List of Papers Browse by Subject Areas Author List
Abstract
Tracking tumor lesions across serial CT scans is essential for oncological response assessment. Existing automated methods face a fundamental trade-off: end-to-end trackers achieve high automation but offer no opportunity to correct silent tracking failures, while decoupled registration–segmentation pipelines permit user verification yet discard the lesion’s prior appearance, limiting accuracy in ambiguous cases. In this work, we propose a Verified Tracking paradigm: a clinician verifies a registration-proposed prompt, which the model leverages alongside the baseline lesion appearance to resolve segmentation ambiguities. We present a unified framework combining early spatial prompt fusion with latent temporal difference weighting for longitudinally-informed segmentation. To address data scarcity, we leverage large-scale synthetic pretraining, proving essential for exploiting longitudinal context, improving performance by up to 4.5 Dice points over training from scratch. Our approach secured first place in the MICCAI autoPET IV challenge. We further curate and release PanTrack, a new longitudinal pancreatic cancer benchmark, to assess out-of-distribution generalization. Experiments show that our model outperforms prior work in both fully automatic and the proposed verified tracking setting offering a clinically safe middle ground between automation and control. Code, model and dataset will be released at https://github.com/MIC-DKFZ/LongiSeg
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026/paper/3814_paper.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to the Code Repository
https://github.com/MIC-DKFZ/LongiSeg
Link to the Dataset(s)
https://huggingface.co/datasets/mrokuss/PanTrack
BibTex
@InProceedings{KirYan_Exploiting_MICCAI2026,
author = { Kirchhoff, Yannick AND Rokuss, Maximilian AND Mertens, Daniel Philipp AND Füller, David AND Hamm, Benjamin AND Schreyer, Andreas AND Ritter, Oliver AND Maier-Hein, Klaus},
title = { { Exploiting Longitudinal Context in Clinician-Verified Interactive Lesion Tracking } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 16887},
month = {September},
page = {pending}
}
Reviews
Review #1
- Please describe the contribution of the paper
- A unified longitudinal tracking framework that has clinician interaction as well as automatic difference modeling for temporal changes.
- The PanTrack dataset is a potentially useful contribution to the community.
- 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.
-
Paper is very well written. Great to see authors mentioning several subtle nitty-gritty implementation details, which are often overlooked. Even small details about dataset, baselines, and evaluation fairness, are all included.
-
Framework is well motivated and grounded by existing literature. The authors clearly highlight the driving factors for each component of their method.
-
Obtaining first place in a MICCAI challenge signifies real-world usability.
-
- 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.
Mostly, the paper is strong enough. maybe a bit more info about the prompting strategies (scribbles or boxes instead of points) can be useful.
- 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?
Paper is strong enough for acceptance.
- 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
This paper proposes a clinically motivated verified tracking setup for longitudinal lesion analysis in CT. The basic idea is to use registration to propose the followup lesion location, allow the clinician to verify or correct that point, and then let the segmentation model delineate the followup lesion using both the corrected prompt and the baseline lesion appearance. The model combines early prompt fusion, latent temporal difference weighting, and synthetic longitudinal pretraining. The paper also introduces PanTrack, a new longitudinal pancreatic cancer dataset for external evaluation.
- 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 tackles a clinically relevant problem and addresses a real weakness of fully automatic longitudinal tracking methods. 2.The ablation study is convincing and does a good job showing why synthetic pretraining and temporal difference weighting matter. 3.The PanTrack release is a useful contribution, since public longitudinal datasets with lesion-level annotations across timepoints are still limited.
- 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 work has limited novelty at the component level. Several ingredients are closely related to prior work already cited by the authors, including registration-based tracking pipelines, temporal difference weighting, and promptable segmentation models. The main novelty here is more in the overall workflow than in a new algorithm. 2.No real user study was provided on the workflow. The verified setting is simulated using the gt follow up centroid. This is useful for the experiments, but it does not show how often a radiologist would need to correct the propagated point or how much effort that would require in practice. 3.Synthetic pretraining is underexplained. Since synthetic longitudinal pretraining seems to be an important part of the method, the paper should provide more detail on how lesion changes such as growth and shrinkage were simulated, and how realistic these synthetic followups are.
- 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.
(4) Weak Accept — marginally above the acceptance threshold, but would not mind if rejected, dependent on rebuttal
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
This work addresses a clinically relevant problem and includes external evaluation on PanTrack, and the dataset release is a valuable contribution. Main limitation is that the novelty is more at the integration level than at the level of a new algorithm.
