List of Papers Browse by Subject Areas Author List
Abstract
Ultrasound imaging is widely used clinically due to its real-time capability, non-invasive nature, and low cost. Novel view synthesis in ultrasound allows better anatomical understanding, consistent evaluation across standard planes, and review of regions poorly captured during scanning. However, generating realistic novel views is challenging due to the prevalence of speckle patterns arising from complex wave–tissue interactions. These fine-grained textures often carry diagnostic significance, yet existing rendering methods frequently oversmooth or distort speckle, limiting performance in downstream tasks that rely on accurate speckle preservation. In this paper, we propose BIGUS, a 3D Gaussian Splatting–based framework for ultrasound novel view synthesis. We redesign the Gaussian rendering pipeline with a physics-aware scattering formulation that better captures the characteristics of ultrasound physics. In addition, we introduce a structure-aware initialization strategy that adaptively samples Gaussian primitives from tissue regions, combined with an error-guided densification mechanism to improve reconstruction fidelity. We evaluate BIGUS on three diverse ultrasound datasets, including fetal brain, forearm, and brain imaging, each with distinct speckle patterns and probe geometries. To better assess speckle preservation, we propose the Backscatter Fidelity Score, a speckle-sensitive evaluation metric that complements conventional image quality measures. Our experiments show that BIGUS achieves 67.26 dB PSNR on the fetal brain dataset, establishing state-of-the-art performance with only 140k Gaussians on a single GPU and training in under 15 minutes. Moreover, it outperforms existing methods by up to 9 dB PSNR on speckle-dominant datasets. Code will be released upon acceptance.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026/paper/5438_paper.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: https://papers.miccai.org/miccai-2026/supp/5438_supp.zip
Link to the Code Repository
https://github.com/YoussifKhaled/BIGUS.
Link to the Dataset(s)
N/A
BibTex
@InProceedings{AbdYou_BIGUS_MICCAI2026,
author = { Abdelaziz, Youssif AND Ismail, Malak AND Mourad, Nada AND Badran, Mohamed AND Torki, Marwan},
title = { { BIGUS: Beam Integrated Gaussian Scattering for Continuous Ultrasound View Synthesis } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 16890},
month = {September},
page = {pending}
}
Reviews
Review #1
- Please describe the contribution of the paper
This paper introduces BIGUS, a new Gaussian Splatting method dedicated for ultrasound novel view synthesis. The proposed method is designed to reconstruct ultrasound volume while preserving speckle patterns. By preserving the speckle patterns, BIGUS could achieve US reconstruction realism. The authors present several experimental results and visual evidence to demonstrate BIGUS’s improvement from previous methods, like UltraGauss.
- 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.Strong quantitative results compared to baselines. 2.The ablation studies prove that the modified densification does improve over the conventional ADC, as the authors suggested. 3.BIGUS also adapted ultrasound physics for rendering by integrating the beam profile.
- 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.GS methods are also vulnerable to inaccurate pose priors, but it’s unclear what the impact of the noisy transducer poses is. 2.There are a few 3DGS methods, such as 2DGS, which are also designed to better align with the object surface by modeling the primitive within a disk. Maybe the authors should include a discussion of variations of 3DGS. 3.It is unclear why the Harris corner detector was used in the proposed method. 4.Lack of clinical validation for the proposed BFS as the evaluation metric.
- 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 authors explain why standard 3DGS fails for ultrasound well, and propose several improvements to address this issue.
- 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 authors propose a new Gaussian Splatting rendering technique for ultrasound images. This techniques solves two issues of current alternatives; 1.considers the beam elevation width (so imaging plane is not infinitely thin), and considers scatterer-produced destructive interference (so speckle is consistent).
- 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 is well written, clear, concise and mathematically sound
- The method is appropriately justified and reasoned,
- The results are significant and well above competing methods
- 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.
- Data is somewhat limited to fetal brain - it would be interesting to see how the technique holds to more challenging scenarios such as moving organs (e.g. the heart), or high resolution ultrasound (e.g. superficial vessels).
- It would be interesting to understand how the method behaves for artefact rich data (e.g. shadows, reverberation) and for low scattering media (e.g. blood pool in cardiac imaging).
- 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.
