List of Papers Browse by Subject Areas Author List
Abstract
Percutaneous thermal ablation is a localized treatment for hepatic tumors for which multi-needle planning remains mostly manual, while the difficulty grows sharply with the number of needles and the complexity of the targets. We present a GPU-accelerated pipeline that jointly optimizes all needle placements and ablation parameters in a single run, to aim for a perfect burn. Our pipeline uses a decomposition-based multi-objective particle swarm optimizer, with four clinical objectives (coverage, efficiency, safety, insertion depth), per-needle activation variables to optimize needle count, a CUDA-based fitness evaluation, and Vulkan-rendered safety maps. Tested on 94 tumors, the method achieves better coverage, larger distance to critical structures, and less collateral damage than the expert plan, in less than 6 seconds. An expert rated 88% of auto-plannings as clinically acceptable as is (≥4/5) and 99% as requiring only minor adjustments (≥3/5), with 97% judged equivalent or superior to manual plans (87% preferred, 10% equivalent).
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026/paper/5166_paper.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to the Code Repository
N/A
Link to the Dataset(s)
Planning dataset: https://figshare.com/s/13c01677c744eb1625c6
BibTex
@InProceedings{MehJon_Planning_MICCAI2026,
author = { Mehtali, Jonas AND Verde, Juan Manuel AND Essert, Caroline},
title = { { Planning the Perfect Burn: Multi-objective PSO for Multi-needle Thermal Ablation } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 16893},
month = {September},
page = {pending}
}
Reviews
Review #1
- Please describe the contribution of the paper
This paper presents a GPU-accelerated pipeline for automatic multi-needle percutaneous thermal ablation planning of tumors in the liver. The proposed method jointly optimizes needle placements, ablation parameters, needle count, and pullback strategies in a single optimization run using a decomposition-based multi-objective particle swarm optimizer (MOEA/D–PSO) with four clinical objectives: coverage, efficiency, safety, and insertion depth. The pipeline incorporates 3D-rendered safety maps, GPU-accelerated fitness evaluation, and per-needle activation variables to adaptively determine needle count. The authors evaluated the workflow on 94 tumors across 50 clinical cases separated into 3 difficulty tiers. Compared to expert planning, the method achieved superior coverage, better safety margins, and less collateral damage. Expert review rated 88% of plans as clinically acceptable and preferred the auto-planner in 87% of cases. The auto-planner executed in a median of 5.2 seconds on a consumer-grade NVIDIA RTX 4080 SUPER, which represents strong clinical potential.
- 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 addresses an important and well recognized clinical gap towards optimizing multi-needle ablation plans across multiple objectives directly relevant to the clinical task. The four primary objectives (coverage, efficiency, safety, depth) are formulated in accordance with the decision criteria used in interventional radiology, and other aspects such as the inclusion of pullback ablations, variable needle count via activation encoding, and safety-aware trajectory selection within the optimization loop represent important innovations that target key clinical processes. Prior works have not adequately addressed these clinical aspects, and the present contribution represents a strong advancement towards a clinically relevant ablation planning utility.
The evaluation on 94 tumors across 50 cases stratified according to COLLISION difficulty provides insight into task planning impact. Paired comparisons against expert plans with quantitative and qualitative experimental endpoints constitutes strong experimental design. Transparency about failure modes (e. g. one unacceptable case, extrahepatic pathing issues, extra needles, etc. ) and cognitive biases in manual planning adds important insights and credibility.
The public release of expert plans alongside the CRLM database is a notable contribution currently missing within the domain. This enable future work and benchmarking, and overall further improves the impact of the paper.
- 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 algorithmic novelty appears to primarily focus on the broad formulation and technical integration rather than methodological innovation. However, the techniques implemented are very strongly grounded and are applied in a rigorous manner into the application at hand, and represents a substantial engineering effort to build a clinically useful pipeline.
The use of manufacturer-calibrated ellipsoids as ablation zone estimates is a substantial simplification. Although the authors acknowledge that “future work will integrate physics-based ablation zone simulation” to account for very real patient-specific variations in the resulting treatment zone, the paper does in my opinion over-claim in promising that it will be able to provide “the perfect burn” without accounting for perfusion differences, steatotic differences, etc.
Minor comments include:
- It is not specified how many years of clinical experience are possessed by the expert reader.
- Statistical reporting includes Cohen’s d and q-values, but it is not described what inferential tests or procedures led to the q-values reported.
- There are many hyperparameters that may impact overall performance, but the space afforded by the MICCAI format is not conducive to deeper characterization, and so this reviewer does not explicitly expect an analysis on parameter sensitivity or optimal operating ranges.
- 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?
This is a well-executed systems paper that integrates multiple engineering contributions (GPU safety maps, CUDA fitness evaluation, MOEA/D–PSO with domain-specific operators) into a clinically useful pipeline for multi-needle ablation planning. The clinical evaluation is more thorough than most comparable work. The paper makes a rigorous practical contribution that has very strong potential clinical impact, and delivers compelling evidence justifying its clinical promise.
- 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 GPU-accelerated approach for optimizing needle placement in thermal tumor ablation, taking into account multiple clinically relevant factors, including treatment coverage, safety, efficiency, and needle insertion depth. The study demonstrates the potential clinical feasibility and significance of the proposed method. The work aligns well with the scope of the MICCAI community and represents a meaningful integration of medical image computing (MIC) and computer-assisted intervention (CAI).
