Skip to content

kits 2019 : kidney segmentation challenge for MICCAI 2019

Notifications You must be signed in to change notification settings

SWKoreaBME/sangwook_kits2019

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

5 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Kits Challenge

Kidney Segmentation in CT scans

Dataset

  1. KiTs Challenge ( MICCAI 2019 )

Characteristics of images

  1. CT images ==> HU units

  2. Voxel Spacing varies among patients BUT slice WIDTH, HEIGHT are same ( 512, 512 except for one case ( case_00160 | (512, 796)))

Models

  1. 3D Unet

    • 3d 를 사용할거면 voxel spacing 을 맞춰주는 것이 좋을 것 같다? ( z 축으로 사이즈가 일정하지 않으니까 / 대신에 ㅑisotropic-Resampling 을 거쳐야 하는데 오래걸릴수도 있다.. )
    1. V-net ( Code available )
  2. 2D Segmentation

    1. DeepLABv3+ ( code available )
    2. Unet++ ( code available )
    3. Attention Unet
    4. Modified-Unit [ https://arxiv.org/pdf/1802.10508.pdf ] -> 1등 ?

Loss Functions

https://lars76.github.io/neural-networks/object-detection/losses-for-segmentation/

Loss Function 에 대해서 생각해 볼 필요가 있다.

Due to highly unbalanced dataset ( little Tumor, Kidney region + Too much Background Values )

Rules

Score

  1. Teams will be awarded a score for each of the 90 test cases equal to the (Kidney Sørensen–Dice + Tumor Sørensen–Dice)/2.

    Label 1 의 Kidney 의 Sørensen–Dice score 와 Label 2 의 Tumor 의 Sørensen–Dice score 의 산술 평균 값을 보고하는 점수로 사용한다.

Sorensen-Dice Score [ wiki ]

Results

Reference

​ [1] https://kits19.grand-challenge.org/rules/

About

kits 2019 : kidney segmentation challenge for MICCAI 2019

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages