MRI Tumor Detection using Deep Learning

A deep learning–based MRI brain tumor classification system built using EfficientNet-B0 and deployed through a Flask web application. The system enables real-time tumor prediction from MRI images and provides an analytical dashboard for evaluating model performance.

This project was developed as part of the MCA Capstone (CMR University, 2023–2025).

---

Table of Contents

• Project Overview
• Features
• Tech Stack
• Dataset
• Project Structure
• Installation
• Usage
• API Documentation
• Model Performance
• Training Workflow
• Future Enhancements
• Acknowledgements
• License

---

Project Overview

This project implements an automated MRI tumor detection system using a deep learning classifier trained on brain MRI images. The trained model (modelfinal.h5) categorizes MRIs into:

• glioma
• meningioma
• notumor
• pituitary

A Flask web application allows users to upload MRI images for prediction and provides an interactive dashboard showing ROC curves, confusion matrices, classification reports, and dataset analytics.

---

Features

• EfficientNet-B0–based deep learning classifier
• Real-time MRI tumor prediction via Flask app
• Dashboard visualizations:
  • Dataset distribution
  • ROC curve
  • Confusion matrix
  • Classification report
  • Training accuracy/loss curves
• Pre-trained model included (modelfinal.h5)
• JSON data exports for evaluation metrics

---

Tech Stack

Backend

• Python
• Flask
• TensorFlow / Keras

Frontend

• HTML5
• CSS3
• Bootstrap
• JavaScript

Visualization

• Matplotlib
• Seaborn

---

Dataset

• Public brain MRI datasets (Kaggle / BRATS-style)
• Classes: glioma, meningioma, notumor, pituitary
• Images resized to 224×224
• Preprocessing:
  • Normalization
  • Augmentation (rotation, zoom, flip)
• Split:
  • 70% training
  • 20% validation
  • 10% testing

---

Project Structure

CAPSTONE_PROJECT/
├── MRI Images/
│   ├── Training/
│   └── Testing/
├── static/
│   ├── charts/
│   ├── data/
│   ├── uploads/
│   └── style.css
├── templates/
│   ├── index.html
│   └── dashboard.html
├── uploads/
├── Brain-Tumor_Detection.ipynb
├── classification_report.json
├── history.json
├── main.py
├── modelfinal.h5
├── test_labels.npy
├── test_predictions.npy
└── README.md

---

Installation

1. Clone the repository

git clone https://github.com/s3ldc/BrainTumorDetection
cd CAPSTONE_PROJECT

2. Create a virtual environment

python -m venv venv

Activate:

# Windows
venv\Scripts\activate

# macOS / Linux
source venv/bin/activate

3. Install dependencies

Create a requirements.txt:

flask
tensorflow
numpy
opencv-python
pillow
matplotlib
scikit-learn
pandas
seaborn
gunicorn

Install the dependencies:

pip install -r requirements.txt

---

Usage

Start the Flask server:

python main.py

Access the application:

http://127.0.0.1:5000/

Pages available:

• Homepage — Upload MRI & get prediction
• Dashboard — Evaluation metrics, dataset analysis, charts

---

API Documentation

POST /predict

Uploads an MRI image and returns a tumor prediction.

Form Data:

• file: MRI image (jpg/png)

Sample Response:

{
  "prediction": "glioma",
  "confidence": 0.9873
}

---

Model Performance

• AUC: 0.99–1.00
• Precision: 0.93–0.99
• Recall: 0.93–1.00
• F1 Score: 0.94–0.99

Visual outputs:

• classification_report.json
• history.json
• dashboard charts

---

Training Workflow

The Jupyter notebook includes:

1. Dataset loading
2. Preprocessing & augmentation
3. EfficientNet-B0 model creation
4. Training with callbacks
5. Evaluation (ROC, confusion matrix, reports)
6. Saving the model and metrics

---

Future Enhancements

• Grad-CAM visual explanations
• Hospital PACS (DICOM) Integration
• Cloud Deployment (AWS, Render, GCP)
• Lightweight mobile version
• Tumor segmentation model

---

Acknowledgements

• Guide: Prof. Aurangazeb Khan
• Institution: CMR University
• TensorFlow, Flask, Bootstrap communities
• MRI dataset contributors

---

License

This project was developed as part of an academic capstone.
For reuse or modification, please credit the author.