- 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 manuscript is well structured and easy to follow.
- The proposed optimization framework is well formulated and demonstrates novelty by jointly optimizing needle placement while considering multiple clinically relevant factors, highlighting its potential clinical significance.
- The validation is clearly presented in the context of the optimization task, utilizing manufacturer-provided ablation parameters.
- 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.
- When multiple needles are inserted in close proximity, the resulting ablation zone is typically influenced by the interaction between individual ablation zones, rather than being a simple superposition of them. This phenomenon is also reflected in manufacturers’ preset models. To clarify this point, the manuscript should include a discussion of these interaction effects and how they may impact the proposed approach.
- The proposed approach involves numerous hyperparameters that require tuning for optimal performance. From a reproducibility perspective, it is unclear how practical it would be for users to tune these parameters in order to achieve results comparable to those reported in this manuscript.
- It is unclear to the reviewer how the safety map is generated in the proposed approach. Additional details on its formulation and underlying assumptions would be helpful. Furthermore, the manuscript does not clearly explain how this safety map is integrated into the needle planning optimization process. It is also unclear whether the safety map needs to be dynamically updated based on the needle position during planning or execution. Clarification on this point would be important for understanding the practical implementation of the method.
- The critical structures need to explicitly describe in this manuscript.
- 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 submission does not mention open access to source code or data, but provides a clear and detailed description of the algorithm to ensure reproducibility.
- 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 study demonstrates the potential clinical feasibility and significance of the proposed method. The work represents a meaningful integration of medical image computing (MIC) and computer-assisted intervention (CAI).
- 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 #3
- Please describe the contribution of the paper
This paper presents an automatic planning method of multi-needle thermal ablation for liver tumors. This is a known and significant clinical problem. Existing methods either optimize needles one by one, fix the needle count before optimization, or are too slow. The proposed method jointly optimizes all needle placements, ablation parameters, and trajectory safety in a single run. It uses a decomposition-based particle swarm optimizer with four clinical objectives: tumor coverage, ablation efficiency, trajectory safety, and insertion depth. Needle count is not fixed in advance. It is calculated from the optimization through per-needle activation variables. Safety is integrated by pre-computed GPU-rendered maps, avoiding expensive constraint checking during search. The system runs in a median of 5.2 seconds per target. The method is evaluated on 94 tumors from 50 cases across three difficulty tiers. The auto-planner achieves full margin coverage in all cases and significantly outperforms expert plans on DSC (0.892 vs. 0.525) and efficiency (0.829 vs. 0.394). A single expert reviewer rated 88% of plans as fully acceptable and preferred the auto-planner in 87% of cases. These are strong results, though the reliance on a single reviewer is a limitation that the authors should acknowledge more explicitly.
- 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.
Strong clinical results with expert validation. The performance gap over expert planning is large and statistically significant. The expert review goes beyond what most planning papers provide. The reviewer’s acknowledgment of their own cognitive biases adds credibility to the analysis.
First joint method addressing all five identified gaps at intraoperative speed. The paper identifies what prior methods are missing and addresses all five gaps in a single system. The 5.2 s runtime is a practical advance over simulation-based methods and makes intraoperative use realistic.
Adaptive needle count without pre-commitment. Needle count is computed from the optimization rather than being fixed in advance. Prior methods commit to needle count before optimization, which can lead to suboptimal plans.
Thorough evaluation, testing on 94 tumors across three difficulty tiers, reporting effect sizes and corrected p-values, and combining quantitative metrics with expert review. The public release of the dataset is a valuable contribution.
Native encoding of pullback burns along a single needle trajectory is clinically relevant and not commonly addressed in automated planning systems.
- 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.
All qualitative assessments come from one clinician. Inter-rater agreement cannot be assessed. The reviewer’s stated biases may influence the comparison with their own manual plans. A second reviewer or inter-rater reliability measure should be reported.
Ablation zone model is not physics-based. The system uses manufacturer-calibrated ellipsoids to model ablation zones. Heat sink effects from nearby vessels and patient-specific tissue properties are not modeled. In many cases the nearest obstacle is a vessel where these effects matter most. The gap between the modeled and actual ablation zone is not quantified.
Recurring failure modes are not fully analyzed. In some of plans the needle exits the liver after entry, have superfluous needles, or have overly tangential approaches. These are systematic issues, and their clinical impact is not discussed.
For the most complex cases (multi-lesion, bilobar), more detail on how the system handles inter-target interactions would strengthen the clinical argument.
No information is given about the expert’s experience level or the time taken to produce manual plans. It is unclear if the comparison represents a typical skilled clinician.
- 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 submission has provided an anonymized link to the source code, dataset, or any other dependencies.
- 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?
This is a strong CAI paper that addresses a real clinical problem with a well-evaluated system. The results are convincing. The public dataset release is an additional community contribution.
- 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
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.
This well-written paper with informative visuals presents a well-engineered percutaneous liver tumor ablation planning system that uses a GPU-accelerated pipeline to jointly optimize needle placement, ablation parameters, and trajectory safety in ~5 seconds per target. Evaluated on 94 tumors across 50 patients with expert validation, the system demonstrates high potential for clinical impact. The public release of the dataset constitutes a valuable contribution to the field.