- 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 introduces a longitudinal lesion tracking framework that combines:
A “verified tracking” paradigm, where registration-proposed prompts are optionally corrected by a clinician A longitudinal promptable segmentation model, integrating: early prompt fusion with latent temporal difference weighting A large-scale synthetic pretraining strategy for promptable longitudinal data
The method is evaluated on autoPET IV and a newly curated dataset, showing improvements over existing automatic and interactive tracking approaches.
- 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.
Clear problem formulation with clinical relevance; introducing a meaningful middle ground between automation and interaction which is both realistic and clinically motivated
A complete pipeline contribution with registration, prompting, temporal modeling and interaction
Clear gains over prior methods: autoPET IV (verified): 73.7 vs 68.7 DSC PanTrack: consistent improvements across metrics, which are statistically and practically meaningful.
Insightful ablation results; proving that naive longitudinal modeling fails, synthetic pretraining is necessary and temporal difference weighting is critical
Dataset contribution : PanTrack dataset (161 scans) is valuable considering longitudinal data is scarce and OOD evaluation is important in such cases.
- 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 early prompt fusion + latent temporal weighting may be considered as a well-engineered system, not a fundamentally new method.
Synthetic data realism is questionable 2,606 synthetic CT pairs generated via deformation fields, Simulated tumor growth/shrinkage
Performance gains may rely on synthetic bias rather than true generalization.
- 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?
The paper presents a well-engineered and clinically relevant framework for longitudinal lesion tracking, with strong empirical results and a meaningful dataset contribution. While the methodological novelty is limited and much of the performance gain may stem from engineering and pretraining, the overall system and evaluation are convincing enough to provide practical value to the community. The work would benefit from deeper analysis and clearer isolation of contributing factors.
- 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
Author Feedback
We would like to thank the reviewers and meta-reviewer for their detailed and constructive feedback, and for their positive assessment of our work. We appreciate the recognition of the clinical motivation, clarity, experimental validation, and dataset contribution. We would like to take this opportunity to briefly clarify several points raised in the reviews.
Novelty: We agree with Reviewers #3 and #4 that some individual components of our framework build on established ideas, including registration-based tracking, promptable segmentation, and temporal modeling. Our main contribution lies in formulating and validating the full verified longitudinal tracking workflow, which integrates these components into a clinically motivated pipeline. Importantly, the proposed method outperforms related approaches that rely on similar underlying components, suggesting that the integration and task-specific design choices are critical for practical longitudinal lesion tracking.
Prompting strategy: In this work, we focused on single positive point prompts as a minimal and clinically lightweight form of interaction. We agree with Reviewer #2 that richer prompts, such as scribbles or bounding boxes, could provide stronger guidance and may further improve performance. Exploring these interaction modes will be an important direction for future work.
Synthetic pretraining: We used the synthetic longitudinal dataset introduced with LesionLocator. Further details on the augmentation procedure can be found in that work. The goal of this pretraining stage is not to perfectly simulate all aspects of longitudinal disease progression, but to help the model learn temporal correspondence between scans, which we found difficult to learn from scratch. Indeed, the pretrained model does not perform competitively when applied to the downstream task without any fine-tuning. To preserve realism in the follow-up target during pretraining, we use the augmented volume as the baseline image and the original real volume as the follow-up image.
User study: We agree that a clinical user study would be valuable for assessing how often clinicians need to correct registration-propagated prompts and how much effort this requires in practice. In this paper, we simulated the verified setting using the ground-truth follow-up centroid in order to isolate the methodological contribution and provide a controlled evaluation. A prospective user study with radiologists is an important next step for validating the practical usability of the proposed workflow.
Finally, we would like to thank the reviewers and meta-reviewer again for recognising our contribution. We are grateful for the opportunity to present this work at MICCAI.
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.
This paper presents an interactive longitudinal lesion tracking framework. It combines registration-based matching proposals, optional clinician correction and promptable segmentation with temporal difference modeling. Additionally, a new longitudinal dataset for pancreatic cancer will be released.
Reviewers agree that the paper is well-motivated, clearly written and clinically relevant. The idea of lesion matching verification is a meaningful middle ground between fully automatic and fully manual workflows. The experimental setup is appropriate and the ablations convincing. The dataset contribution is considered valuable.
Reviewers’ concerns are related to limited methodological novelty (R#3) and the use of synthetic longitudinal pretraining, which could be clarified further.
Overall, the reviewers agree that the paper has string practical value and presents solid experimental evidence. My recommendation is accept.