Skip to content
/ CombTR Public

A paper utilizing an ensemble model for multi-organ segmentation.

Notifications You must be signed in to change notification settings

Ian-VA/CombTR

Repository files navigation

CombTR

image

CombTR is an ensemble model developed to automatically segment 13 organs based off of 3D CT scans. The following models are used:

  • UNETR
  • SwinUNETR
  • SegResNet

CombTR uses a "stacking" model architecture, where each model's output is the input into a meta-learner, which learns from the ensemble's mistakes and corrects them over time. This model beat previous research from NVIDIA and Vanderbilt University by 5% in DICE score.

Winner of Solano County Science Fair 2023

3rd place UCLA Brain Research Institute Award @ California State Science and Engineering Fair 2023

About

A paper utilizing an ensemble model for multi-organ segmentation.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages