List of Papers Browse by Subject Areas Author List
Abstract
Accurate cell segmentation in pathology images typically requires dense pixel-wise annotations, which are costly and time-consuming to obtain. This challenge is especially important for emerging biological imaging modalities and multiplexed datasets with variable channel configurations, where expert-labeled data are scarce. In this work, we introduce ImPartial, a deep learning framework designed to achieve state-of-the-art segmentation performance in low-annotation regimes using sparse scribbles and limited supervision. ImPartial augments the segmentation objective via self-supervised multi-channel quantized imputation. This approach leverages the observation that perfect pixel-wise reconstruction or denoising of the image is not needed for accurate segmentation, and thus, introduces a self-supervised classification objective that better aligns with the overall segmentation goal. We demonstrate that ImPartial achieves performance at par with fully supervised models while requiring substantially fewer annotations. Extensive experiments on benchmark multiplexed cellular imaging and single-plex clinical brightfield immunohistochemistry datasets show consistent improvements over strong baselines with only partial annotations. All benchmark datasets and code are available via our Github: https://github.com/nadeemlab/ImPartial.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026/paper/3854_paper.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to the Code Repository
https://github.com/nadeemlab/ImPartial
Link to the Dataset(s)
https://zenodo.org/records/4751737#.YKRTS0NKhH4
https://doi.org/10.7303/syn27624812
BibTex
@InProceedings{ShrGun_ImPartial_MICCAI2026,
author = { Shrivastava, Gunjan AND Nadeem, Saad},
title = { { ImPartial: Multi-channel Whole-Cell Segmentation Using Partial Annotations } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 16891},
month = {September},
page = {pending}
}
Reviews
Review #1
- Please describe the contribution of the paper
The authors present a novel method for semi supervised training of a segmentation network. The core idea is to combine segmentation with blind spot denoising, by having the blind spot network predict a probability distribution over class labels and then having one predicted Gaussian intensity for each class, which is optimised to predict the masked pixel value. The predicted classes are then used for segmentation. A second traditional cross entropy loss is used to train with sparse segmentation labels.
- 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.
I think this is a highly original and elegant and well motivated approach.
The paper is clearly written, to the point and precise.
I appreciate the large number of datasets and baselines
- 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.
I did not completely understand how exactly the scribbles where obtained. Are they part of the original datasets? Have they been added by the authors? What exactly does 10% scribbles mean? Does it mean 10% of pixels are annotated or does it mean that of a set of scribbles we are keeping 10%? How exaclt are the scribbles sub sampled?
- Please rate the clarity and organization of this paper
Good
- 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 authors claimed to release the source code and/or dataset upon acceptance of the submission.
- Based on your review and your understanding of the MICCAI Scientific Code of Ethics, do you believe this submission may involve a potential ethics concern or violation?
N/A
- 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/2026/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?
I think this is an interesting and elegant method achieving good results compared to the baselines. I am confident the question I have can be cleared up in the rebuttal.
- 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 #2
- Please describe the contribution of the paper
The paper proposes a framework to achieve SOTA performance on a variety of image modalities for digital pathology data using only weakly annotated data (scribbles).
- 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.
S1) The paper addresses an important issue of the domain: cumbersome annotations of small objects at large scale, with possibly involving many channels (multiplexed image data)
S2) A variety of evaluation metrics are stated, along with qualitative evaluation (examples, resulting masks, uncertainty maps).
- 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.
W1) It is mentioned in section 3, part “Uncertainty Quantification.” that there is an ensemble approach to guide user interaction and facilitate iterative refinement.
Several things are unclear:
Does the approach in this paper propose an active learning scheme with user-interaction? If so, this should be addressed more prominently.
The last sentence, “ImPartial was run for a single iteration with ensemble of 3.”, suggests that such a regime is implemented but was not used for the experiments in this work? (Fig. 1 also makes no mention of possible human-in-the-loop iterations)
Furthermore, some information on the time consumption for a multi-iteration setup would be favorable.
W2) The “amount” of scribbles was stated per experiment in the respective tables. While this is positive, the interested reader is left wondering how the scribbles were generated exactly. While for the CPDMFCI dataset, it is explicitly stated that ImageJ was used for annotation, there is no information regarding the other two datasets.
The following questions arise: How did the human annotator reach 10% scribbles exactly? (10% of what? full images? patches?) If for the other two datasets the ground truth was employed and artificially reduced, how? If for the other two datasets new scribbles were created, how?
- 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 authors claimed to release the source code and/or dataset upon acceptance of the submission.
- Based on your review and your understanding of the MICCAI Scientific Code of Ethics, do you believe this submission may involve a potential ethics concern or violation?
N/A
- 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/2026/en/REVIEWER-GUIDELINES.html
- page 2, row 5; “self-supervised” too wide for LNCS format
-
page 3.row 11 from bottom; “Noise2Void” too wide for LNCS format
- page 3, row 7 from bottom; “(I b = {i : x̄ bi ← x bj(i) , j(i) ̸ = i}) i.e. the blind spot patch, x̄ b , assigns random local values to pixel locations in I b )”: there appears to be an unopened bracket “)”
- 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?
While the paper leaves some questions (see Weaknesses) unanswered, the overall approach seems sound and valuable to the digital pathology community.
- 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
The paper proposes a weakly supervised cell segmentation framework, ImPartial, which requires only a small amount of scribble annotation rather than dense pixel-wise masks to perform whole-cell and nucleus segmentation, thereby substantially reducing annotation cost. In addition, the paper introduces a self-supervised multi-channel quantized imputation objective, using Gaussian mixture components to constrain foreground/background representations.
- 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 paper addresses an important problem in biomedical image analysis, namely cell segmentation under limited annotation. 2.On three benchmarks, the proposed method achieves performance comparable to, and in some settings better than, several fully supervised methods while using substantially fewer annotations. 3.The proposed multi-channel quantized imputation objective is interesting and reasonably novel.
- 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 overall presentation and writing are not sufficiently clear. The logical flow is sometimes difficult to follow, and some tables are missing clear headers, which hurts readability. 2.The experimental comparison is not fully convincing because the baselines are relatively weak for the claimed setting. In particular, the “Scribble Loss Only” variant in Table 2 already outperforms some methods in Table 1, which raises questions about the strength and relevance of the chosen baselines. More importantly, the paper does not compare against more directly relevant partial-supervision or weak-supervision methods. 3.The ablation study on the proposed multi-channel quantized imputation component is insufficient. The paper would be stronger if it provided a more detailed analysis of how this component contributes to the final performance.
- Please rate the clarity and organization of this paper
Poor
- 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 authors claimed to release the source code and/or dataset upon acceptance of the submission.
- Based on your review and your understanding of the MICCAI Scientific Code of Ethics, do you believe this submission may involve a potential ethics concern or violation?
N/A
- 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/2026/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.
(3) Weak Reject — marginally below the acceptance threshold, but would not mind if accepted, dependent on rebuttal
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
Overall, I think the paper presents a meaningful weakly supervised framework for cell segmentation that only requires sparse scribble annotations instead of dense pixel-wise masks. The proposed self-supervised multi-channel quantized imputation objective is interesting. However, the empirical evaluation is not sufficiently persuasive because the paper lacks comparisons with stronger and more directly relevant baselines. As a result, although the idea is promising, I find the current evidence insufficient to support the claims fully, and this leads me to a weak rejection recommendation.
- 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
- Please justify your decision. In case you deviate from the reviewers’ recommendations, explain in detail the reasons why. In case of an invitation for rebuttal, clarify which points are important to address in the rebuttal.
The reviewers agree that the paper addresses an important and practically relevant problem in biomedical image analysis: reducing the annotation burden for large-scale segmentation of small objects, particularly in multiplexed or multi-channel image data. They found the proposed approach to be original, elegant, well motivated, and reasonably novel, especially the multi-channel quantized imputation objective. The paper was also viewed as clearly written, concise, and precise. Empirically, the reviewers appreciated the extensive evaluation across multiple datasets, baselines, and metrics, and noted that the method achieves performance comparable to, and in some cases better than, fully supervised methods while requiring substantially fewer annotations.
The reviewers also raised several concerns. These include insufficient clarity regarding the scribble annotation protocol, questions about the strength and relevance of the experimental comparisons, particularly the lack of direct comparison to partial- or weakly supervised methods, and limited ablation analysis of the proposed imputation component. There is also some ambiguity regarding the role of uncertainty-guided iterative refinement.
Overall, despite these weaknesses, the paper makes a strong contribution to an important problem, and the main concerns appear addressable through clarification and additional discussion in the revision. Although one reviewer recommends rejection, the other two reviewers are highly positive. I also find that the strengths outweigh the weaknesses. Taking these considerations together, I recommend acceptance.