StrokeSentry

Inspiration

The project was inspired by a personal story. One of our team members had a very close family member who suffered from a stroke. Fortunately, they were lucky enough to survive. But in those moments he wondered how this could have just happened. That experience made him realize how important it is to recognize stroke symptoms quickly and accurately to prevent them. He thought, "What if someone had a tool right in their pocket or on their phone to overcome the long and arduous process of detecting a stroke"? That question is what led to StrokeSentry, an app that helps detect stroke symptoms in real time.

What it does

Stroke Sentry is a mobile app that uses your phone camera and microphone to check for stroke symptoms. It is built based off the vastly proven F.A.S.T method: -Face Drooping

Arm weakness
Speech Difficulty
Time to act

The app looks for things such as facial asymmetry and drooping, slurred speech , or changes in posture. It provides the user with quick, clear and concise feedback so they can take immediate action if needed.

How we built it

We built the app using Swift for iOS. For real time face and pose estimation we used Apple's Vision Framework. Our backend the models consist of:

Convolutional Neural Networks to detect facial and pose/ posture changes
Speech- to- text API's to analyze voice input and catch slurring
CoreML and TensorFlow Lite to run everything on the device

Challenges we ran into

The hardest part was making sure the app could recognize stroke symptoms accurately. Faces come in all kinds of shapes and lighting cannot always be perfect, and speech might vary a lot from person to person.

Accomplishments that we're proud of

We are very proud that we built a working app that detects stroke symptoms in real time, reached over a 75% accuracy in tests, designed a clean and simple UI for anyone to use and created something that could potentially save many lives.

What we learned

We learned just how powerful AI can be when applied to real world problems. Strokes are just one of many health risks, there are many medical health risks where early detection can make a huge difference. We also realized that not every problem has been solved yet. There is still so much room for AI and other software tools to grow, especially in healthcare. This project made us more motivated to keep building tools that help people when it matters most.

What's next for Stroke Sentry

We plan on refining the app and the releasing it on the App Store, work on publishing our results in a reputable and credible journal and explore new features to detect other health risks in real time.