Inspiration

6.5 million people in the US suffer from chronic wounds which results in $25 billion spent annually on wound care. The average cost of outpatient wound care is $5500. We hope to reduce the cost of outpatient wound care through accurate diagnosis of the severity of wounds.

What it does

We are starting off with a diagnosis of burn wounds based on the different degrees of burn. We allow users to take a picture of their burn wounds and we can advise them if they require professional care or self-care. We can also advise them on the appropriate treatments based on the severity of the burn and connect them to a doctor immediately through our platform.

How I built it

We built it using a React frontend and a Python backend.

On the frontend, we built a React app to allow users to upload images of their burn wounds. They would then receive a diagnosis of their wound severity and some advice on further medical action.

On the backend, we deployed on Microsoft Azure. We also used Azure's Custom Vision service to train our burn wound severity classification model. The image that the user upload would be given to the model to infer the severity of the wound. We made use of the Flask framework to send & receive HTTP requests to & fro the Custom Vision service and our React app.

Challenges I ran into

We encountered challenges in various areas. Initially, we wanted to train our own models with Tensorflow & Keras. However, we had difficulties deploying these libraries onto our Flask app on Microsoft Azure. Thus, we had to look around for an alternative and found the Custom Vision service which is provided on Microsoft Azure. It allowed us to train a burn wound classification model in a short time and thus we avoided wasting time trying to deploy our own models.

There were also difficulties in figuring out how to design the classification model and the use cases. We had to reduce our expectations and focus on differentiating 1st and 3rd degree burns only in order to make it in time for the deadline.

Accomplishments that I'm proud of

We have successfully built and trained our models to test for burn victims and deployed our apps on Microsoft Azure.

What I learned

We came out of this hackathon with a better understanding of wound care for burn victims and could properly identify the different degrees of burns. On the technical side, we have gained new knowledge on TensorFlow, collaboration on Github, Microsoft Azure, React and Flask.

What's next for CrazyRichAsian

We plan to expand & improve our models to diagnose various wounds and provide more detailed care information. We are also considering further integrations with healthcare providers to provide on-demand telehealth services. If things go well, we hope to venture into clinical trials and FDA approval.

Built With

Share this project:

Updates