MedicAI

Home Page
Contact Page Hover
Contact Page
Contact link
Localization Dropdown
Backend Demo

Inspiration

Coming into this project we wanted a way to address today's health concerns. After some digging we came accross MNIST's Humans against Machines 10000 dataset which is a dataset of 10000 pictures of cancerous and benign skin lessions. We took this opportunity to train a Neural Network model that can predict cancer on a few given datapoints.

What it does

Our website allows users to upload their image along with their age, sex, and localization (where the lession is located on the body) and our systems then report back the type of lession it may be. Not all lessions are cancerous so it may give piece of mind to identify these lessions.

How we built it

We made our backend using Python Flask and trained our Neural Network model using Tensorflow. Our frontend was made with HTML and CSS. Flask was used to render the frontend.

Challenges we ran into

We had lots of trouble processing data from the MNIST dataset, setting up photo upload api for our frontend and running our tensorflow model within our flask app. For this reason we were not able to fully complete the project. We were about 80% complete.

Accomplishments that we're proud of

Our machine learning model is able to predict the type of lession given with about 77% accuracy. Working with large picture datasets is more difficult than expected.