Skip to main content

Kaggle Gold Kernel

Project Details

General

Writer

Screengrab

Description

Your patient comes in. They're having a stinging pain in the chest. The X-Ray shows a shadow on the lung. And that shadow is air. If not treated a "collapsed lung"(-ish), or "air on the wrong side of the lung", can result in death.

What's a Pneumothorax?

Problem is in an X-Ray, air is usually the thing you ignore. The general idea is:

  • Black: Air
  • Gray: Fluids and Tissue
  • White: Bone and Solids

So the issue is that an air enclosure may just be a mild disturbance in the chest xray. Considering convolutional neural networks are exceptional at identifying abnormalities, we may want them to take a look, as not to miss these tiny abnormalities.

In this challenge, we get chest xrays and masks. In challenges I link below, only bounding boxes were available. Here, we actually get to do dense prediction.

Visit Site Fork on Github