MelaScan - AI Powered Melanoma Detection

Overview

An AI-powered mobile application for early detection of melanoma skin cancer using image classification.

MelaScan is an AI-powered melanoma detection application that enables users to assess their nevi's skin cancer risk without visiting a doctor. The app aims to reduce the cost and need for medical consultation in the pre-diagnosis stage of melanoma, the 5th most common cancer. Users can take photos of their nevi using the mobile application, which then utilizes a CNN classifier to analyze the images and provide risk assessments with visual explanations.

System Architecture

MelaScan employs a three-layer architecture:

• AI Layer: Utilizes a custom-trained convolutional neural network (CNN) model to analyze skin lesion images and classify them based on melanoma risk factors.
• Mobile Application Layer: A React Native application that provides the user interface for capturing images, displaying results, and managing user profiles.
• Backend Layer: A Spring Boot application that handles authentication, stores user data and scan history, and communicates with the AI model for image processing.

The layers communicate through RESTful APIs, with the mobile app sending images to the backend, which then processes them through the AI model and returns the results to be displayed to the user.

Features

• AI-powered melanoma risk assessment using CNN classification
• Image preprocessing and segmentation for accurate detection
• Visual explanations with Grad-CAM, Heat Map, and Saliency Map
• User-friendly mobile interface built with React Native
• Secure backend with Spring Framework and JWT authentication
• Cloud storage integration with AWS S3
• Traditional ABCDE approach combined with deep learning methods