Skin Cancer Detection.

This project presents a deep learning-based image classification system to detect and classify various types of skin cancer using dermatoscopic images. Leveraging Convolutional Neural Networks (CNN), this project aims to provide an efficient and accurate tool to assist dermatologists in early diagnosis of skin cancer.

---

🌐 Project Overview

The classification model is trained on the HAM10000 dataset containing dermatoscopic images of skin lesions. The pipeline includes data preprocessing, augmentation, class balancing using SMOTE, CNN-based model building, and performance evaluation using multiple metrics.

---

📃 Dataset

• Source: HAM10000 - Human Against Machine with 10000 training images
• Contains over 10,000 dermatoscopic images categorized into seven skin disease types:
  • Melanocytic nevi
  • Melanoma
  • Benign keratosis-like lesions
  • Basal cell carcinoma
  • Actinic keratoses
  • Vascular lesions
  • Dermatofibroma
• Includes patient metadata like age, sex, and localization

---

📊 Features

• CNN Model Architecture:
  • Includes Conv2D, MaxPooling2D, Dropout, BatchNormalization, and Dense layers
• Data Preprocessing:
  • Label encoding and one-hot encoding
  • Handling missing values
  • SMOTE (Synthetic Minority Over-sampling Technique) to handle class imbalance
• Training Optimization:
  • Image augmentation using ImageDataGenerator
  • Learning rate scheduling using ReduceLROnPlateau
• Evaluation Metrics:
  • Confusion matrix
  • ROC-AUC Score
  • Classification Report

---

🎓 Technologies Used

• Python
• TensorFlow/Keras
• Scikit-learn
• Pandas & NumPy
• Matplotlib & Seaborn
• SMOTE (from imblearn)

---

🚀 How to Run the Project

1. Clone the repository:

git clone https://github.com/your-username/skin-cancer-detection.git
cd skin-cancer-detection

2. Install dependencies:

pip install -r requirements.txt

3. Download the dataset and place it in the project folder
4. Run the Jupyter Notebook:

jupyter notebook project.ipynb

---

🚀 Sample Results

• Visualizations of class distributions and metadata
• High accuracy in skin cancer classification
• ROC-AUC and classification reports validate model performance

---

🙌 Acknowledgments

• Dataset courtesy of Kaggle (HAM10000)
• Inspired by applications of AI in dermatology

---

🚀 Future Enhancements

• Deployment as a web or mobile application
• Use of advanced architectures like ResNet or EfficientNet
• Integration with patient history for context-aware diagnosis

---

👤 Author

Litesh Perumalla
Master's in Data Science | University of North Texas
Python | Machine Learning | Deep Learning | Data Analysis

Feel free to connect or raise an issue if you have suggestions or questions!

---

This project is part of my portfolio to showcase deep learning applications in the healthcare domain.