Sweet Heart

• 

  Home Screen

• 

  Parent View

• 

  Child View

• 

  Child Data

• 

  Parent View with Houndify

• 

  Blood Glucose vs Time

• 

  Fourier Transform

• 

  Combined Fourier & Morlet Analysis (reveals regularity in data)

• 

  Morlet Reconstruction Contours

• 

  4th Level Haar Transform

Inspiration

Approximately 1.25 MILLION children live with Type-I diabetes in the United States ALONE. These children bear the trials and tribulations of a painful and often socially alienating disease. However, parents of young children with diabetes have it arguably worse. They live with the fear that their little "sweetheart" could have severe health repurcussions for something as simple as forgetting to count calories or measure glucose levels. We live in a highly connected world, and we decided we could leverage this connectedness to bring some peace of mind to parents of children with diabetes.

What it does

Sweetheart gives parents a direct link to the health of their child. Concerned moms and dads interface with our natural language interface to get Real Time vital data about their children (from the OTR API). Using voice recognition and natural language processing from the Houndify API, they can ask questions about the status of their children (sends back latest blood glucose level and visualization of recent readings) and remind forgetful children to measure(sends message to child if not). On the other hand, the child will have a daily checklist of tasks to complete, such as having testing strips, counting his calories, and measuring his blood sugar levels, and it also provides a day streak for the child's progress. When the child completes their tasks, the application automatically sends an sms message to his/her parent(s) informing them of their actions.

Essentially, it is a simple, smart, and intuitive way for parents to get the info they want about their children's diabetes and for the child to be reminded of what he or she needs to do.

How we built it

We built a native Android application implementing the Houndify and Johnson & Johnson OTR APIs. On the backend, we also used Parse.

Furthermore, we used Matlab and R to create the visualizations and exploratory analysis of the dataset.

Challenges we ran into

The user design and experience part of the Android app development itself was quite challenging, as it required a lot of constant testing and tweaking in order for it to match what we wanted.

Accomplishments that we're proud of

Figuring out how to use the Johnson & Johnson OTR API was a major stepping point in our project, as it enabled us to have access to lots of data for our app and analysis.

What we learned

When we first approached the Johnson & Johnson data, we took a simple plot of the blood glucose concentration against time, as presented in the gallery, to gain insight about structure of the data. Upon realizing that the data was incredibly volatile, we further implemented Fourier analysis in Matlab, also in the gallery, which only showed a huge spike at the origin and a relatively even distribution of frequencies elsewhere. Then, upon conducting online research, we implemented a Morlet wavelet transform, split into 6 frequency densities. Upon examination of the 4th and 5th density curves, we started finding more regularities in the curves, as shown in the gallery. In the future we plan to implement a neural network on the curve histories to try to predict future anomalous blood glucose concentrations.

In terms of building an app, we learned a lot about the complexities of implementing multiple APIs in Android.

What's next for Sweetheart

We hope to make Sweetheart have a more user-friendly interface with more options on the parent and child view points. In addition, we hope to implement several other APIs, like the Postmates for delivering strips when the child runs out.

We plan to implement a neural network on the curve histories to try to predict future anomalous blood glucose concentrations.

Built With

• android
• houndify-api
• otr-api
• parse