REVOLUTIONIZING BRAIN TUMOR DETECTION WITH
CONVOLUTIONAL NEURAL NETWORKS
Presented By:- Guided By:-
Aakankshya Nayak Ast. Prof. Prakash Kumar
Anwesha Panda
Goutami Mohapatra
Payal Patra
Manisha Priyadarshini Sahoo
CONTENTS TABLE
SL NO. TOPICS
1 INTRODUCTION
2 OBJECTIVES
3 MOTIVATION
4 LITERATURE SURVEY
5 PROPOSED METHODOLOGY
6 SYSTEM ARCHITECTURE
7 DATA FLOW DIAGRAM
8 TRAINING DATASET MODELS
9 RESULTS
10 COMPARISON
11 FUTURE SCOPE
12 CONCLUSION
13 REFERENCES
INTRODRUCTION
WHAT IS BRAIN TUMORS ?
Brain tumors are abnormal growths of cells within the
brain or its surrounding structures, such as the meninges
or pituitary gland.
Types of Brain Tumors:
Primary Brain Tumors
Secondary Brain
Tumors
3
TYPES OF BRAIN TUMORS
4
OBJECTIVES
Improve Diagnostic Accuracy
Early Detection and Intervention
CNNs aim to enhance the precision and reliability
of brain tumor detection by automatically
Enables faster and more accurate identification of
identifying subtle patterns in medical images that
tumors at early stages, improving chances of
may be missed by human experts.
successful treatment.
Reduction in Human Error and Workload Scalability and Accessibility
Automates the image analysis process, reducing CNN-based tools can be deployed across various
dependency on radiologists and minimizing healthcare settings, improving access to quality
interpretation errors. diagnostics globally.
5
GOAL AND IMPACT OF OUR PROJECT
Develop an automated system using Convolutional Neural Networks (CNNs) to detect and
classify brain tumors in MRI images, accurately identifying tumor presence and type (e.g.,
glioma, meningioma, pituitary) without manual intervention.
Faster Diagonsis Better Accuracy
6
MOTIVATION
Ahmed et al. (2018)
Reported that delays in diagnosing brain tumors significantly impact survival
rates. Emphasized the need for faster, more accurate diagnostic tools.
Pereira et al. (2016)
Compared manual segmentation to CNN-based methods and showed that
manual analysis is not only time-consuming but also inconsistent across
radiologists.
Esteva et al. (2017)
Demonstrated that CNNs could match or exceed dermatologist-level accuracy in
image-based diagnosis, highlighting their medical imaging potential.
Chilamkurthy et al. (2018)
Used deep learning to detect critical findings in head CT scans across hospitals in
India, showing CNNs' potential in resource-limited environments.
7
LITERATURE SURVEY
References Purpose Model Limitations/future works
Examining the execution of the
To classify different types of brain designed neural network in the
Badža et al. [5] tumours using a convolutional CNN mentioned study, as well as
neural network enhanced ones in different
medical images
Lack of comparison of the
To classify brain tumours using a
Gumaei et al. [23] RELM technique used in this study and
hybrid feature extraction method
other machine learning methods
Proposing three architectures of Convolutional neural Explore other essential deep
convolutional neural networks networks (AlexNet, neural network’s architectures for
Rehman et al. [22]
(alexnet,Googlenet, and vggnet) to GoogLeNet, and brain tumour classification with
classify brain tumours VGGNet) less time complexity
8
PROPOSED METHODOLOGY
Data Collection and CNN Model Design Tumor Evalution and
Pre-processing and Training Classification Validation
Gather MRI brain Design a deep CNN Use the trained CNN to Evaluate model
scan datasets and architecture or use a classify images into performance using
apply preprocessing pre-trained model (e.g., categories such as metrics like accuracy,
techniques such as VGG, ResNet) and train glioma, meningioma, precision, recall, F1-
noise removal, it using labeled MRI pituitary tumor, or no score, and cross-
resizing, images. tumor. validation techniques.
normalization, and
data augmentation.
9
CNN ARCHITECTURE OVERVIEW
10
11
12
DATASET AND MRI IMAGES
13
TRAINING THE MODELS
Use early stopping to prevent overfitting.
# Early stopping callback
early_stop = EarlyStopping(monitor='val_loss', patience=5,
restore_best_weights=True)
# Train the model
history = model.fit(
data_gen.flow(X_train, y_train, batch_size=32),
validation_data=(X_val, y_val),
epochs=20,
callbacks=[early_stop]
)
14
RESULTS AND ANALYSIS
15
Comparison
Feature / Aspect Traditional Methods CNN-Based Methods
Accuracy Moderate (varies with expertise) High (often >90% with large datasets)
Feature Extraction Manual (requires domain knowledge) Automatic (learned from data)
Speed Time-consuming (manual analysis) Fast (real-time or near real-time)
Consistency Subject to human error and variability Highly consistent and repeatable
Highly scalable with computational
Scalability Limited (requires expert radiologists)
resources
Adaptability Difficult to update or generalize Easily fine-tuned with new data
Expensive (requires ongoing expert Cost-effective long-term (after
Cost Efficiency
involvement) setup/training)
16
MODEL LOSS AND ACCURACY GRAPHS
17
FUTURE SCOPE
18
REAL WORLD APPLICATIONS
Hospital and Clinics
IOT Based Diagonsis
Telemedicine
Medical Research Global Healthcare
19
CONCLUSION
Brain tumour detection is a critical application of deep learning in the medical
field, aiming to assist healthcare professionals with precise and efficient
diagnosis. This project implemented a Convolutional Neural Network (CNN) to
automate the detection of brain tumours using MRI images, showcasing the
potential of artificial intelligence in enhancing diagnostic accuracy and speed.
This project emphasizes the transformative role of AI in healthcare, addressing
the limitations of manual diagnosis, which can be prone to errors and time-
consuming. By reducing dependency on human intervention for initial
diagnosis, the solution has the potential to significantly improve clinical
workflows and patient outcomes.
20
References
Pereira, S., Pinto, A., Alves, V., & Silva, C. A. (2016).
Brain tumor segmentation using convolutional neural networks in MRI
images.
IEEE Transactions on Medical Imaging, 35(5), 1240–1251.
Chilamkurthy, S. et al. (2018).
Deep learning algorithms for detection of critical findings in head CT scans.
The Lancet, 392(10162), 2388–2396.
Hossain, M. S., & Muhammad, G. (2019).
Brain tumor detection using CNNs from MRI images: A comparative study.
Future Generation Computer Systems, 94, 87–94.
21
THANK YOU
