Scribd
Upload
11 views

Project Slides

This document discusses the critical need for early detection of brain tumors and presents a study employing advanced machine learning techniques, including a convolutional neural network, to analyze a dataset of 3284 MRI images. The study demonstrates improved accuracy in brain tumor detection, achieving a 93% accuracy rate with the CNN model, while comparing various machine learning models and their performance. The findings emphasize the potential of machine learning and deep learning in enhancing diagnostic methods for brain tumors.

Uploaded by

Sunny Govardhan
Copyright
© © All Rights Reserved
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
11 views

Project Slides

This document discusses the critical need for early detection of brain tumors and presents a study employing advanced machine learning techniques, including a convolutional neural network, to analyze a dataset of 3284 MRI images. The study demonstrates improved accuracy in brain tumor detection, achieving a 93% accuracy rate with the CNN model, while comparing various machine learning models and their performance. The findings emphasize the potential of machine learning and deep learning in enhancing diagnostic methods for brain tumors.

Uploaded by

Sunny Govardhan
Copyright
© © All Rights Reserved
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 24

BRAIN

TUMOR
DETECTION
Content
• Motivation
• Introduction
• Literature Review
• Proposed Methodology
• Results
• Conclusion
• References
Motivation
• Brain tumors, a perilous form of cancer, affect individuals across all age groups.
• Early detection is vital for better patient outcomes, but traditional methods like
eye analysis of MRI data have inherent limitations.
• The alarming 300% increase in brain tumor-related deaths over the past 30
years underscores the urgent need for timely identification.
• Untreated brain tumors can lead to fatal consequences, emphasizing the critical
importance of accurate and non-invasive diagnostic methods.
• Magnetic Resonance Imaging (MRI) stands out as the most effective tool for
diagnosing brain cancers.
Introduction
• Recent advancements in machine learning, particularly deep learning, enable precise
recognition of medical imaging patterns.
• This study employs diverse machine-learning techniques, including a convolutional neural
network with data augmentation and feature extraction, on a substantial MRI dataset.
• Evaluation of multiple machine-learning models on a dataset comprising 3284 brain tumor
(MRI) images provides nuanced insights into their performance.
• This research significantly contributes to ongoing endeavors in brain tumor detection,
utilizing state-of-the-art machine learning and deep learning techniques, and the results
are rigorously compared with existing literature to advance our understanding of effective
approaches in brain tumor detection.
Literature Review

Refernces Models Accuracy Rate

Malarvizhi et al.(2022) K-means clustering, SVM classification 80%

23-layer CNN, Fine-tuned CNN with


Md. Saikat Islam Khan (2022) 97.8%, 100%
VGG16
Divyamary.D and Gopika.S,
Naive Bayes Classifier 84%
(2020)

Rajni H et al., (2020) Naive Bayes Classifier 94%

Ramdas Vankdothu and Mohd Hybrid ANFIS and SVM, FCM


99.24%
Abdul Hameed, (2022) segmentation, GWO, SSO, GA
Influence from Previous Work: MRI-
based Brain Tumor Detection
Previous Study
• Title: “MRI-based Brain Tumor Detection using Convolutional Deep Learning Methods”
(2023)
Methodology:
• Dataset: 3264 MRI images (glioma, meningioma, pituitary tumors, healthy brains)
• Techniques: 2D CNN, Auto-encoder Network, Machine Learning (MLP, KNN)
Results:
• Acuracy: 2D CNN (93%), Auto-encoder (90%)
• KNN achieved the highest ML accuracy (86%)
Our Contributions
Overview
• Extending prior work, our study implements state-of-the-art ML and deep learning
algorithms.
Results
• Improved accuracy, efficiency, and applicability in diverse scenarios.
Comparison:
• Highlight key improvements or differences
• Make predictions
• Use visuals to enhance understanding
Proposed Methodology
Dataset Overview
Dataset Information: Total Images: 3284 brain tumor
(MRI) images
Dataset after up-sampling: 3788

• Data Split: Applied an 80-20 train-test split.


Train (3030) and Test (758)
Source: Publicly available dataset on Kaggle
MRI Images of Four Classes:
• No tumor (Healthy Brain)
• Meningioma Tumor
• Pituitary Gland Tumor
• Glioma Tumor
Machine Learning Models
• Naive Bayes: Limited performance • SVM (Support Vector Machine): Strong
suggesting simplicity may not capture overall performance. Effective for both
complex data patterns effectively. linear and non-linear patterns.

• KNN (K-Nearest Neighbors): Solid, • MLP (Multi-Layer Perceptron): High


balanced performance across metrics. accuracy and balanced metrics.

