Inspiration
What inspired us was the state of the healthcare system in Greece. The pandemic meant that the scarce healthcare resources left after more than a decade of austerity measures would have to manage a large number of cases requiring hospitalisation. At the same time the vast majority of COVID-19 cases would remain at home with minimal symptoms.
These people would be scared and rush to the hospital, taking precious time out of nurses and doctors. We wanted to avoid that. How? By making it easier for people to stay in a virtual health line, control their own situation, and hopefully even offer a helping hand to others. Most importantly, we wanted the cases remaining at home to be adequately managed with minimal requirements in healthcare professionals. At the same time, we wanted an early detection system for all such deteriorating cases where further interventions may be required. This would give both patients and the overstretched healthcare system the best chance for success.
In the end though, it became obvious that even the most robust healthcare systems can be crippled by such a pandemic. This showed us that our platform was relevant to a much wider audience, which prompted us to take this step.
What it does
Our system tries to make it easy for the user (patient, someone under quarantine, or just any concerned person) to self-report their health status. This is done using a number of subjective (symptoms/underlying diseases questionnaires etc.) and objective (temperature, respiratory audio files) criteria. The built-in AI evaluates these criteria and alerts healthcare personnel if further actions are needed.
How we built it
We chose Django for the fast development experience, and continuous iteration. It allows us to stay robust and roll out new features daily. For the decision making, we chose Python which has various options to implement a fuzzy system build on top of SciPy/NumPY.
Challenges we ran into
Our main challenge is to fine-tune the AI to make the best decisions in each case. As it is a new threat, there is a lot new and often conflicting information. We therefore need the assistance of patient management experts for COVID-19 cases, to help us at training and validating the AI (symptoms and audio files).
Accomplishments that we are proud of
We are proud that we have developed this product in a matter of days, in the spirit of the Greek Hackathon, and going strong since. The video was developed within 2 days, during the current hackathon. The first iteration is already giving decent results, but as mentioned, we're rolling out new ideas and features constantly.
We are also very proud of its potential to save lives and healthcare resources, and that at no significant cost for the user or the healthcare system, as no gadgets, peripherals etc. are required.
What we learned
When we started working on the initial idea, we tried to solve a local problem (limited healthcare resources) for another mostly local issue (a virus mostly affecting China, with some localised incidents in other countries, just beginning to affect Italy and Spain). We were amazed at how fast this platform gained global significance.
What's next for StayingHealthy
Finalisation of the platform. Fine-tuning and Validation of the AI (symptoms-audio). Expansion of the AI capability to be able to train itself based on feedback from doctors, healthcare professionals, institutions etc.,so that the platform could quickly adapt to any symptoms change (from a different strain or even a completely different pandemic).
Log in or sign up for Devpost to join the conversation.