/Refed

Refed (Rib Extaction and Fracture Detection Model) aims to extract ribs and detect fracture from CT images.

REFED_logo

Refed

Refed (Rib Extaction and Fracture Detection Model) by Youyou Jiang and Shiye Lei in AI lab, Matrix Inc. Our model has a great performance on extracting ribs and detecting fracture from CT images. The model is based on python 3.6.

Refed is consist by two modules: rib extraction module and fracture detection module. First, we design algorithm based on computer vision for extracting ribs from CT images. Second, we design DNN based on faster-rcnn for detecting fracture location with these ribs.

Performance

rib extraction module: extract ribs from source CT image

  1. Input CT image

source CT image

Note: the source CT image is handled by HU value binary process (HU threshold = 400)

  1. Cut sternum with envelope line

sternum envelope line

  1. remaining spine

remaining spine

  1. collect ribs

collect ribs

you can also view every single rib as follows:

single rib

  • operation system: Linux (our OS is Ubuntu 16.04)
  • interpreter: More advanced than python 3.6.0
  • some python packages: skimage, opencv, tensorflow, sklearn

Note

REFED is still in progress. So we cannot open source code of REFED due to business reasons. If you are intereted in the program and want to know more details, please send an e-mail to me. Thanks for your understanding.