Skin Cancer Detector Web Application

An advanced web-based framework for classifying various skin cancer types using deep learning and TensorFlow.js. Users can upload dermoscopic images of skin lesions to receive predictive insights into potential skin cancer diagnoses.

---

Purpose

This application leverages state-of-the-art computer vision techniques to deliver the top three diagnoses for a given skin lesion.

Note: This tool is for educational purposes and is not intended as a substitute for professional medical consultation or diagnosis.

---

Features

• 🕒 Real-Time Detection: Instantaneous predictions with TensorFlow.js.
• 📋 Multi-Type Classification: Identifies multiple skin cancer types.
• 🖱️ Drag-and-Drop Interface: Easy image upload functionality.
• 🖼️ Demo Images: Evaluate the model's capabilities with sample images.
• 📊 Detailed Predictions: Top 3 classification results with confidence scores.
• 📱 Responsive Design: Optimized for mobile and desktop.
• 📈 Interactive Visualization: View predictions through dynamic charts.

---

Technologies Used

• 🧠 TensorFlow.js: Real-time inference with deep learning.
• 🌐 HTML5/CSS3: For structure and design.
• 📜 JavaScript: Implements client-side logic.
• 📉 Chart.js: Visualizes prediction probabilities.
• 🏗️ MobileNet: Efficient, pre-trained image classification model.

---

Lesion Types

The model identifies the following skin lesion categories:

1. nv: Melanocytic nevi (benign melanocyte neoplasms).
2. mel: Melanoma (malignant melanocyte neoplasms).
3. bkl: Benign keratosis lesions.
4. bcc: Basal cell carcinoma.
5. akiec: Actinic keratoses/intraepithelial carcinoma.
6. vasc: Vascular lesions.
7. df: Dermatofibroma (benign cutaneous lesion).

Descriptions are derived from dermatological research.

---

Disclaimer

This tool is designed for educational purposes only. It is not certified for medical diagnostics. Consult qualified healthcare professionals for medical concerns.

---

Getting Started

Prerequisites

• 🌐 A modern web browser (e.g., Chrome, Firefox, Safari, Edge).
• 📶 An active internet connection.

Running Locally

1. Clone the repository:

   git clone https://github.com/Arianrezaz/SkinCancerTypeDetection.git
   cd SkinCancerTypeDetection

2. Launch a local server:

   Using Python:

   python -m http.server 8000

   Using Node.js:

   npx http-server

3. Open the application in your browser:

   • Python: http://localhost:8000
   • Node.js: http://localhost:8080

---

Usage

1. Demo Testing:
   • Select "Use Demo Image" to test pre-loaded images.
2. Image Upload:
   • Drag and drop or click to upload a dermoscopic image.
3. Generate Predictions:
   • Click "Detect" for classification results.
4. Visualize Results:
   • Review the interactive chart for probability distribution.

---

Model Information

The application uses a fine-tuned MobileNet model trained on dermatological datasets, including:

• ISIC Dataset
• HAM10000 Dataset

Datasets Used

Public datasets (ISIC and HAM10000) are used under a CC BY-NC-SA 4.0 license. The datasets contain labeled dermoscopic images for training and validation.

---

Screenshots

Home Page

Benign Results

Malignant Results

---

Contributing

Contributions are welcome! Submit pull requests to propose changes or address issues.

---

Acknowledgments

• 💡 Special thanks to the TensorFlow.js team.
• 🏗️ Acknowledgment to MobileNet developers.
• 📊 Recognition of ISIC and HAM10000 dataset contributors.

---

Contact

Arian Rezazadeh
Email: [email protected]