List of Papers Browse by Subject Areas Author List
Abstract
Existing studies on weakly supervised pathological tissue segmentation predominantly rely on class activation maps (CAMs) to generate pixel-level pseudo-masks from image-level labels. However, CAMs tend to emphasize only the most discriminative regions, resulting in boundary noise that undermines the quality of pseudo-masks and degrades segmentation performance. To address these challenges, we propose a novel weakly supervised pathological tissue segmentation framework: Edge-semantic Synergy Fusion and Adaptive Noise-aware (ESFAN) mechanism. In the classification phase, the Edge-semantic Synergy Fusion (ESF) improves the quality of pseudo-masks by incorporating four synergistic components. The hybrid edge-aware transformer refines boundaries, while the pyramid context integrator captures multi-scale context. The context channel amplifier fine-tunes semantic features, and the adaptive fusion gating balances feature map contributions using learnable spatial weights. In the segmentation phase, we propose an Adaptive Noise-aware Mechanism (ANM) that incorporates adaptive weighted cross-entropy, uncertainty regularization, and spatial smoothing constraints to mitigate noise in pseudo-masks and enhance segmentation robustness. Extensive experiments on the LUAD-HistoSeg and BCSS datasets demonstrate that ESFAN significantly outperforms state-of-the-art methods. The code is available at: \url{https://github.com/Sameer-815/ESFAN}.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2025/paper/2692_paper.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to the Code Repository
https://github.com/Sameer-815/ESFAN
Link to the Dataset(s)
N/A
BibTex
@InProceedings{ZhaHua_Edgesemantic_MICCAI2025,
author = { Zhang, Hualong and Feng, Siyang and Huan, Zihan and Wang, Huadeng and Liu, Zhenbing and Lan, Rushi and Pan, Xipeng},
title = { { Edge-semantic Synergy Fusion and Adaptive Noise-aware for Weakly Supervised Pathological Tissue Segmentation } },
booktitle = {proceedings of Medical Image Computing and Computer Assisted Intervention -- MICCAI 2025},
year = {2025},
publisher = {Springer Nature Switzerland},
volume = {LNCS 15967},
month = {September},
page = {162 -- 171}
}
Reviews
Review #1
- Please describe the contribution of the paper
This paper introduces a new weakly-supervised pathology image segmentation framework, which incorporates a new edge-enhancing strategy and a new pseudo label denoising strategy to improve the model performance.
- Please list the major strengths of the paper: you should highlight a novel formulation, an original way to use data, demonstration of clinical feasibility, a novel application, a particularly strong evaluation, or anything else that is a strong aspect of this work. Please provide details, for instance, if a method is novel, explain what aspect is novel and why this is interesting.
1.The code is released. 2.The proposed method is evaluated on two relative large datasets. 3.The proposed method to enhance the pseudo labeling at the boundary and denoising noise in pseudo labels is reasonable. 4.Essential ablation studies are conducted.
- Please list the major weaknesses of the paper. Please provide details: for instance, if you state that a formulation, way of using data, demonstration of clinical feasibility, or application is not novel, then you must provide specific references to prior work.
1.The proposed method utilizes different version of ResNet (200) from most comparing methods (ResNet 50/101), which makes the comparison unfair. It is not clear if the performance improvement comes from larger backbone or the proposed pseudo label enhancing techniques. 2.In Fig. 2, it seems that the final performance of the proposed method is still weak when comparing to the ground-truth.
- Please rate the clarity and organization of this paper
Satisfactory
- Please comment on the reproducibility of the paper. Please be aware that providing code and data is a plus, but not a requirement for acceptance.
The submission has provided an anonymized link to the source code, dataset, or any other dependencies.
- Optional: If you have any additional comments to share with the authors, please provide them here. Please also refer to our Reviewer’s guide on what makes a good review and pay specific attention to the different assessment criteria for the different paper categories: https://conferences.miccai.org/2025/en/REVIEWER-GUIDELINES.html
N/A
- Rate the paper on a scale of 1-6, 6 being the strongest (6-4: accept; 3-1: reject). Please use the entire range of the distribution. Spreading the score helps create a distribution for decision-making.
(4) Weak Accept — could be accepted, dependent on rebuttal
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
novel method while using different backbone may cause unfairness in comparison
- Reviewer confidence
Very confident (4)
- [Post rebuttal] After reading the authors’ rebuttal, please state your final opinion of the paper.
N/A
- [Post rebuttal] Please justify your final decision from above.
N/A
Review #2
- Please describe the contribution of the paper
The paper presents a new method called ESFAN for weakly supervised pathological tissue segmentation, aiming to solve issues with unclear boundaries and noise in pseudo-masks from class activation maps (CAMs). It improves boundary precision with a technique called Edge-semantic Synergy Fusion (ESF), which combines multiple components like edge detection and context integration. It also uses an Adaptive Noise-aware Mechanism (ANM) to reduce noise in the pseudo-masks.
- Please list the major strengths of the paper: you should highlight a novel formulation, an original way to use data, demonstration of clinical feasibility, a novel application, a particularly strong evaluation, or anything else that is a strong aspect of this work. Please provide details, for instance, if a method is novel, explain what aspect is novel and why this is interesting.
The paper’s major strengths include its novel ESFAN framework for weakly supervised pathological tissue segmentation, which combines Edge-semantic Synergy Fusion (ESF) and Adaptive Noise-aware Mechanism (ANM) to address challenges like unclear boundaries and noise in pseudo-masks. This work proposed an adaptive approach without needing manual parameter tuning.
