🩺 Melanoma Detection Using Hybrid Features & Machine Learning

A hybrid-feature melanoma detection pipeline combining advanced image preprocessing, handcrafted features, and multiple machine learning classifiers. Designed to deliver accurate and efficient skin cancer recognition

A complete machine learning pipeline for melanoma classification using hybrid texture, color, morphological, and intensity features All executable code exists inside one Python file

---

📑 Table of Contents

1. Project Overview
2. Project Structure
3. Dataset Description
4. Methodology
5. Results
6. How to Run
7. Future Work
8. References
9. License

---

Project Overview

This project implements an automated melanoma detection system using:

• Hybrid feature extraction (HOG + LBP + Color + Shape)
• Classical ML models (SVM, Logistic Regression, Decision Tree, KNN)
• Performance evaluation using accuracy, precision, recall, F1-score, and confusion matrices

All code is located in:
👉 Notebook/melanoma.py

---

Project Structure

---

Dataset Description here

Dataset Folder Includes:

• Colored/ — Original skin lesion images
• labels.xls — Actual class labels
• features.csv — Pre-extracted combined features:
  • HOG
  • LBP
  • Shape features (area, perimeter, compactness, eccentricity)
  • Color statistics (mean & std)

Using precomputed features greatly speeds up training.

---

Methodology

1. Preprocessing

• Resize images
• Noise removal
• Convert to grayscale
• Segmentation
• Feature normalization

2. Feature Extraction

• HOG — texture
• LBP — micro-texture
• Color features — RGB means & standard deviations
• Shape features — area, perimeter, eccentricity, compactness

3. Machine Learning Models

• SVM
• Logistic Regression
• Decision Tree
• KNN

Evaluation Metrics

• Accuracy
• Precision
• Recall
• F1-score
• Confusion Matrix

---

Results

Model	Accuracy	Precision	Recall	F1-Score
SVM	0.78	0.82	0.75	0.77
Logistic Regression	0.77	0.81	0.70	0.74
Decision Tree	0.75	0.72	0.70	0.71
KNN	0.70	0.72	0.70	0.71

Key Insight:

• Perimeter is a highly important shape feature → chosen as the root node in Decision Tree.

---

How to Run

1. Install dependencies

pip install -r requirements.txt

2. Run the main script

python Notebook/melanoma.py

---

Future Work

Add CNN-based deep learning

Add Random Forest & XGBoost

Build a Streamlit UI

Add explainability (Grad-CAM, SHAP)

Deploy the system using FastAPI or Flask

References

Full research documentation:

📄 References/Report.pdf

License

This project is licensed under the MIT License..

You are free to modify and use it with attribution