🧬 Identifying Colon Cancer Using Deep Learning

This project leverages Convolutional Neural Networks (CNNs) to automatically identify colon cancer from histopathological images. It is designed for reproducibility, deployment, and accessibility on multiple platforms.

---

📌 Project Links

• 🔗 GitHub Repository: Identifying-Colon-Cancer-Using-Deep-Learning
• 📊 Kaggle Notebook: Colon Cancer Classifier on Kaggle
• 🐳 Docker Image: Coming soon

---

📁 Project Structure

├── Colon_cancer.ipynb        # Jupyter Notebook with training pipeline
├── train.csv                 # Dataset with labeled image paths for training
├── pred.csv                  # Dataset with image paths for inference
├── best_checkpoint.model     # Saved PyTorch model checkpoint
├── graphs/                   # Visualizations and training plots
├── example.csv               # A sample dataset format (for reference)
├── requirements.txt          # Dependencies file (install this first)
├── Dockerfile                # Docker configuration to replicate environment
└── README.md                 # Project documentation

---

🔬 Project Pipeline

1. Data Preparation

   • train.csv: labeled images (image_id, label)
   • pred.csv: for model inference
   • Image preprocessing includes resizing, normalization, and augmentation.

2. Model Architecture

   • CNN built using PyTorch or with transfer learning (e.g., ResNet18)
   • Loss: CrossEntropyLoss, Optimizer: Adam
   • Checkpointing used to store best model

3. Evaluation

   • Accuracy, Precision, Recall metrics
   • Visuals: Loss vs Epoch, Accuracy vs Epoch
   • Prediction samples visualized for interpretability

---

⚙️ Setup and Installation

✅ Option 1: Manual Setup

1. Clone the repository:

git clone https://github.com/Arpangpta/Identifying-Colon-Cancer-Using-Deep-Learning.git
cd Identifying-Colon-Cancer-Using-Deep-Learning

2. Set up virtual environment:

python -m venv venv
source venv/bin/activate      # Windows: venv\Scripts\activate

3. Install dependencies:

pip install -r requirements.txt

📄 View requirements.txt

4. Launch Jupyter Notebook:

jupyter notebook Colon_cancer.ipynb

---

🐳 Option 2: Docker-Based Setup

1. Build the Docker image:

docker build -t colon-cancer-dl .

2. Run the container:

docker run -p 8888:8888 colon-cancer-dl

Access Jupyter Notebook at localhost:8888

---

🧪 Dataset Format

• train.csv — Columns: image_id, label
• pred.csv — Column: image_id (no labels)
• Example schema available in example.csv

---

📉 Results & Graphs

🔁 Accuracy vs Epoch

📉 Loss vs Epoch

---

📊 Demo Predictions

Image Sample	True Label	Predicted Label
	Non-Cancer	Non-Cancer
	Cancer	Cancer

---

🧠 Model Checkpoint

The best trained model is saved as:

best_checkpoint.model

You can load it in PyTorch using:

import torch
model = torch.load("best_checkpoint.model")
model.eval()

---

👥 Author

• 👤 Arpan Gupta
• 📓 Notebook: Kaggle

---

📄 License

This project is licensed under the MIT License. See the LICENSE file for details.