FolDetect

Inspiration

We were inspired by watching a Ted Talk by Manu Sharma, the founder of the FoldScope idea. We were amazed by how accessible this device was, especially for vulnerable and/or underprivileged populations who might not have easy access to lab tests specifically designed for disease diagnosis.

What it does

Our app uses a machine learning model to analyze a simple blood sample and predicts by looking at a blood cell whether the user is positive or negative for malaria.

How we built it

We used the CreateML platform in order to build an image classification model, which we deployed into an iOS app using Swift.

Challenges we ran into

Initially, we planned to code a Convoluted Neural Network using the Keras library from Tensorflow, or by using SciKit Learn. However, we soon found out that the layers and connections of the CNN became too complicated for us to understand and implement in a short time frame. Also, these platforms were much harder to deploy into either an iOS or Android app. Therefore, we decided to switch to using the CreateML platform to train a model and focused on integrating it into our iOS app. We also faced difficulties with UI - specifically, we had issues with distorted appearances of our UI on the Simulator or testing device. However, with the help of the mentors and some extra research, we were able to fix these issues.

Accomplishments that we're proud of

We are proud of being able to give the user an option to take their own picture of their microscopic sample or use existing pictures from their photo gallery. As beginners working with Swift and CreateML, this was a very good learning experience for us. We are also proud of being able to integrate the machine learning model with our app itself, and being able to present a complete prototype on an actual iOS device.

What we learned

We definitely learned a lot about how to work with the Swift Storyboard and integrate our UI views and tools with the code for them. We also learned about machine learning models and libraries - even if we didn't use all of the tools we researched, this really contributed to our learning experience. We also learned how to research the right way to find answers to our questions on both concepts and programming, especially by understanding tutorials and articles and adapting them for our own unique applications.

What's next for FolDetect

We hope to extend this app to detect many more diseases, especially diseases with high transmission rates that can be contained by early detection and treatment. We also hope to develop a telehealth platform, in which we will give the user an option to send their results to their primary doctor and request a video consultation.

Built With

• createml
• swift