Multi Class Skin Cancer Detection and Localization Using Deep Learning

Project Domain / Category

Image processing

Abstract / Introduction

Skin cancer is a growing global health concern, with multiple types such as melanoma, basal cell carcinoma (BCC), and squamous cell carcinoma (SCC) requiring early detection for effective treatment. Traditional diagnostic methods are time-consuming, subjective, and dependent on dermatologists' expertise. This project aims to develop an automated deep learning-based system that classifies and detects skin lesions in medical images. The system will not only classify skin cancer types but also localize lesions using object detection techniques. Convolutional Neural Networks (CNNs) and object detection frameworks like YOLO (You Only Look Once) and Faster R-CNN will be used to detect, classify, and highlight affected skin areas in real-time.

Functional Requirements:

1.    Preprocess skin images (noise removal, contrast enhancement, normalization).

2.    Implement multi-class classification for skin lesion types (Melanoma, BCC, SCC, Benign Lesions).

3.    Integrate object detection to localize and highlight affected skin regions.

4.    Utilize deep learning models (ResNet, MobileNet, EfficientNet) for classification.

5.    Apply object detection techniques (YOLO, Faster R-CNN) for lesion localization.

6.    Develop a GUI interface to allow users to upload images for classification.

7.    Evaluate model performance using metrics like accuracy, precision, recall, F1-score, and IoU (Intersection over Union).

8.    Provide visualization tools like Grad-CAM to interpret predictions.

Tools:

Programming Languages: Python

Frameworks & Libraries: TensorFlow, Keras, OpenCV, Scikit-learn Development Environment: Jupyter Notebook, Google Colab Dataset: ISIC (International Skin Imaging Collaboration), HAM10000

https://www.isic-archive.com/ https://www.kaggle.com/datasets/kmader/skin-cancer-mnist-ham10000

Object Detection Models: YOLOv5, Faster R-CNN, SSD (Single Shot MultiBox Detector)

Supervisor:

Name: Laraib Sana

Email ID: Laraib.sana@vu.edu.pk

Skype ID: Laraib.sana