We all know COVID19 is extremely contagious by now, and shortage in healthcare provider is already happening in first world, this would only further escalate in 3rd world nations who do not have much resources. By being part of UN Sustainable Development Goal 3, we want to innovate in this area and able to be able to help everyone in the world.
Telemedicine have been maturing for the past decade, and many have expected after COVID19, this will be pushed towards mainstream. This also allows doctors working remotely to treat people in developing nations, that's in dire need of healthcare resources right now. Every doctor wears a stethoscope, they can use that to diagnose symptoms for the patients.
The digital stethoscopes in the market is generally not affordable at individual level, especially in Developing nations and struggling communities, they are a great tool for the doctor, but in telemedicine, we need the tool to be able to diagnose patients remotely. In case of no doctors available for Developing Nations, we need AI that's affordable to do that diagnosis.
We are building AI Responsibly by giving AI access to developing nations as well as developed nations.
What it does
So we want to build an open sourced digital stethoscopes that's under $1.00 and can easily be given to everyone in the world, so that we can empower doctors via telemedicine to treat more people.
In this project, we will also train an AI to use sound classification so that patients can do a self checkup anytime, and when the AI detects possible symptoms, the audio data is recorded for the patients to send it to a doctor for a further checkup.
How I built it
We built our digital stethoscope by cutting apart stethoscope and a cheap microphone, both can be bought for under $10. This allows us to record sound from the lungs.
We then got data from kaggle, and trained sound recognition system using TF2.2. Building an Android app that does sound classification. The entire built instruction can be seen at https://www.hackster.io/339769/digital-stethoscope-ai-1e0229
Challenges I ran into
TF 2.2 had a lot of backward compatibility issues such as tensorboard, which if we were to use we had to downgrade.
Accomplishments that I'm proud of
We are able to build a digital stethoscope under $1, this is affordable now to everyone in the world, especially developing nations.
What I learned
Most expensive tools can be broken down and made cheaper, and stethoscope in people's hand can really help wit telemedicine.
What's next for Digital Stethoscope AI
We are going to work with UNDP to try to get this into everyone's hand. And with this we believe we can make an impact to the telemedicine world.