-
-
-
-
-
-
-
-
-
Our Team!
Introduction
Cardiovascular diseases are the number 1 cause of death globally, taking an estimated 17.9 million lives each year. On top of this worrisome statistic, COVID-19 has had an unexpected impact - in addition to lung damage, patients are also developing heart problems and dying of cardiac arrest, and more doctors and experts have come to believe the virus can directly infect the heart muscle (citations at the bottom).
The need to diagnose potential heart issues is now higher than ever, but with social distancing rules and busy doctors offices, many patients are skipping their visits and not getting the preventive care they need.
Teladocs are helpful in the age of social distancing, but checking heart heath is a challenge from far away. We needed a solution. This is why we created heart.ai.
Heart.ai uses artificial intelligence and machine learning to address this need, and helps doctors detect potential heart defects in patients they can't see in person.
Project Features & Technologies
- Front end for communicating with your doctor, checking your own heart's health, listing your health history, reporting your own heartbeat, or watching your own heartbeat with a frequency response graph. All built with React.
- Homemade stethoscope for recording your heartbeat in an affordable and cheap way for everyone.
- The math for calculating the best homemade stethoscope design in a PDF formatted with Latex.
- Instructions for making your own stethoscope!
- TensorFlow machine learning model for classifying the stethoscope's audio signal as normal or abnormal. Trained with a Kaggle dataset and trained on Google Cloud's Deep Learning VMs.
- OpenCV external application used to detected heart rate by monitoring blood flow through capturing a video of the user's face.
Math Breakdown
All the mathematics that went behind approximating the performance of our hardware can be found detailed in this paper.
In order to achieve an accurate model, we carefully measured the dimensions of our hardware and accurately predicted how it would behave as a stethoscope, to ensure it would function as we intended and deliver the correct results when listening for a heartbeat, free of white noise or other kinds of audio disruptions.
Here is a preview:
Furthermore, find instructions on how we made this here.
Challenges we ran into
Kestrel: During our prototyping phase, the 3D printer's power supply failed and left us with incomplete parts and no 3D printer. However, this turned out to be a blessing in disguise as it encouraged us to create a stethoscope that anyone can make from home, without a 3D printer!
Belle: Figuring out how to grab audio from the user from the front end was an ordeal that took several hours. We got it working, even tho the visualization wasn't exactly how we'd pictured it!
Citations
https://abcnews.go.com/Health/health-struggle-covid-19-patients-heart-failure/story?id=70002186
https://www.dicardiology.com/article/cardiovascular-impact-covid-19
Built With
- google-cloud
- jupyter
- kaggle
- latex
- matlab
- python
- react
- solidworks
- tensorflow
Log in or sign up for Devpost to join the conversation.