The Covid-19 pandemic has wreaked havoc on the world. It has caused widespread job loss, taking away the livelihoods of 20.6 million people. It has infected 131 million people, only 74.5 million of whom have recovered and will probably continue to feel the effects of the Covid-19 virus on their health for years to come and killed nearly 3 million people. It has shut down schools and halted students' learning and caused increased feelings of depression and anxiety, tripling the rate of depression amongst US adults of all age groups. Personally, I've seen the impact Covid-19 has had on my family members as well as friends and I know how terrible it can be, especially for my older relatives with weaker immune systems. I wanted to create a hack that tackled this issue by providing a publicly available machine learning model for predicting whether an individual has Covid-19 so that radiologists, who often find themselves overwhelmed with Covid-19 patients, can have a machine learning model quickly predict, based on chest X-rays, whether a patient has Covid-19 to double check their work and make sure that their assessment was accurate.
What it does 💻🖱
Covid-Care has a machine learning model that can predict whether or not a patient has Covid-19 based on uploaded chest X-rays. It also has a tab that displays nearby Covid-19 treatment centers to users as well as a tab that educates about vaccines due to the misinformation regarding the vaccine and it also debunks popular Covid-19 myth.
How we built it 👩💻
I built the webpage using HTML and CSS. I used the Google Maps API as well as the Places API to embed a Google Map displaying nearby Covid-19 treatment centers on the Hospitals page. For the machine learning model, I used Google Cloud Auto ML to train and test a machine learning model on a Kaggle dataset of 1,392 images with a 97% accuracy. It predicts 95% of the positives correctly and 99% of the negatives correctly.
Challenges we ran into 🤦😡
I ran into many challenges with my machine learning model. First, I had to find a dataset that had enough data to build an accurate machine learning model as the more data a model has, the better it will perform. The images took several hours to train and test so this cut back on my time to add other features to the hack. Last night, when I was trying to record my demo, our power went out for several hours and luckily it came back just in time for me to submit my project.
Accomplishments that we're proud of 🥳
I'm proud that I completed my hack on time and that I was able to create a hack that has a real world impact and can help people!
What we learned 👩🏻🏫
I learned that it's a good idea to plan ahead and that errors will occur, the important part is that I don't get discouraged by them and keep working!
What's next for Covid Care
I want to continue to improve my model as well as create a tab that has a test of symptoms and lets the user know if it is likely that they have Covid-19 and should get checked based on the number of systems that they've realized apply to them.