Overview / Usage

By using wearable devices we monitor basic vital signs like heart rate, HRV, blood saturation or movement to analyze your heart condition. IRBA performs continuous analysis of provided data using machine learning algorithms to detect vital signs patterns and recognize life-threatening heart behaviour. When IRBA detects a life-threatening state, it will notify the user right away by displaying medical assessment. If he does not respond in proper time or his assessment is positive in the direction of sudden cardiac states it calls for adequate help - emergency contacts like friends or relatives, medical support services and immediately provide them with your personal details (including allergies, taken drugs, your diseases, insurance number) that can be helpful for paramedics. Every worrying, but not life-threatening the performance of your circulatory system will be also detected and saved to be presented for you and your physician during next follow-up visit which can help diagnose many chronic diseases and fit the most effective therapy. IRBA improves first aid at every level. It calls for help immediately after it detects SCA or any other life-threatening state. It calls and provides your personal details to EMS faster and more precisely than relatives and bystanders. It also calls for your own emergency contacts like your son or neighbour who have a chance to reach an emergency scene before ambulance arrival and assists them in CPR. Gathered data from emergency moment can be valuable for physicians in diagnosing process and important to enter the right treatment. That all shorten the time of response and provide immediate
smartphone and wearable device. All gathered data are valuable for research purposes, especially in detecting early symptoms of cardiovascular diseases, improving diagnose algorithms, and optimizing treatment methods.

Methodology / Approach

Data directly from life-threatening states and many others are rare and most often in low quality but crucial to make the next step in medicine and save millions of lives. In IRBA we have created an app compatible with most of the Bluetooth 4.0 wearables, because of their popularity and low price, and we have started works on compatibility with smart bands working on their own SDK. We wrote a basic algorithm based on statistic data analysis and deep neural network searching for dangerous patterns on heart rate, heart rhythm and movement level to call help and collect significant data for further development. This simple app at this point can detect major life-threatening states and call for help with 30-60s delay, but its greatest strength lies in its accessibility all around the world and the amount of data that we can gather in a short time to improve our AI algorithms and make the global game-changing product in emergency medicine and cardiology.

Heart rate is analyzed using deep neural network trained on expert-labeled data, achieving over 93 per cent sensitivity on separate validation subset.

Technologies Used

All models for IRBA project have been implemented using open-source scikit-learn and Keras machine learning frameworks.

Technologies to integrate IRBA app with iOS ecosystem are following: Swift 4.0, CoreBluetooth for BT 4.0 communication and CoreML for on-device machine learning integration.

Other links

• Our website