R-Scan | Devpost

The device
Lens
Testing
Illuminating the Retina
What the user sees
What the user sees
What the user sees

Inspiration

Did you know that over half of people with Diabetes will develop diabetic retinopathy ? The inspiration for this project is to make a not only cost efficient but time efficient device that has the possibility to detect diabetic retinopathy before it gets to a severe case because during the first stages there are no physical symptoms.

What it does

The device uses a 30 DD lens and a phone camera with flash on to take pictures of your eye using our app and then this image will be fed into a Machine Learning Image classification CNN model that uses efficientNetB0 to classify the image and its severity.

How we built it

We trained the model using Colab, the app using React-Native, and based our physical devices on current retinal scanning equipment but using everyday supplies.

Challenges we ran into

The first challenge we ran into was taking the retinal scans. Retinal scans often require pupil dilation to take clear photos as the pupil is a window to the back of the eyeball. Another challenge we ran into was transferring our model into the app. Although it was working on its own, we had run into difficulties trying to apply it to a different platform. Lastly, building the app posed a big challenge as it was the first time many of our members had used react and as such, had to learn the basics of the language in addition to building an app.

Accomplishments that we're proud of

We are proud that with our supplies we were able to get light into the eye and illuminate the retina, if we had the dilution drops the image would have been clear but getting any sort of insight with our tools and the time had was something to be proud of.

What we learned

The biggest thing we learned was that it can be difficult to integrate different parts together at the end as something is bound to not work or break. Thus, it is important to integrate our different parts throughout the building process, rather than at the end. In addition, we learned about the importance of managing our time and learning when to move onto the different aspects of the project.

What's next for R-Scan

The next step for R-Scan would be making the app have more features such as video taking with flash on and also training the model more so the accuracy rate is higher.