2024 DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING 14
 RESULT

10/02/2024 15
 RESULT

10/02/2024 16
 RESULT

10/02/2024 17
FUTURE DIRECTIONS
 Refining algorithms:
 Continuously optimize deep learning algorithms to further improve accuracy and
efficiency.
 Explore advanced techniques such as transfer learning and ensemble methods to
enhance model performance.
 Incorporating additional data sources:
 Integrate diverse datasets, including multi-modal imaging data and genetic
information, to capture a more holistic view of brain tumors.
 Collaborate with healthcare institutions to access larger and more diverse patient
cohorts for training and validation.
 Collaborating with healthcare professionals:
 Partner with radiologists, neurosurgeons, and oncologists to ensure the clinical
relevance and usability of the system.
 Conduct clinical trails to validate the effectiveness of the system in real-world
settings and gather feedback for further improvements.
10/02/2024 18
Conclusions
 In conclusion, our project represents a significant
advancement in brain tumor detection and analysis. By
improving accuracy and providing comprehensive insights,
we’re empowering clinicians to make more informed
decisions and ultimately improve patient outcomes.
 The enhancements made in our system pave the way for more personalized and
effective treatment strategies for patients with brain tumors.
 With ongoing collaboration and innovation, we are committed to further
refining our system to meet the evolving needs of clinicians and patients in the
fight against brain cancer.
5/10/2024 19
 REFERENCES
 Gumaei A, Hassan MM, Hassan MR, Alelaiwi A, Fortino G. A Hybrid feature extraction method with
regularized extreme learning machine for brain tumor classification. IEEE Access. 2019; 7:36266–73.
 Malathi M, Sinthia P. Brain tumour segmentation using convolutional neural network with tensor flow.
Asian Pac J Cancer Prevent APJCP. 2019;20(7):2095.
 Mittal M, Goyal LM, Kaur S, Kaur I, Verma A, Jude HD. Deep learning based enhanced tumor
segmentation approach for MR brain images. Appl Soft Comput. 2019; 78:346–54.
 Meng Y, Tang C, Yu J, Meng S, Zhang W. Exposure to lead increases the risk of meningioma and brain
cancer: a meta-analysis. J Trace Elem Med Biol. 2020; 60:126474.
 Muhannad Faleh Alanazi, Muhammad Umair Ali, Shaik Javeed Hussain, Amad Zafar, Mohammed
Mohatram, Muhammad Irfan, Raed AlRuwaili, Mubarak Alruwaili, Naif H. Ali, and Anas Mohammad
Albarrak. Brain tumor/mass classification framework using magnetic-resonance-imaging-based isolated
and developed transfer deep-learning model. Sensors 2022;22(1):372.

5/10/2024 DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING 20

 THANK YOU

5/10/2024 21