Denny Loevlie

Denny Loevlie

ELLIS PhD student interested in multimodal tabular foundation models

SkinsAI: Melanoma Detection Tool

Published:

SkinsAI Logo

Overview

SkinsAI is a free-access diagnostic tool that helps users classify moles as benign or malignant using deep learning. Developed during The Pitt Challenge Hackathon in Fall 2022.

Demo Video

Problem

Early detection of melanoma is critical for successful treatment, but access to dermatologists can be limited. SkinsAI provides an accessible screening tool to help identify potentially dangerous moles.

Solution

  • Convolutional Neural Network: Built and trained in PyTorch for image classification
  • Web Interface: User-friendly interface for uploading and analyzing mole images
  • Free Access: No cost to users, democratizing skin cancer screening

Technical Details

  • Framework: PyTorch for model development
  • Model: Custom CNN architecture trained on medical imagery
  • Deployment: Django web framework, hosted on Heroku

Recognition

  • 2nd place out of 24 teams at The Pitt Challenge Hackathon ($2,000 prize)
  • Recognized for impact on healthcare accessibility

Impact

SkinsAI demonstrates how deep learning can be applied to critical healthcare problems while maintaining accessibility. The tool helps users make informed decisions about seeking professional medical evaluation.

Disclaimer: This tool is meant for screening purposes only and should not replace professional medical diagnosis.