- Please list the major weaknesses of the paper. Please provide details: for instance, if you state that a formulation, way of using data, demonstration of clinical feasibility, or application is not novel, then you must provide specific references to prior work.
- While the paper showcases strong segmentation performance, it lacks analysis on the computational efficiency and scalability of the proposed method. The time and computational cost of training and inference with the ESFAN framework are not discussed.
- The ESFAN framework incorporates multiple components, such as the Pyramid Context Integrator and Context Channel Amplifier, which perform similar functions of feature fusion and context enhancement. They might not need to be completely separate.
- Please rate the clarity and organization of this paper
Satisfactory
- Please comment on the reproducibility of the paper. Please be aware that providing code and data is a plus, but not a requirement for acceptance.
The submission has provided an anonymized link to the source code, dataset, or any other dependencies.
- Optional: If you have any additional comments to share with the authors, please provide them here. Please also refer to our Reviewer’s guide on what makes a good review and pay specific attention to the different assessment criteria for the different paper categories: https://conferences.miccai.org/2025/en/REVIEWER-GUIDELINES.html
N/A
- Rate the paper on a scale of 1-6, 6 being the strongest (6-4: accept; 3-1: reject). Please use the entire range of the distribution. Spreading the score helps create a distribution for decision-making.
(4) Weak Accept — could be accepted, dependent on rebuttal
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
The paper is reasonably written, but the novelty is somewhat lacking. It’s difficult to determine whether the performance improvement is due to the effectiveness of the structural design or merely the result of an increased number of parameters.
- Reviewer confidence
Confident but not absolutely certain (3)
- [Post rebuttal] After reading the authors’ rebuttal, please state your final opinion of the paper.
N/A
- [Post rebuttal] Please justify your final decision from above.
N/A
Review #3
- Please describe the contribution of the paper
Existing studies on weakly supervised pathological tissue segmentation predominantly rely on class activation maps (CAMs) to generate pixel-level pseudo-masks from image-level labels. However, CAMs tend to emphasize only the most discriminative regions, resulting in boundary noise that undermines the quality of pseudo-masks and degrades segmentation performance. To address these challenges, this paper propose a novel weakly supervised pathological tissue segmentation framework: Edge-semantic Synergy Fusion and Adaptive Noise-aware (ESFAN) mechanism, which addresses critical challenges in weakly supervised pathological tissue segmentation, including boundary ambiguity and noise sensitivity in pseudo-masks.
- Please list the major strengths of the paper: you should highlight a novel formulation, an original way to use data, demonstration of clinical feasibility, a novel application, a particularly strong evaluation, or anything else that is a strong aspect of this work. Please provide details, for instance, if a method is novel, explain what aspect is novel and why this is interesting.
The paper proposes an edge semantic collaborative fusion method, which improves tissue segmentation accuracy and boundary accuracy by generating high-quality pseudo masks. In addition, the paper introduces an adaptive noise perception mechanism to mitigate the impact of noisy labels, effectively suppressing noise while preserving important information from other regions. Extensive experiments conducted on two publicly available datasets have demonstrated the outstanding performance of the method. This method achieves 79.29% mIoU on the LUAD-HistoSeg dataset and 71.41% on the BCSS dataset, setting state-of-the-art performance with only image-level labels.
- Please list the major weaknesses of the paper. Please provide details: for instance, if you state that a formulation, way of using data, demonstration of clinical feasibility, or application is not novel, then you must provide specific references to prior work.
- The network architecture is somewhat complex, and the range of the adaptive threshold τ in ANM, as well as the initial and final weight values in the weighted sum of the final loss, require supplementary theoretical or experimental justification.
- It is unclear whether the gradient smoothing loss in Equation (9) uses squared differences (the formula currently describes absolute value summation, but the text mentions “squared gradient differences”). Ensure consistency between mathematical notation and textual descriptions.
- Please rate the clarity and organization of this paper
Satisfactory
- Please comment on the reproducibility of the paper. Please be aware that providing code and data is a plus, but not a requirement for acceptance.
The submission has provided an anonymized link to the source code, dataset, or any other dependencies.
- Optional: If you have any additional comments to share with the authors, please provide them here. Please also refer to our Reviewer’s guide on what makes a good review and pay specific attention to the different assessment criteria for the different paper categories: https://conferences.miccai.org/2025/en/REVIEWER-GUIDELINES.html
N/A
- Rate the paper on a scale of 1-6, 6 being the strongest (6-4: accept; 3-1: reject). Please use the entire range of the distribution. Spreading the score helps create a distribution for decision-making.
(5) Accept — should be accepted, independent of rebuttal
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
The paper proposes an edge semantic collaborative fusion method, which improves tissue segmentation accuracy and boundary accuracy by generating high-quality pseudo masks. In addition, the paper introduces an adaptive noise perception mechanism to mitigate the impact of noisy labels, effectively suppressing noise while preserving important information from other regions. The paper also conducted extensive experiments to demonstrate the performance of the method.
- Reviewer confidence
Somewhat confident (2)
- [Post rebuttal] After reading the authors’ rebuttal, please state your final opinion of the paper.
N/A
- [Post rebuttal] Please justify your final decision from above.
N/A
Author Feedback
N/A
Meta-Review
Meta-review #1
- Your recommendation
Provisional Accept
- If your recommendation is “Provisional Reject”, then summarize the factors that went into this decision. In case you deviate from the reviewers’ recommendations, explain in detail the reasons why. You do not need to provide a justification for a recommendation of “Provisional Accept” or “Invite for Rebuttal”.
N/A