Anaphlax

• 

  Anaphlax - 3D Model

• 

  Data

• 

  Logistic Regression

Inspiration

I’m allergic to a lot of things, like nuts, goat cheese, and some other strange things. For nuts and goat cheese, I go into anaphylactic shock and for several hours I’m in severe pain and my throat tightens up. The most recent event was in Spain when I didn’t have an EpiPen because it was so hard to carry around. I wanted to have a smaller device that would always be with me without me knowing.

What it does

Our device collects data (Blood Pressure, BPS, Maximum Heart Rate, Electrocardiographic results, etc.) and checks whether the person is going into anaphylactic shock based on our machine-learning model. It then notifies the user that it is injecting adrenaline and activates a small spring-loaded piston which administers the adrenaline.

How we built it

In order to design this device, we used Fusion 360, Scikit, and Jupyter Notebook in order to cad prototypes and train our machine learning model. To train our model we used a labeled data set and supervised machine learning to generate a logistic regression.

Challenges we ran into

Machine learning and using deep learning was really hard, but Scikit and Jupyter Notebook as well as others at the Hackathon helped us to learn. We struggled to find data sets for our project.

Accomplishments that we're proud of

In a day and a half, we created a whole model of our device. We became well-researched on how it worked and went from nearly no machine learning knowledge to being able to train a neural network to determine if somebody is going into shock given data about their heart rate, blood pressure, etc.

What we learned

Machine learning is powerful and can easily be learned, given the right tools and curiosity. Taran showed me how to use Fusion 360, a Cad-like tool for creating 3-D models of what you want to make. We both learned how to create neural networks and pump in data.

What's next for Anaphlax

We want to train our machine learning model further and create an app side for the device, to make it more user-friendly and up-to-date. We also would like to refine our actual design of the project.

Built With

• autodesk-fusion-360
• jupyter
• python
• scikit-learn
• vscode