Don Bosco Institute of Technology

Kumbalagodu, Mysuru Road, Bengaluru – 560074

Department of Information Science & Engineering

“Brain Tumor Detection and Classification Using ML

And DL Approaches”
Presented by Under the Guidance of
Project Batch :B-31 Mr. V Senthil Kumar,
1. Manjunath Pujeri 1DB22IS078 Asst. Professor,
2. Pranav B 1DB22IS106 Dept. of ISE, DBIT
3. Pratham B Noola 1DB22IS108
4. Ayush S 1DB23IS401
 Don Bosco Institute of Technology
Kumbalagodu, Mysuru Road, Bengaluru – 560074
Department of Information Science & Engineering

Agenda
•Introduction
•Literature Survey
•Aim & Objective
• Methodology
•Existing System
•Proposed System
•Application & Advantages
•References
 Don Bosco Institute of Technology
Kumbalagodu, Mysuru Road, Bengaluru – 560074
Department of Information Science & Engineering

INTRODUCTION

•Brain tumors are abnormal cell growths in the brain; can be benign or malignant.
•Early and accurate detection is critical for effective treatment and patient survival.
•Manual MRI scan analysis is time-consuming and prone to human error.
•Convolutional Neural Networks (CNNs) excel in image recognition tasks.
•CNNs automatically extract features and classify tumor types from MRI images.
•Offers faster, more reliable, and automated brain tumor diagnosis.
•Helps support radiologists and improve clinical decision-making.
 Don Bosco Institute of Technology
Kumbalagodu, Mysuru Road, Bengaluru – 560074
Department of Information Science & Engineering

INTRODUCTION

The bar chart shows brain cancer net survival rates by

The image shows pie charts of brain tumor types
age group.
and global distribution of new brain cancer cases
by region.
 Don Bosco Institute of Technology
Kumbalagodu, Mysuru Road, Bengaluru – 560074
Department of Information Science & Engineering

AIM
•To develop a deep learning-based system for automatic brain tumor detection using
MRI scans.
•To apply Convolutional Neural Networks (CNNs) for accurate classification of
different brain tumor types.
•To support radiologists by reducing manual diagnostic effort and improving
consistency.
•To enable early diagnosis, improve treatment planning, and enhance patient outcomes.
 Don Bosco Institute of Technology
Kumbalagodu, Mysuru Road, Bengaluru – 560074
Department of Information Science & Engineering

OBJECTIVES

•To collect and preprocess MRI brain scan datasets for training and testing the model.
•To design and train a CNN architecture capable of detecting and classifying brain tumors.
•To evaluate model performance using metrics such as accuracy, precision, recall
•To compare the proposed model with existing techniques for reliability and efficiency.
•To deploy the model in a user-friendly interface for practical medical use.
 Don Bosco Institute of Technology
Kumbalagodu, Mysuru Road, Bengaluru – 560074
Department of Information Science & Engineering

LITERATURE SURVEY
Paper : 1
Title: An Advanced Brain Tumour Detection Framework Using Improved Fast Mask R-CNN Techniques
Authors: G. Shankar et al.
Year Published: 2024, IEEE ICSCAN
Proposed Idea:
Uses an Improved Fast Mask R-CNN for high-precision detection and boundary segmentation of
brain tumors.
Architecture:
Fast Mask R-CNN + SVM,Advanced image preprocessing and feature extraction
Methodology:
Dataset: 218 MRI images,Preprocessing, segmentation, classification using IFMRCNN Validation by
neurologist and ground truth comparison
Drawback:
High performance on small datasets, but not tested on larger or varied data—limits generalizability.
 Don Bosco Institute of Technology
Kumbalagodu, Mysuru Road, Bengaluru – 560074
Department of Information Science & Engineering

LITERATURE SURVEY
Paper : 2
Title: MRI Images Based Brain Tumor Detection Using CNN for Multiclass Classification
Authors: Nikita Mahajan, Hariram Chavan
Year Published: 2023,IEEE ASIANCON
Proposed Idea:
Presents a 9-layer CNN model for multiclass classification with improved decision support for doctors.
Architecture:
9-layer CNN (convolution, pooling, FC, Softmax),Pretrained on grayscale 150x150 MRI images
Methodology:
Dataset: Kaggle (3264 MRI images),Evaluated with metrics like accuracy, precision, recall, F1,Applied
80-20 split and training-validation routines
Drawback:
Relatively simple model; might underperform on complex or noisy real-world clinical data.
 Don Bosco Institute of Technology
Kumbalagodu, Mysuru Road, Bengaluru – 560074
Department of Information Science & Engineering

