Abstract:A histopathological image is a microscopic image applied to examine cellular and tissue structures and identify any abnormalities or disease processes. Histopathological image segmentation is a prerequisite step for analyzing histopathological images that can divide an image into meaningful regions or objects to accurately classify and analyze tissue structures, cellular regions, or particular histological entities. However, the existing deep learning based pathological image segmentation methods require huge annotation efforts from the pathologists, which is labor-intensive and time-consuming. In this scenario, it has become a hotspot to leverage abundantly available unlabeled data to help learn segmentation models given limited labeled data. In this paper, we propose a global-local consistent semi-supervised segmentation (GLCS) model that enforce the consistency of the segmentation results with weak and strong perturbations on unlabeled data. In GLCS, we firstly generate different weak perturbations for each unlabeled sample, and then add a regularization term to ensure the segmentation consistency among different weak perturbations. Next, different from the existing studies applying the regression methods to match the segmentation results among different perturbations, our methods are based on the generative adversarial learning that can keep the global structure consistency among unlabeled data with different strength of perturbations. Finally, we also add a patch-correlation based regularization term to preserve the local structure similarity among different perturbations images. We validate our GLCS on three datasets, i.e. Glas, Crag and MoNuSeg. The experimental results show that our method can achieve to the dice ratio of 90.35, 82.61 and 81.60 with 1:1 proportion of labeled data, which are significantly superior to the state-of-the-art semi-supervised histopathological image segmentation methods. Our code is public available at https://github.com/ISBELLAG/GLCS .

Intra- and Inter-Pair Consistency for Semi-Supervised Gland Segmentation

Pairwise Relation Learning for Semi-supervised Gland Segmentation

Gland Instance Segmentation by Deep Multichannel Side Supervision

Morphology-inspired Unsupervised Gland Segmentation via Selective Semantic Grouping

Weakly-Supervised Gland Instance Segmentation based on Point Labeling

Gland Segmentation Via Dual Encoders and Boundary-Enhanced Attention

GlandSAM: Injecting Morphology Knowledge into Segment Anything Model for Label-free Gland Segmentation

Gland Instance Segmentation by Deep Multichannel Neural Networks

Dual-consistency guidance semi-supervised medical image segmentation with low-level detail feature augmentation

Dual Consistency Semi-Supervised Nuclei Detection Via Global Regularization and Local Adversarial Learning.

Multi-domain consistency constraint model for semi-supervised lesion segmentation in automatic breast ultrasound (ABUS)

Gland Instance Segmentation Using Deep Multichannel Neural Networks

Inherent Consistent Learning for Accurate Semi-supervised Medical Image Segmentation

Uncertainty-Guided Mutual Consistency Learning for Semi-Supervised Medical Image Segmentation

Global-local consistent semi-supervised segmentation of histopathological image with different perturbations

Deep Segmentation-Emendation Model for Gland Instance Segmentation

Semantic Segmentation of Colon Gland with Conditional Generative Adversarial Network.

IntrNet: Weakly Supervised Segmentation of Thyroid Nodules Based on Intra-image and Inter-image Semantic Information.

Online Easy Example Mining for Weakly-supervised Gland Segmentation from Histology Images

Multi-Consistency Training for Semi-Supervised Medical Image Segmentation

A Hybrid Feature Enhancement Method for Gland Segmentation in Histopathology Images