Omni-Seg: a scale-aware dynamic network for pathological image segmentation

This is the official implementation of Omni-Seg: A Scale-aware Dynamic Network for Pathological Image Segmentation.

Overview
Docker

Get our docker imageGet our docker image

https://github.com/ddrrnn123/Omni-Seg

Get our docker image

<code>sudo docker pull lengh2/omni_seg</code>Code language: HTML, XML (xml)

Run Omni-Seg

You can run the following command or change the input_dir, then you will have the final segmentation results in output_dir.

# you need to specify the input directory. 
export input_dir=/home/input_dir   
# make that directory
sudo mkdir $input_dir
# set output directory
export output_dir=$input_dir/output
#run the docker
sudo nvidia-docker run --shm-size 64G -it --rm -v $input_dir:/INPUTS/ -v $output_dir:/OUTPUTS lengh2/omni_seg Code language: PHP (php)

Description

Comprehensive semantic segmentation on renal pathological images is challenging due to the heterogeneous scales of the objects. For example, on a whole slide image (WSI), the cross-sectional areas of glomeruli can be 64 times larger than that of the peritubular capillaries, making it impractical to segment both objects on the same patch, at the same scale. To handle this scaling issue, we propose the Omni-Seg network, a scale-aware dynamic neural network that achieves multi-object (six tissue types) and multi-scale (5$\times$ to 40$\times$ scale) pathological image segmentation via a single neural network.

The contribution of this paper is three-fold:
(1) a novel scale-aware controller is proposed to generalize the dynamic neural network from single-scale to multi-scale;
(2) semi-supervised consistency regularization of pseudo-labels is introduced to model the inter-scale correlation of unannotated tissue types into a single end-to-end learning paradigm;
(3) superior scale-aware generalization is evidenced by directly applying a model trained on human kidney images to mouse kidney images, without retraining.

Reference

Omni-Seg: A Scale-aware Dynamic Network for Pathological Image Segmentation
Ruining Deng, Quan Liu, Can Cui, Tianyuan Yao, Jun Long, Zuhayr Asad, R. Michael Womick, Zheyu Zhu, Agnes B. Fogo, Shilin Zhao, Haichun Yang, Yuankai Huo.
IEEE Transactions on Biomedical Engineering

Omni-Seg: A Single Dynamic Network for Multi-label Renal Pathology Image Segmentation using Partially Labeled Data
Ruining Deng, Quan Liu, Can Cui, Zuhayr Asad, Haichun Yang, Yuankai Huo.
MIDL 2022

An Accelerated Pipeline for Multi-label Renal Pathology Image Segmentation at the Whole Slide Image Level
Haoju Leng*, Ruining Deng*, Zuhayr Asad, R. Michael Womick, Haichun Yang, Lipeng Wan, and Yuankai Huo.
SPIE 2023