List of Papers Browse by Subject Areas Author List
Abstract
Patient-specific hemodynamic assessment is critical for intracranial aneurysm rupture risk stratification, yet standard computational fluid dynamics (CFD) pipelines are too time-consuming for routine clinical use. Existing learning-based surrogates enable fast inference but often produce over-smoothed velocity fields, obscuring near-wall gradients and fine-scale vortical structures that are important for clinically relevant biomarkers. We propose PICA, a boundary-conditioned implicit neural field for continuous 3D hemodynamic prediction directly from voxelized aneurysm geometries. PICA encodes signed distance field (SDF)-augmented inputs with a multi-scale backbone and conditions the representation on patient-specific inlet parameters via feature-wise modulation, enabling robust adaptation across flow regimes. An implicit coordinate-based decoder queries the learned representation at arbitrary 3D locations to generate dense velocity fields without meshing or geometric reconstruction. To improve physical fidelity, we incorporate physics-guided regularization that enforces incompressibility and steady-state Navier–Stokes consistency, together with a progressive two-stage training curriculum to enhance cross-anatomy generalization. On the Aneumo benchmark, PICA achieves sub-second inference and reduces Mean Normalised Absolute Error by 25.4% compared to DeepONet-SwinT, while better recovering complex intra-saccular vortices and near-wall flow patterns.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026/paper/6349_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)
https://huggingface.co/datasets/SAIS-Life-Science/Aneumo
BibTex
@InProceedings{HuMen_PICA_MICCAI2026,
author = { Hu, Mengfan AND Li, Siyu AND Zhang, Zeng},
title = { { PICA: Physics-Guided Implicit Neural Fields with Boundary Conditioning for Intracranial Aneurysm Hemodynamics } },
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
The paper proposes a method to obtain 3D blood velocity fields in intracranial aneurysms. The approach uses CFD simulation data to train a neural surrogate based on implicit neural representation. Clear contributions of the paper are: conditioning the model on boundary conditions, incorporating physics loss (also through the vorticity transport equation), and leveraging two-stage curriculum learning to improve 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.
The approach does not need meshing since it is based on INRs. The authors use a clever input representation and boundary conditioning via FiLM. The use of the vorticity transport equation as a physics loss is smart, since it does not require any information about pressure. Good comparison against other neural operator approaches.
- 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.
Hyperparameters are not discussed, such as the weights for the loss components. There are more rigorous ways to approach this. Also, the weight for the flow loss is several orders of magnitude smaller than the others, questioning its true utility. No investigation of pressure or WSS fields is a pity, since also the authors themselves stress the importance of WSS for clinical purposes. Generalization with respect to anatomical variability could have been studied better, perphaps including more datasets. It is strongly advised to release code open source.
- 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?
As stated above, the paper presents a novel workflow for estimating blood flow velocity fields from geometry information and boundary conditions.
- 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 addresses the clinically relevant problem of fast patient-specific hemodynamic prediction for intracranial aneurysms, with the goal of replacing the time-consuming conventional CFD pipeline and improving practical clinical usability. Overall, the proposed framework is designed to address the common issue of over-smoothed velocity fields in existing learning-based surrogate models, which often fail to recover near-wall flow details and complex vortex structures. In this regard, the integration of an implicit neural field, boundary-condition modulation, SDF-based geometric encoding, and physics-guided constraints into a unified framework is reasonably well motivated and has some task-specific practical value.
- 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 shows some novelty at the task-modeling level. By targeting the issues of over-smoothed velocity fields and limited clinical practicality in intracranial aneurysm hemodynamic prediction, the authors combine implicit neural fields, boundary-condition modulation, SDF geometric encoding, and physics-guided constraints into a unified framework. This design is reasonably targeted and has some practical relevance.
- 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 methodological novelty appears limited. Although the overall framework is relatively complete and tailored to the target task, most of its main components are based on established ideas, including implicit field representation, FiLM-style conditioning, SDF-based geometric encoding, and physics-guided regularization. As a result, the contribution seems to be more about integrating and adapting existing techniques to this application, rather than introducing a substantially new method. The authors should more clearly define the true novelty of the work and distinguish it more convincingly from prior studies. 2.The construction of the boundary-condition vector used in FiLM is unclear. The paper mentions that patient-specific boundary vectors, such as the inlet centroid position, surface normal direction, and volumetric flow rate QQQ, are used as inputs to the FiLM module. However, it is not clearly described how these quantities are obtained. Are they automatically computed from the geometry, or do they require additional annotations, preprocessing, or CFD-derived information? This point is important for reproducibility, automation, and eventual clinical applicability. The authors should provide more details on how these boundary-condition vectors are generated in practice. 3.The comparison with baseline methods is not sufficient. The current experimental comparison appears to include only two baseline methods, which is not enough to convincingly demonstrate the advantage of the proposed approach. For a paper on fast hemodynamic prediction, stronger experimental support would require comparison with a broader set of representative methods, especially recent learning-based CFD surrogate models or related continuous field / operator learning approaches. 4.The ablation study is still incomplete. Although the authors report ablations for w/o Geometric Encoding (Mask only), w/o Hierarchical FiLM (Late fusion), w/o Multi-scale Querying, w/o Physics Constraints, w/o Two-stage Training, and the full model, the current ablation study is still not sufficient to fully support the necessity of each design choice. In particular, for the geometric representation, the paper emphasizes the importance of SDF-enhanced input, but it does not report more fine-grained comparisons such as using SDF only or SDF + mask. Without these results, it is difficult to clearly assess the independent contribution of SDF itself and whether combining SDF with the mask is indeed superior to other geometric input configurations. Additional ablation results would strengthen the paper.
- 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 does not provide sufficient information for 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.
(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?
Overall, the paper addresses a meaningful problem and presents a reasonably well-motivated framework with some practical value. However, the current version still needs improvement in terms of novelty clarification, methodological detail, and the completeness of experimental validation. I encourage the authors to better articulate the main contribution, clarify key implementation details, and strengthen the comparison and ablation studies to improve the overall persuasiveness of the work.
- 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
An implicit neural field framework for 3D aneurysm hemodynamic prediction that combines SDF-enriched geometry encoding, hierarchical boundary conditioning, coordinate-based decoding, and physics-guided regularization to achieve fast, high-fidelity velocity field prediction.
- 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 tackles a clinically important problem with a well-designed approach. The vorticity transport loss directly targets the over-smoothing problem in existing surrogates rather than applying generic physics constraints.
- 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 baseline comparison is too narrow (two methods from the same family). Steady-state assumption contradicts the clinical motivation around OSI and temporal WSS. No clinically relevant downstream metrics are evaluated.
- 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.
(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?
Strong, well-executed methodological contribution, yet the experiment section is lacking baseline comparison and downstream task evaluation.
- 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.
The reviewers consistently agreed on the clinical motivation of the targeted problem and recognized the technical contributions of this work within the claimed scope. They also raised shared concerns regarding the evaluation, addressing which would meaningfully strengthen the impact and value of the work. Overall, I recommend acceptance.