LungCancerDetection

Lung Cancer Detection Using Deep Learning

This project focuses on building a deep learning-based system to detect lung cancer from CT scan images. The goal is to leverage Convolutional Neural Networks (CNN) and transfer learning models such as ResNet50, VGG16, and MobileNetV2 to assist in early cancer detection and classification.

---

Features

• Handles large-scale CT scan image datasets
• Supports multiple CNN architectures (ResNet50, VGG16, MobileNetV2)
• Detects and classifies cancerous vs non-cancerous lung tissues
• Visualizes model performance (accuracy, confusion matrix, loss curves)
• Includes trained models for quick inference
• Modular code — easy to extend or retrain with custom datasets

---

Project Structure

LungCancerDetection/
│
├── data/                     # Dataset (not included in repo)
│   ├── train/
│   ├── test/
│
├── models/                   # Saved trained models (.pkl / .h5)
│   ├── model_resnet50.pkl
│   ├── model_vgg16.pkl
│
├── notebooks/                # Jupyter notebooks for training & analysis
│   ├── lung_cancer_training.ipynb
│
├── src/                      # Python source code
│   ├── data_preprocessing.py
│   ├── train_model.py
│   ├── predict.py
│
├── requirements.txt          # Python dependencies
└── README.md                 # Project documentation

---

Models Used

Model	Description	Framework
CNN (Custom)	Basic convolutional architecture built from scratch	TensorFlow/Keras
ResNet50	Deep residual network (50 layers) for high accuracy	TensorFlow/Keras
VGG16	Simpler deep CNN architecture	TensorFlow/Keras
MobileNetV2	Lightweight CNN for mobile deployment	TensorFlow/Keras

---

Tech Stack

• Language: Python
• Libraries: TensorFlow, Keras, NumPy, OpenCV, Matplotlib, Scikit-learn
• IDE: Jupyter Notebook / VS Code
• Frameworks: Deep Learning (CNN, Transfer Learning)

---

Installation and Setup

1. Clone the repository

   git clone https://github.com/gschandrasekhar/LungCancerDetection.git
   cd LungCancerDetection

2. Create a virtual environment (optional)

   python -m venv venv
   venv\Scripts\activate    # On Windows

3. Install dependencies

   pip install -r requirements.txt

4. Run the notebook or script

   jupyter notebook notebooks/lung_cancer_training.ipynb

   or

   python src/train_model.py

---

Results

Model	Accuracy	Notes
CNN (Custom)	85%	Baseline model
VGG16	91%	Moderate accuracy, higher training time
ResNet50	94%	Best performing model
MobileNetV2	92%	Lightweight, good trade-off

(Results may vary depending on dataset and training parameters.)

---

Future Work

• Deploy model as a web app using Streamlit or Flask
• Integrate Grad-CAM visualization for explainability
• Add real-time prediction from uploaded CT images
• Optimize for mobile and edge devices

---

References

• Kaggle: Lung Cancer CT Scan Dataset
• Research Paper: “Deep Learning for Lung Cancer Detection Using CT Scans”
• TensorFlow / Keras Documentation

---

🧑‍💻 Author

Chandra Sekhar G 📧 Email Me 🌐 LinkedIn 💻 GitHub

---

🪪 License

This project is licensed under the MIT License — feel free to use and modify for research or learning purposes.