Brain Tumor Detection System

This project utilizes deep learning to detect and classify brain tumors from MRI images. The primary objective is to determine whether a tumor is present and, if so, classify its type with high accuracy.

Features

• Dataset Handling: Prepares and processes the MRI images for training.
• Deep Learning Model: Trains a Convolutional Neural Network (CNN) for tumor classification.
• Prediction System: Detects the presence of a tumor and identifies its type with an accuracy score.
• Future Prediction: Provides insights into the likelihood of tumor formation.

Workflow

1. Dataset Preparation: Curated MRI image dataset is preprocessed.
2. Model Training: A deep learning model is trained and saved as an .h5 file for later use.
3. Prediction: An MRI image is input, and the system determines:
   • If a tumor is present.
   • The type of tumor and its likelihood.
4. Results: Outputs a classification report and accuracy score.

Technologies Used

• Backend: Flask is used to create the web server and handle requests.
• Frontend: HTML, CSS, and JavaScript are used to create the user interface for uploading images and displaying results.
• Machine Learning: TensorFlow and Keras libraries are utilized to build and train the CNN model.
• Data Visualization: Matplotlib and Seaborn are used for plotting graphs and visualizing model performance.
• Database: SQLite is integrated to store user inputs and results (optional).
• Deployment: The app can be deployed on platforms like Heroku or AWS for broader accessibility.

🔄 Workflow Diagram

graph TD
A[Load Dataset] --> B[Preprocess Dataset]
B --> C[Train Deep Learning Model]
C --> D[Save Model as .h5 File]
D --> E[Input MRI Image]
E --> F{Is Tumor Present?}
F -- Yes --> G[Determine Tumor Type and Accuracy Score]
F -- No --> H[Healthy - Display Result]
G --> I[Provide Insights on Tumor Formation]

Loading

How the App Works

1. Input: The user uploads an MRI image through the web interface.
2. Processing:
   • The image is preprocessed and passed to the trained deep learning model.
   • The model predicts whether a tumor is present.
3. Output:
   • If a tumor is detected, the app determines the type of tumor and displays the classification along with a confidence score.
   • If no tumor is detected, the app informs the user that the MRI scan appears healthy.
4. Insights:
   • The app provides additional insights, such as the likelihood of tumor formation based on the analysis.

Project Structure

• app.py: The main backend script to deploy the trained model and process user requests.
• project3.ipynb: Jupyter notebook for training and evaluating the CNN model.
• design.png: Visual representation of the project's workflow.
• models/: Directory for storing saved models (.h5 files).
• static/: Static files such as images and CSS for the web interface.
• templates/: HTML templates for rendering the web pages.

Setup Instructions

Prerequisites

• Python 3.8+
• Required Python libraries: tensorflow, flask, numpy, pandas, matplotlib, seaborn, and sklearn.

Installation

1. Clone the repository:

   git clone <repository_url>
   cd <repository_directory>

2. Install dependencies:

   pip install -r requirements.txt

3. Train the model (optional):
   • Open project3.ipynb and run all the cells to train and save the model.
4. Start the application:

   python app.py

5. Access the application: Open a browser and navigate to http://127.0.0.1:5000.

Usage

1. Upload an MRI image through the web interface.
2. View the prediction results:
   • Tumor presence.
   • Tumor type classification.
   • Prediction likelihood.

Results

• Achieved an accuracy of 99%.
• Demonstrates the potential for early detection and diagnosis of brain tumors.

Future Enhancements

• Integration with cloud storage for large-scale datasets.
• Real-time MRI image processing with edge devices.
• Incorporating additional diagnostic metrics.

License

This project is licensed under the MIT License. See the LICENSE file for details.