Thea

Inspiration

We are all interested in the intersection between technology and medicine, primarily in using machine learning to aid chronic disease management.

What it does

Thea is a pre-screening tool for diabetics who want to gain insight on diabetic retinopathy using non-invasive, at-home technology. This is not a substitution for an eye doctor's visit nor a diagnostic tool by any means but simply serves as a way to empower diabetic patients in taking control of their condition.

How I built it

-Conducted user research on the diabetic population (with and without diabetic retinopathy) to gain insight on how this web app could best benefit them. -Mocked-up wireframes and high fidelity prototypes using Sketch, Invision, Zeplin.

-Translated these designs into working web pages using HTML5, CSS3, and Javascript.

Challenges I ran into

-Converting a high fidelity prototype into a working web product in such a short amount of time was challenging and required compromises.

• We wanted to create a cohesive web application that had many components to it, but ultimately found that it was difficult to ensure that each part of our application was to completion -Division of labor could have been more clearly communicated and distributed amongst team members ## Accomplishments that I'm proud of -Built a highly functional machine learning model within the limited time frame
• Implemented an image-based machine learning model based on convolutional neural networks for the first time
• Learned how to use Flask for python
• One of our team members is a designer and had very limited coding experience, she ended up building almost all of the front-end portion of our application!
• Made new friends! ## What I learned
• To be a high performing team, it's very helpful to have people who are specialized in different areas (front-end, back-end, data science)
• It's better to be realistic with your goals and know what each person can accomplish.

What's next for Thea

• We would like to continue with our prototyping and implementation of a more interactive eye exam that's more custom tailored to retinopathy patients.
• Would like to improve on the machine learning algorithm's image preprocessing steps as well as trying other models.
• It would be great to turn Thea into a patient portal with patient login access and historical data.

Built With

• cnn
• css3
• fastai
• flask
• html5
• javascript
• machine-learning
• matplotlib
• pandas
• python
• sketch