AI Pneumothorax Segmentation

Medical image classification with Deep Neural Networks

This application uses a Deep Neural network to identify Pneumothorax. It takes an Thorax x-ray as input, the resulting image will have the Pneumothorax zones hightlighted.

Using an example X ray image ———–> and the result will be:

——>

This application can be found in: Heroku. You can test it with your own images!

The Deep Learning neural network module has been trained with images from the SIIM-ACR Pneumothorax Segmentations from kaggle. The software used is PyTorch with fast.ai.

This repo is based in fast.ai and has been adapted to be used on Heroku.

You can test locally installing Docker and using the command:

docker build -t pneumo . && docker run --rm -it -p 5000:5000 pneumo