LITERATURE SURVEY
Paper : 3
Title: Brain Tumor Detection System using Convolutional Neural Network
Authors: Shubham Koshti et al.
Year Published: 2022,IEEE PuneCon
Proposed Idea:
A web-based brain tumor detection system using CNN and Transfer Learning for early-stage tumor
detection
Architecture:
Custom CNN model,Connected to a web app via API
Methodology:
Dataset: Kaggle (3264 MRI images, 4 classes),Transfer Learning, callbacks, and image
segmentation,Deployed model in a 4-route web application
Drawback:
Accuracy depends heavily on high-quality dataset; false positives/negatives possible if poor data is used.
 Don Bosco Institute of Technology
Kumbalagodu, Mysuru Road, Bengaluru – 560074
Department of Information Science & Engineering

LITERATURE SURVEY
Paper : 4
Title: Implementing a Novel Low Complexity CNN Model for Brain Tumor Detection
Authors: Nawras Al-Ani, Omran Al-Shamma
Year Published: 2022, IEEE 8th ICCITM
Proposed Idea:
Introduces an improved CNN model that balances high accuracy and low complexity for
detecting brain tumors using MRI.
Architecture:
• Based on AlexNet,Enhanced over standard CNNs
Methodology:
• Dataset: Kaggle (3350 MRI images),Preprocessing: resizing, normalization, labeling
Drawback:
Focuses only on binary classification not generalized for multi-class tumor types.
 Don Bosco Institute of Technology
Kumbalagodu, Mysuru Road, Bengaluru – 560074
Department of Information Science & Engineering

LITERATURE SURVEY
Title Authors Year Proposed Architecture Methodology Drawback
Published Idea
INCS: A Novel Brain Mohan V. M. Thota, B. 2024 Deep learning-assisted CNN + INCS CNN compared with High accuracy, verified
Tumor Detection Ravi Prasad Intelligent Neural INCS; multiclass via cross-validation
Methodology Design Classification Strategy classification of glioma,
(INCS) for classifying meningioma, pituitary,
brain tumors and healthy

Advanced Brain Tumor Mohan V. M. Thota, B. 2024 INCS: Intelligent Neural CNN + INCS Combines CNN with High classification
Detection using Improved Ravi Prasad Classification Strategy intelligent neural accuracy for 3 tumor
Fast Mask R-CNN using deep learning classifiers for multiclass types + healthy MRI,
tumor detection cross-validation with
CNN

Brain Tumor Detection Adnan Sakib et al. 2023 Hybrid model combining VGG16 + custom CNN Transfer learning on ccuracy: 96.52%
and Classification Using VGG16 with additional VGG16; MRI images
Hybrid VGG Network CNN layers classified into 4 categories

Brain Tumor Detection Shubham Koshti et al. 2022 Web-based CNN CNN + Transfer Learning image preprocessing + Accuracy: 97%
System using CNN detection system with user web deployment; CNN
input MRI scans trained on 80:10:10 split

Detection of Brain Tumor G. Shankar et al. 2024 Enhanced FMRCNN with mproved Fast Mask R- FMRCNN + OpenCV + 100% (meningioma),
using CNN image processing and CNN SVM for segmentation 87.5% (glioma), Avg
SVM and classification score: 94.6%
 Don Bosco Institute of Technology
Kumbalagodu, Mysuru Road, Bengaluru – 560074
Department of Information Science & Engineering

LITERATURE SURVEY
Title Authors Year Proposed Architecture Methodology Drawback
Published Idea
Detection of Brain Tumor Tapesh Kumar et al. 2022 CNN trained on CNN Grayscale conversion + Better accuracy than state-
using CNN preprocessed Kaggle MRI Gaussian blur + train/test of-the-art models
dataset split

Comparison of VGG-19 Dr. J.K. Periasamy et al. 2023 Comparative analysis of VGG-19, ResNet-50 MRI processing and Both perform well;
and RESNET-50 VGG-19 and ResNet-50 CNN-based classification detailed comparison of
Algorithms in Brain on same dataset metrics provided
Tumor Detection

MRI-Based Brain Tumor Nikita Mahajan, Hariram 2023 9-layer CNN for 9-layer CNN Texture/location/shape- Improved accuracy, CNN
Detection Using CNN for Chavan multiclass classification of based classification using advantageous for small
Multiclass Classification brain tumors Kaggle data datasets

Brain Tumor Diagnose Hanming Hu et al. 2021 YOLO + CNN for VGG, ResNet, AlexNet, Used multiple CNNs, Demonstrated viability of
Applying CNN through localization and GoogleNet, YOLO implemented YOLO, CNN for tumor diagnosis,
MRI classification observed accuracy drop but noted YOLO's
with YOLO drawbacks
 Don Bosco Institute of Technology
Kumbalagodu, Mysuru Road, Bengaluru – 560074
Department of Information Science & Engineering

