🧠 Brain Tumor MRI Classification using EfficientNet-B3

An AI-powered web application for classifying brain MRI scans into tumor categories using deep learning and transfer learning with EfficientNet-B3.

---

📌 Project Overview

This project uses a Convolutional Neural Network (CNN) based on EfficientNet-B3 to classify brain MRI images into four categories:

• Glioma
• Meningioma
• Pituitary Tumor
• No Tumor

The trained model is deployed using Streamlit to provide a clean and interactive web interface with real-time predictions and visual analytics.

---

🚀 Features

• ✅ Transfer Learning using EfficientNet-B3
• ✅ Fine-tuned CNN for medical image classification
• ✅ Real-time MRI image upload and prediction
• ✅ Interactive visualizations (Gauge, Radar, Donut charts)
• ✅ Clinical information display (Symptoms, Treatment, Statistics)
• ✅ Clean and professional UI
• ✅ Cached model loading for performance optimization

---

🏗️ Model Architecture

The model uses:

• EfficientNet-B3 (Pretrained on ImageNet)
• GlobalAveragePooling2D
• BatchNormalization
• Dense (256 units, ReLU activation)
• Dropout (0.5)
• Dense (4 units, Softmax activation)

Why EfficientNet-B3?

• Balanced accuracy and computational efficiency
• Compound scaling (depth, width, resolution)
• Designed for 300×300 input images
• Strong feature extraction capability for medical imaging

---

🧠 Training Strategy

Dataset

• Training & Validation split: 80–20
• Separate test dataset
• Image size: 300 × 300
• Batch size: 32

Techniques Used

• Data augmentation (Flip, Rotation, Zoom, Contrast)
• Transfer learning (frozen base model initially)
• Fine-tuning last 50 layers
• Adam optimizer
• Sparse categorical crossentropy
• EarlyStopping
• ReduceLROnPlateau
• ModelCheckpoint

Evaluation Metrics

• Accuracy
• Precision
• Recall
• F1-score
• Confusion Matrix

---

📊 Web Application (Streamlit)

The app allows users to:

• Upload an MRI scan (JPG/PNG)
• Run AI-powered classification
• View probability distribution
• See confidence gauge
• Analyze radar chart comparison
• Explore clinical insights

Visual Components

• 📈 Confidence Gauge
• 🥯 Donut Probability Chart
• 📊 Radar Analysis
• 📋 Clinical Tabs (Symptoms, Treatment, Statistics)

---

🛠️ Tech Stack

Backend / Model

• Python
• TensorFlow / Keras
• EfficientNet-B3
• NumPy
• scikit-learn

Frontend / Deployment

• Streamlit
• Plotly
• HTML/CSS Styling
• Pillow (PIL)

---

📂 Project Structure

Brain-Tumor-Classifier/
│
├── tumor_model.keras
├── app.py
├── training_script.py
├── requirements.txt
└── README.md

---

⚙️ Installation & Setup

1️⃣ Clone Repository

git clone https://github.com/your-username/brain-tumor-classifier.git
cd brain-tumor-classifier

2️⃣ Install Dependencies

pip install -r requirements.txt

3️⃣ Run Streamlit App

streamlit run app.py

---

🧪 Model Performance

• Training Accuracy: ~87–90%
• Validation Accuracy: ~85–88%
• Test Accuracy: ~85%+
• Balanced precision and recall across tumor classes

(Exact numbers depend on dataset version.)

---

🔬 How the Model Works

1. MRI image is uploaded
2. Image resized to 300×300
3. Preprocessing using EfficientNet normalization
4. CNN extracts deep features
5. Classification head predicts probabilities
6. Highest probability class selected
7. Results visualized with interactive charts

---

🎯 Key Learning Outcomes

• Deep understanding of CNNs
• Transfer learning and fine-tuning
• EfficientNet architecture
• Handling medical imaging datasets
• Model evaluation beyond accuracy
• Deployment using Streamlit
• UI/UX for AI applications

---

⚠️ Disclaimer

This application is for educational and research purposes only.
It is not a substitute for professional medical diagnosis.
Always consult qualified healthcare professionals for medical advice.

---

👩‍💻 Author

Developed as part of a deep learning and medical AI project.

---