🧠 Brain Tumor MRI Classification with Deep Learning

An advanced deep learning system for automated brain tumor classification from MRI scans, achieving 96.89% accuracy.

---

🌟 Key Features

• Dual Model Architecture
  • Transfer Learning with Xception (96.89% accuracy)
  • Custom CNN (95.65% accuracy)
• Interactive Web Interface
  • Real-time predictions
  • Saliency map visualization
  • Confidence scores
• AI-Powered Explanations
  • Medical context generation
  • Region-specific analysis

🎯 Overview

The system classifies brain MRI scans into four categories with exceptional accuracy:

Tumor Type	Description	Model Performance
Glioma	Most common malignant brain tumor	97% accuracy
Meningioma	Usually benign, slow-growing	90% accuracy
Pituitary	Occurs in pituitary gland	99% accuracy
No Tumor	Healthy brain scan	100% accuracy

🔬 Dataset

The project utilizes the Brain Tumor MRI Dataset from Kaggle:

• Size: 7,023 MRI images
• Resolution: High-quality .jpg format
• Distribution: Balanced across classes
• Augmentation: Applied for robust training

🏗️ Architecture

Transfer Learning Model (Xception)

model = Sequential([
    Xception(include_top=False, weights='imagenet'),
    GlobalMaxPooling2D(),
    Dropout(0.3),
    Dense(128, activation='relu'),
    Dropout(0.25),
    Dense(4, activation='softmax')
])

Custom CNN Architecture

• 6 convolutional layers
• Batch normalization
• Skip connections
• Dropout regularization

📊 Performance Metrics

Xception Model

Accuracy: 96.89%
Precision: 0.97
Recall: 0.96
F1-Score: 0.97

Custom CNN

Accuracy: 95.65%
Precision: 0.96
Recall: 0.95
F1-Score: 0.96

💻 Installation

# Clone repository
git clone https://github.com/sheicky/Brain_Tumor_AI.git
cd Brain_Tumor_AI

# Create virtual environment
python -m venv venv
source venv/bin/activate  # Linux/Mac
venv\Scripts\activate     # Windows

🚀 Usage

Access via Browser

• Run the code on google colab
• Open http://localhost:8501
• Upload MRI scan
• View predictions and analysis

📱 Web Interface Features

• Upload: Drag & drop MRI scans
• Model Selection: Choose between Xception and Custom CNN
• Visualization:
  • Saliency maps
  • Confidence scores
  • Region highlighting
• Analysis: AI-generated medical explanations

📈 Results & Metrics

Our models achieve exceptional performance:

Metric	Xception	Custom CNN
Accuracy	96.89%	95.65%
Precision	0.97	0.96
Recall	0.96	0.95
F1-Score	0.97	0.96

📧 Contact

Sheick - @sheicky

Project Link: https://github.com/sheicky/Brain_Tumor_AI

---

Made with ❤️ by Sheick | Advancing Medical Imaging with AI