Colon Cancer Detection Using CNN

This repository contains the code and trained models for Colon Cancer Detection using Convolutional Neural Networks (CNNs). The model is trained to classify histopathological images as Colon Adenocarcinoma (Cancerous) or Colon Benign Tissue (Non-Cancerous).

Model Download:

To download Models, Gmail me at "syedmuhammadhassan367@gmail.com"!

Project Overview

This project aims to develop an AI-based classifier that helps in early detection of colon cancer by analyzing medical images. The deep learning model was trained on a dataset of Colon Adenocarcinoma and Colon Benign Tissue images, achieving high accuracy.

Repository Structure

📦 Colon-Cancer-Detection
│── 📂 notebooks/                     # Jupyter Notebooks for training and evaluation
    │── 📜 colon-c-training.ipynb               # Model Traing Notebook
    │── 📜 Report Testing.ipynb                 # Model Evaluation (Accuracy, loss, and confusion matrix analysis)
│── 📂 Report/                        # Trained models testing reports
│── 📂 Training Graph/                # Model performance graphs
│── 📜 requirements.txt               # Python dependencies
│── 📜 README.md                      # Project documentation (this file)

Features

✅ Binary Classification: Detects whether an image is Cancerous or Benign ✅ Deep Learning Model: CNN-based architecture trained using TensorFlow/Keras ✅ Data Augmentation: Improves model generalization ✅ Model Evaluation: Accuracy, loss, and confusion matrix analysis ✅ Pre-trained Models: Ready to use for inference

Models Performance

1. C_normal_1.h5 (Normal Training)

• Training Accuracy: 93.54%
• Validation Accuracy: 93.78%
• Test Accuracy: 94.89%

2. C_large_1.h5 (Data Augmentation with Droupout)

• Training Accuracy: 97.70%
• Validation Accuracy: 93.78%
• Test Accuracy: 93.33%

3. C_WDA_1.h5 (Fast Feature Extraction Without Data Augmentation)

• Training Accuracy: 100%
• Validation Accuracy: 99.56%

4. C_DA_1.h5 (Feature Extraction with Data Augmentation: 1st Try)

• Training Accuracy: 100%
• Validation Accuracy: 99.89%
• Test Accuracy: 99.89%

5. C_DA_2.h5 (Feature Extraction with Data Augmentation: 2st Try)

• Training Accuracy: 100%
• Validation Accuracy: 100%
• Test Accuracy: 100%

Model performance has been validated using accuracy/loss curves, confusion matrices, and classification reports!

Installation & Setup

Clone the Repository

git clone https://github.com/Muhammad-Hassan12/Colon-Cancer-Prediction-CNN-Model.git
cd Colon-Cancer-Prediction-CNN-Model

Install Dependencies

pip install -r requirements.txt

Run the Training Notebook (If you want to train your self!)

jupyter notebook colon-c-training.ipynb

Run the Model on Test Data for Evaluation

jupyter notebook Report Testing.ipynb

Dataset

The Dataset I used is "Lung and Colon Cancer Histopathological Images" by "Larxel" from Kaggle! You can download it yourself: https://www.kaggle.com/datasets/andrewmvd/lung-and-colon-cancer-histopathological-images The dataset consists of Colon Adenocarcinoma and Colon Benign Tissue images. The images are resized to 256x256 before training.

Model Training Details

• Framework: TensorFlow/Keras
• Optimizer: RMSprop
• Loss Function: Binary Cross-Entropy
• Evaluation Metrics: Accuracy, Precision, Recall, F1-score

Results and Visualizations

1. C_normal_1.h5

1. Training & Validation Accuracy & Loss:

• 

2. Confusion Matrix:

• 

2. C_large_1.h5

1. Training & Validation Accuracy & Loss:

• 

2. Confusion Matrix:

• 

3. C_WDA_1.h5

1. Training & Validation Accuracy & Loss:

• 

4. C_DA_1.h5

1. Training & Validation Accuracy & Loss:

• 

2. Confusion Matrix:

• 

5. C_DA_2.h5

1. Training & Validation Accuracy & Loss:

• 

2. Confusion Matrix:

• 

Example Predictions

• Input: Colon biopsy image
• Model Output: Probability score (Cancerous vs. Benign)

Future Improvements

🔹 Improve dataset diversity 🔹 Experiment with different CNN architectures 🔹 Deploy the model as a web app for easy access

License

This project is open-source under the MIT License. Feel free to contribute!

Special Thanks To!

Kaggle Cloud Jupyter Notebook!