• Decision Tree: Balanced performance in • Random Forest: Excellent overall


accuracy, recall, precision, and F1. performance with high metrics. The
ensemble approach enhances robustness.
• Logistic Regression: Consistent, robust
• XGBoost: Top-tier performance with high
performance. Simple and interpretable,
accuracy and balanced metrics. Gradient
suitable for various tasks.
boosting optimizes generalization.
No tumor: 0,
Confusion Matrix of ML Models Glioma tumor: 1,
Meningioma Tumor: 2,
1/4 Pituitary Tumor: 3

SVM Confusion Matrix Naïve Bayes Confusion Matrix


No tumor: 0,
Confusion Matrix of ML Models Glioma tumor: 1,
Meningioma Tumor: 2,
2/4 Pituitary Tumor: 3

Random Forest Confusion K-NN Confusion Matrix


Matrix
No tumor: 0,
Confusion Matrix of ML Models Glioma tumor: 1,
Meningioma Tumor: 2,
3/4 Pituitary Tumor: 3

Logistic Regression Confusion XGBoost Confusion Matrix


Matrix
No tumor: 0,
Glioma tumor: 1,
Confusion Matrix of ML Models 4/4 Meningioma Tumor: 2,
Pituitary Tumor: 3

Decision Tree Confusion Matrix Multilayer Perceptron Confusion


Matrix
Results on Machine Learning Models

Model Accuracy Avg Recall Avg Precision F1 Score


Naive Bayes 53% 0.54 0.58 0.5
KNN 84% 0.84 0.85 0.84
Decision Tree 83% 0.83 0.83 0.83
Logistic Regression 88% 0.88 0.88 0.88
SVM 88% 0.88 0.88 0.88
MLP 90% 0.9 0.89 0.89
Random Forest 92% 0.92 0.93 0.92
XGBoost 92% 0.92 0.93 0.92
Comparative Analysis

Accuracy

0.9 0.92 0.92


0.88 0.88
0.84 0.83

0.53

NB KNN DT LR SVM MLP RF XGBoost

Results we have achieved Results from paper, [Saeedi et al.,


(2023)]
Convolutional Neural Network (CNN)

Model Architecture:
• Sequential model with distinct layers.
• Conv2D Layers: Utilized 32 filters of 3x3 kernels with ReLU activation.
• Max Pooling Layers: Applied 2x2 max pooling to reduce spatial
dimensions.
• Flattened Layer: Transformation of output into a 1D array.
• ReLU Activation: Introduced non-linearity for better generalization on
complex data.
• Output Layer: Comprised of 4 neurons with softmax activation for
multi-class classification.
Model Compilation:
• Optimizer: Adam optimizer employed.
• Loss Function: sparse_categorical_crossentropy chosen based on
existing literature.
Results on CNN
• On CNN we got 93% accuracy
Model Predictions

Actual Images: Pituitary Tumor Actual Images: Meningioma Tumor


Model Predictions

Actual Images: Glioma Tumor Actual Images: No Tumor


Summary
Machine Learning Models:
• Naive Bayes and KNN showed lower performance.
• Decision Tree, Logistic Regression, SVM, MLP, Random Forest, and XGBoost
demonstrated varying levels of accuracy.
• Top performers were Random Forest and XGBoost.
Convolutional Neural Network (CNN):
• Achieved a high accuracy of 93%.
• Outperformed traditional ML models on image data.
References 1/2
• El-Dahshan, E. S. A., Mohsen, H. M., Revett, K., & Salem, A. B. M. (2014). Computer-aided diagnosis of human
brain tumor through MRI: A survey and a new algorithm. Expert Systems with Applications, 41(11), 5526–5545.

• Badža, M. M., & Barjaktarović, M. Č. (2020). Classification of brain tumors from MRI images using a
convolutional neural network. Applied Sciences, 10(6), 1999.

• Divyamary, D., Gopika, S., Pradeeba, S., & Bhuvaneswari, M. (2020, March). Brain tumor detection from MRI
images using Naive classifier. In 2020 6th international conference on advanced computing and communication
systems (ICACCS) (pp. 620-622). IEEE.

• H, R., B, S., Reddy, V. K., & M, V. S. (2020). Brain Tumor Detection using Naïve Bayes Classification. International
Research Journal of Engineering and Technology (IRJET), 07(06), 6851.

• Malarvizhi, A. B., Mofika, A., Monapreetha, M., & Arunnagiri, A. M. (2022, August). Brain tumour classification
using machine learning algorithm. In Journal of Physics: Conference Series (Vol. 2318, No. 1, p. 012042). IOP
Publishing.
References 2/2
• Vankdothu, R., & Hameed, M. A. (2022). Brain tumor segmentation of MR images using SVM and fuzzy classifier in
machine learning. Measurement: Sensors, 24, 100440.

• Khan, M. S. I., Rahman, A., Debnath, T., Karim, M. R., Nasir, M. K., Band, S. S., ... & Dehzangi, I. (2022). Accurate
brain tumor detection using deep convolutional neural network. Computational and Structural Biotechnology
Journal, 20, 4733-4745.

• Saeedi, S., Rezayi, S., Keshavarz, H., & R. Niakan Kalhori, S. (2023). MRI-based brain tumor detection using
convolutional deep learning methods and chosen machine learning techniques. BMC Medical Informatics and
Decision Making, 23(1), 16.

• Sartaj. (2020). Brain tumor classification (MRI). Retrieved from


https://www.kaggle.com/datasets/sartajbhuvaji/brain-tumor-classification-mri?select=Training
THANK YOU
FOR YOUR
ATTENTION.

You might also like