LITERATURE SURVEY

The bar chart displays the number of literature sources using different
brain tumor detection methods, with Region Growing and U-Net
variants being the most commonly used, followed by Custom CNN.
 Don Bosco Institute of Technology
Kumbalagodu, Mysuru Road, Bengaluru – 560074
Department of Information Science & Engineering

EXISTING SYSTEM

•CNN: Automates tumor detection from MRI images using deep feature extraction.
•VGG16/VGG19: Pre-trained deep networks for accurate classification.
•ResNet: Residual networks for deeper, stable training.
•U-Net: Efficient tumor segmentation model.
•MobileNet: Lightweight model for real-time inference.
•3D CNN: Analyzes volumetric MRI data.
•Transfer Learning: Reduces training data needs.
•DenseNet: Enhances accuracy via dense layer connections.
 Don Bosco Institute of Technology
Kumbalagodu, Mysuru Road, Bengaluru – 560074
Department of Information Science & Engineering

PROPOSED SYSTEM

• Automated Image Processing with CNN: Use Convolutional Neural Networks (CNNs) to
automatically detect and classify brain tumors from MRI scans. Models like U-Net, ResNet,
or custom CNNs can identify tumor types such as glioma, meningioma, pituitary tumors, or no
tumor.
• Advanced Preprocessing:Apply techniques like skull stripping, image normalization, and
augmentation to clean and standardize MRI scans before training.Helps improve model
accuracy and robustness.
• Multi-Class Classification: The CNN model is trained to distinguish between multiple tumor
types, not just tumor vs. no tumor.Increases clinical usefulness by helping doctors identify
specific tumor types early.
 Don Bosco Institute of Technology
Kumbalagodu, Mysuru Road, Bengaluru – 560074
Department of Information Science & Engineering

APPLICATIONS

• Detects and classifies brain tumors using CNN.

•Uses MRI images.
•Trained on labeled dataset.
•Predicts tumor type.
•Built with Python, TensorFlow.
•Deploy via Streamlit.
 Don Bosco Institute of Technology
Kumbalagodu, Mysuru Road, Bengaluru – 560074
Department of Information Science & Engineering

ADVANTAGES
• High Accuracy: CNNs can achieve very high precision and recall in detecting and classifying
tumors when trained on quality MRI datasets.Helps reduce false positives and false negatives in
diagnosis.
• Automatic Feature Extraction: Unlike traditional methods, CNNs automatically learn and
extract complex patterns from MRI images—no need for manual feature engineering.
• Fast and Scalable Diagnosis: Once trained, CNN models can analyze MRI scans in seconds,
enabling faster decision-making and scalable screening in hospitals.
• Remote Diagnosis Possibility: CNN-based tools can be deployed in cloud systems or mobile
apps, helping with diagnosis in rural or resource-limited settings.
• Supports Multi-Class Classification: CNNs can identify different types of brain tumors
(e.g., glioma, meningioma, pituitary) instead of just detecting tumor vs. no tumor.
 Don Bosco Institute of Technology
Kumbalagodu, Mysuru Road, Bengaluru – 560074
Department of Information Science & Engineering
REFERENCES
1. M. Tamilselvi, "Human facial emotion recognition using graphical cascaded convolutional neural network," in Proc. 2024
Ninth Int. Conf. Sci. Technol. Eng. Math. (ICONSTEM), Chennai, India, 2024
2. A. A. Asiri et al., "Brain tumor detection and classification using fine-tuned CNN with ResNet50 and U-Net model: A study
on TCGA-LGG and TCIA dataset for MRI applications," Life, vol. 13, no. 7, p. 1449, 2023
3. Siar, M., Teshnehlab, M.: Brain Tumor Detection Using Deep Neural Network and Machine Learning Algorithm. In: ICCKE,
pp. 219–224 (2019)
4. Sajid, S., Hussain, S., Sarwar, A.: Brain Tumor Detection and Segmentation in MR Images Using Deep Learning. Arabian
Journal for Science and Engineering, 1–10 (2019)
5. Amin, J. et al.: Detection of Brain Tumor based on Features Fusion and ML. J. Ambient Intell. Hum. Comput. (2018)
6. Basha, D.M.M. et al.: Brain Tumor Detection Using Machine Learning. Int. J. Comput. Sci. Mobile Comput. 11(1), 146–152
(2022)
7. Amin, J. et al.: Brain Tumor Detection and Classification Using ML: A Survey. Complex Intell. Syst. 8(3), 3161–3183 (2022)
8. Kshirsagar, P.R. et al.: MRI Image Based Brain Tumor Detection Using ML. TEST Engineering, 81, 3672–3680 (2020)
 Don Bosco Institute of Technology
Kumbalagodu, Mysuru Road, Bengaluru – 560074
Department of Information Science & Engineering

THANK
YOU