Inspiration

  • Skin cancer is the “easiest to cure if diagnosed and treated early”, according to the Skin Cancer Foundation.
  • Every hour one person dies of Melanoma according to the American Cancer Society.
  • More people are diagnosed with skin cancer each year than every other type of cancer COMBINED.
  • Skin cancer is both incredibly common and lethal and can be diagnosed with a self-screening to decrease the mortality rate.

What it does

  • Derm.AI allows a user to take a picture of a mole and then compare that picture with our Machine Learning Model to predict if that mole is cancerous.
  • Derm.AI can help find a nearby dermatologist and oncologist professional and provide their contact information to schedule an appointment.
  • An easy to use UX and UI for a better experience for users of any age.
  • Users can locate a nearby hospital by just clicking on their address provided in the hospital list.

How we built it

We used the standard ISIC dataset and trained our model over a subset of around 4,000 cancerous and non-cancerous skin cancer images. The image classification model achieved an accuracy of 82%.

We built an iOS app. The model was deployed locally for real-time classification. We also used BetterDoctor’s API to recommend dermatologists near you with all their required details.

Challenges we ran into

  • Gathering required dataset.
  • Attaining a feasible accuracy for the ML model.

Accomplishments that we're proud of

We are really proud of that we were able to come up with a product which can actually assist the people in need and provide this service for free which otherwise is very expensive. We are happy to be able to give back to the community using technology.

What we learned

  • The adverse effects, causes and possible cure for skin cancer.
  • How to identify cancerous moles using image classification.
  • Developing Machine Learning models.
  • iOS development.

What's next for Derm.AI

  • Improve accuracy and precision of our model by factoring in user’s family history of skin cancer as well as increasing the data set we train with.
  • Generate a user report that will be sent to the physician of your choice to allow a professional to double check our findings before you arrive.
  • Utilize our API’s filter by an insurance provider to make sure all suggested doctors accept the user’s insurance.
  • Allow skin care professionals to connect with their current patients through our app

Built With

Share this project:
×

Updates