Skin cancer is one of the world's most deadly diseases, causing over 50,000 deaths worldwide per year. Most of these deaths are due to late or wrong diagnosis of this disease. If diagnosed early and efficiently, it can be fully cured. We believe that machine learning can be used for this purpose of diagnosing skin cancer.
What it does
Our project is a mobile app that uses machine learning to diagnose skin cancer from a picture taken by the user. We provide information about over 7 classes of skin cancer in our predictions. We also have a google maps page, where the user can find nearby hospitals or cancer specialists for further treatment. Lastly, we have a patient data management system for storing patient records.
How I built it
- React native for mobile app front-end
- Keras + TensorFlow CNN model for diagnosis (> 85% accuracy on test dataset)
- MongoDB Atlas to store user information and patient records
- Google Cloud for hosting our backend and endpoints
- Google places API to retrieve nearby hospitals.
Challenges I ran into
- Getting the model to a high accuracy
- Cloud hosting
- Integrating with react native.
Accomplishments that I'm proud of
- Creating an effective model using TensorFlow.
- Creating a clean UI
- Integrating google cloud.
What I learned
- google cloud hosting
- Google places API
- TensorFlow CNN models
- React Native.
What's next for LesionAI
- Higher accuracy model
- Better patient data system
- More training data.