Automated classification for COVID-19 pneumonia

• 

  COVID-19 pneumonia lungs

• 

  interstitial pneumonia lungs

• 

  normal lungs

• 

  lungs tissues distribution based on HU-values

At the beginning

We started our project to automate radiologists routines. Tons of CT-images processed by our team to start develop something. And now we have a lot of algorithms to deal with them. That's why we decided to join CGH with our project.

What it does

We offer the principle of lungs CT-images quantification with segmentation preprocessing

1. Lungs CT-image preprocessed for segmentation
2. Lungs masked and segmented
3. Mask applied and lungs tissue HU-values calculated
4. Magic here!
5. COVID-19/Interstitial/Norma classifier

How we built it

We build principles with Colab/Local IPython and implement in our servers

Challenges we ran into

The first challenge is to get properly descripted CT-scans. Most ordinary scans have no description at all. And here is another challenge for our medic teammates. They have to describe all of these scans.

Accomplishments that we're proud of

Collected a huge dataset of lungs-CT

What we learned

Data! The data is the key! The key to successful analysis and results

What's next for Automated classification for COVID-19 pneumonia

Create automated classification tool for radiologists

Built With

• python