Scribd
Upload
40 views11 pages

Brain Tumor Classification Using Convolutional Neural Network

The document outlines a project focused on building a Convolutional Neural Network (CNN) model to classify brain tumor images using a Kaggle dataset. It details the steps of data preprocessing, model architecture, and evaluation, highlighting the successful binary classification of tumors. Future work includes extending the model to multi-class classification and testing with additional datasets.

Uploaded by

expacc007
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
40 views11 pages

Brain Tumor Classification Using Convolutional Neural Network

The document outlines a project focused on building a Convolutional Neural Network (CNN) model to classify brain tumor images using a Kaggle dataset. It details the steps of data preprocessing, model architecture, and evaluation, highlighting the successful binary classification of tumors. Future work includes extending the model to multi-class classification and testing with additional datasets.

Uploaded by

expacc007
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 11

Brain Tumor

Classification using
Convolutional
Neural Network
Kaggle Dataset Analysis

VITBhopal
Team Members
Vijay Vinod Mane 22BSA10048
Aman Deep Singh 22BSA10011
Arnish Baruah 22BSA10014
Aastha Vastani 22BSA10036
Kshatriya Nandini Kuldeep Singh 22BSA10057
Jayesh Dubey 22BSA10058
Akshat Barve 22BSA10066
Mohan Chaturvedi 22BSA10091
Vanshpreet kaur 22BSA10170
Rishabh Baruah 22BSA10330
Introduction
Purpose: Build a CNN model to classify brain tumor
images.

Tools and Technologies: Python, Keras, PIL,


Matplotlib.

Dataset: Kaggle dataset by Navneil Chuckerberti


with labeled images of tumors.

Key Steps:
Data preprocessing
Model building
Training and evaluation
Dataset Overview
Large dataset with labeled images.

Categories: Tumor (Yes) and No Tumor (No).

Significance:
High-quality images enable effective
training
Access: Dataset available on Kaggle.
Data Preprocessing

Steps Involved:
- Import necessary libraries (OS, PIL, Keras, etc.).
- Traverse folders ('Yes' and 'No') using OS library.
- Resize images to 128x128 pixels.
- Convert images to NumPy arrays for compatibility.
- Encode labels (0 for Tumor, 1 for Normal) using One-Hot Encoder.
Lists Created
- Data: Stores image data in NumPy format.
- Paths: Stores file paths of images.
- Result: Stores target classes.
Data Processing Logic
Folder Navigation:
- Loop through 'Yes' and 'No' folders.
- Append image paths and labels to respective lists.
Image Conversion:
- Read and resize images using PIL.
- Convert resized images to NumPy arrays.
Final Data Structure:
- Data list contains image data.
- Result list contains encoded labels
Model Architecture
Model Type: Sequential CNN
Layers:
- Conv2D: Extracts features from images.
- MaxPooling: Reduces spatial dimensions.
- Batch Normalization: Normalizes input.
- Dropout: Prevents overfitting.
- Flatten: Converts 2D matrix to 1D.
- Dense: Fully connected layer with softmax
Parameters:
- Input shape: (128, 128, 3)
- Filters: 32, 64
- Kernel size: (2, 2)
- Activation: ReLU and Softmax
Results and Evaluation
Testing:
- Randomly select images from test set.
- Predict using trained CNN model.
- Reshape images to required input dimensions.
Output:
- Binary classification: Tumor (0) or No Tumor (1).
- Confidence level for predictions.
Case Studies

Examples:
- Test Image 1: No Tumor (100% confidence).
- Test Image 2: Tumor (99.99% confidence).

Analysis:
- High accuracy for test cases.
- Robust predictions for both classes.
Conclusion
Summary:
- Successfully built and evaluated a CNN model for
tumor classification.
- Preprocessing and architecture were key to
performance.
Future Work:
- Extend to multi-class classification.
- Test with additional datasets.
Thank you
very much!

You might also like