One day I was reading the news that CoronaVirus is spreading very quickly and highly developed countries like USA and Italy are not capable of handling this virus. Most striking news was earlier it takes up to 2 days for just testing whether the patient is COVID positive or not. Then after some advancements also, it is taking some hours to test. So, I thought there should be the minimum time to test this virus so that actions can be taken as soon as possible because of its highly contagious nature.

What it does:

It takes Chest X-ray of the patient and within seconds it will try to predict whether the patient may be positive for coronavirus or not.

How I built it:

I developed this system using Deep learning technology (which is a subset of AI). I used Convolutional Neural Networks to train my model from scratch on various layers to achieve the highest possible accuracy. I also used Flask to deploy this model as a web app. Currently, I'm using 's server for the deployment of this web app.

Challenges I ran into:

Accuracy was the greatest challenge because Xray of a pneumonic patient and a person having COVID can be visually similar because pneumonia can be a symptom of COVID-19. So, to classify COVID cases from pneumonic cases separate was a big challenge in itself.

Accomplishments that I'm proud of:

What makes my model different from other detectors out there is I didn't use transfer learning for training my model. I tested my model on unseen images of COVID and Non-COVID Xrays and achieved 91% accuracy on training set and 90% accuracy on the test set.

What I learned:

During developing this system, I learned how new technologies like AI can be helpful in any field like healthcare.

What's next for COVID-19 X-RAY DETECTOR:

For future work, I will add CT scan image processing also in this system so that Not just Xray, CT scan can also be used to predict COVID-19 Positive cases. Also, I am doing more research on how to make this system even more accurate so that govt. and hospitals can make use of this system at a higher scale and people can be benefited from this.

Built With

+ 11 more
Share this project: