Quantum XAI for Brain Tumor Detection

A Hybrid Quantum-Classical Explainable AI Framework for Brain Tumor Detection using ResNet50, Variational Quantum Circuits (VQC), Grad-CAM, and Integrated Gradients. This project combines Deep Learning, Quantum Machine Learning, and Explainable AI to provide accurate and interpretable brain tumor diagnosis from MRI images.

---

📌 Features

• Brain MRI Tumor Detection
• Hybrid Quantum-Classical Architecture
• ResNet50 Feature Extraction
• Variational Quantum Circuit (VQC) Classification
• Explainable AI using Grad-CAM
• Explainable AI using Integrated Gradients
• Performance Evaluation using multiple metrics
• Quantum Circuit Visualization
• Training & Evaluation Plots

---

📂 Project Structure

Quantum-XAI-Brain-Tumor-Detection/
│
├── Dataset/
├── notebooks/
│   └── Quantum_XAI_Brain_Tumor.ipynb
│
├── images/
│   ├── architecture.png
│   ├── gradcam.png
│   ├── integrated_gradients.png
│   ├── overlay_visualization.png
│   └── quantum_circuit.png
│
├── models/
│   └── hybrid_quantum_model.pth
│
├── results/
│   ├── confusion_matrix.png
│   ├── roc_curve.png
│   ├── accuracy_plot.png
│   └── loss_plot.png
│
├── README.md
└── requirements.txt

---

Architecture

Brain MRI Dataset
        │
        ▼
Image Preprocessing
        │
        ▼
Data Augmentation
        │
        ▼
Train-Test Split
        │
        ▼
Feature Extraction (ResNet50)
        │
        ▼
Feature Reduction
        │
        ▼
Variational Quantum Circuit
        │
        ▼
Tumor Classification
        │
        ▼
Grad-CAM & Integrated Gradients
        │
        ▼
Performance Evaluation
        │
        ▼
Final Diagnosis

---

Dataset

Dataset Name

Brain MRI Images for Brain Tumor Detection

Source

Kaggle

Dataset Statistics

Property	Value
Total Images	253
Classes	2
Tumor Images	155
Non-Tumor Images	98
Image Format	JPG
Original Image Size	Variable
Input Size	224 × 224 × 3

---

Technologies Used

• Python
• PyTorch
• Torchvision
• PennyLane
• Quantum Machine Learning
• ResNet50
• Variational Quantum Circuits
• Grad-CAM
• Captum
• Integrated Gradients
• Matplotlib
• Seaborn
• Scikit-Learn

---

Model Workflow

1. Brain MRI Dataset Collection
2. Image Preprocessing
3. Data Augmentation
4. Train-Test Split
5. Feature Extraction using ResNet50
6. Feature Reduction
7. Variational Quantum Circuit Classification
8. Tumor Prediction
9. Explainability using Grad-CAM
10. Explainability using Integrated Gradients
11. Performance Evaluation

---

Evaluation Metrics

• Accuracy
• Precision
• Recall
• F1-Score
• ROC-AUC Score
• Confusion Matrix

---

Explainable AI

The model incorporates Explainable AI (XAI) techniques to improve transparency and trust in medical diagnosis.

Grad-CAM

• Highlights important tumor regions.
• Generates activation heatmaps.
• Improves model interpretability.

Integrated Gradients

• Computes pixel-level feature importance.
• Produces attribution maps.
• Explains model predictions.

Overlay Visualization

• Combines the original MRI image with the attribution map.
• Shows the exact regions influencing the prediction.

---

Quantum Machine Learning

The project uses a 6-Qubit Variational Quantum Circuit (VQC).

Quantum Pipeline

Reduced Features
        │
        ▼
Angle Embedding
        │
        ▼
Strongly Entangling Layers
        │
        ▼
Pauli-Z Measurements
        │
        ▼
Quantum Outputs

---

Results

The proposed hybrid model achieved promising performance for brain tumor classification.

Metric	Score
Accuracy	80.39%
Precision	80.77%
Recall	80.77%
F1-Score	80.77%
ROC-AUC	89.53%

---

Installation

Clone the repository

git clone https://github.com/your-username/Quantum-XAI-Brain-Tumor-Detection.git

cd Quantum-XAI-Brain-Tumor-Detection

Create virtual environment (Optional)

python -m venv venv

Activate environment

Windows

venv\Scripts\activate

Linux / Mac

source venv/bin/activate

Install dependencies

pip install -r requirements.txt

---

Run the Project

Launch Jupyter Notebook

jupyter notebook

or

jupyter lab

Open

Quantum_XAI_Brain_Tumor.ipynb

Run all cells.

---

Required Libraries

pip install torch torchvision
pip install pennylane
pip install captum
pip install grad-cam
pip install matplotlib
pip install seaborn
pip install scikit-learn
pip install numpy
pip install pandas
pip install pillow

---

Future Improvements

• Multi-class brain tumor classification
• Real Quantum Hardware Implementation
• Quantum Transformers
• Streamlit Web Application
• Clinical Decision Support System
• SHAP & LIME Explainability
• Cloud Deployment

---

License

This project is developed for academic and research purposes.