(6) Strong Accept — must be accepted due to excellence
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
The paper is well articulated, proposes a novel technique that is interesting and well validated. I think this is a solid paper.
- 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
1.The paper introduces a physics-aware ultrasound renderer with signed reflectivities and exact elevation-beam integration for Gaussian splatting.
2.The paper adds structure-aware Harris initialization and error-attributed densification to better place and refine Gaussian primitives.
3.The paper proposes BFS, a new metric intended to measure speckle fidelity, and evaluates the method on three ultrasound datasets.
- 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 because preserving speckle is relevant for ultrasound realism and potentially for downstream analysis.
2.The core method is technically novel: specifically, the closed-form beam integration and signed scattering formulation.
3.The reported improvements are consistent on the shown benchmarks.
4.The method appears computationally efficient.
- 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 ablation study is incomplete because it does not isolate the main claimed innovations: signed reflectivity, exact beam integration, and Harris initialization are not individually ablated.
2.The physics model is still approximate: interference is reduced to signed scalar reflectivities plus rectification, not a full RF/phase model.
3.Reproducibility details are incomplete, including limited experimental protocol detail.
- 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?
The paper makes a concrete contribution, with robust evaluation and demonstrated improvements. I lean towards acceptance for this 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
Author Feedback
We sincerely thank all reviewers for their constructive and insightful feedback. We address each concern below.
Regarding the incomplete ablation study [Reviewer 3 W1; Reviewer 1 W3]: We agree with the reviewer on the importance of additional ablation studies. We have added new ablation experiments that individually isolate the effects of signed reflectivity, exact beam integration, and Harris initialization, and these results will be included in the camera-ready version. These experiments were conducted during the development of our method and demonstrated the importance and contribution of each component individually.
Regarding our physics model is approximate [Reviewer 3 W2]: Our model is a deliberate approximation that balances physical accuracy with optimization tractability. A full RF/phase model, while theoretically richer, is extremely difficult to optimize end-to-end, as per our initial experiments, where we tried to develop full RF/phase model but it failed due to the following: Gauge symmetry in (a,φ) ≡ (a,φ+2πk), causing the optimizer to drift. A phase saddle at φ=0 where ∂C/∂φ vanishes leaves no escape signal. B-mode envelope detection discards phase, so an explicit φ has no supervisable target. We therefore opt for a simpler but more physics-aware approximation, which produces the higher-fidelity novel view synthesis evidenced by our state-of-the-art results.
Regarding the reproducibility details [Reviewer 3 W3]: Complete implementation details, hyperparameters, and experimental protocols will be made fully available through the code repository linked in the camera-ready version, ensuring complete reproducibility.
Regarding the vulnerability to inaccurate pose priors [Reviewer 1 W1]: Gaussian splatting is an explicit representation; therefore, the reconstruction quality is inherently dependent on the accuracy of the input poses. Consequently, errors in pose estimation propagate to the reconstructed volume and affect reconstruction fidelity. In our experiments, we evaluated the method on freehand ultrasound datasets that naturally contain pose noise arising from the acquisition process itself. Therefore, all reported results already reflect realistic pose inaccuracies present in practical settings. Furthermore, to explicitly study the sensitivity of the method to pose perturbations, we added additional synthetic pose errors and observed a degradation in reconstruction quality, confirming the strong dependency between pose accuracy and reconstruction fidelity. Nevertheless, pose estimation is treated as an external input to our pipeline, and handling pose correction or refinement lies outside the scope of the current work.
Regarding the clinical validation of BFS [Reviewer 1 W4]: Clinical validation is an unreliable process for measuring speckle-pattern fidelity. Speckle is noise-like by construction, and subjective judgments between two speckled reconstructions do not consistently discriminate well-preserved patterns from poorly-preserved ones. We therefore introduce BFS as a qualitative and standalone measure of speckle preservation. The signal-level preservation it promotes is not only aimed at clinical viewing, it is also aimed at downstream tasks that consume speckle as information, such as sensorless pose estimation via speckle decorrelation, tissue characterization, and speckle-based tracking and segmentation.
Regarding method behavior in different and artifact-rich scenarios[Reviewer 2 W1,2]: Our method requires posed ultrasound datasets that are scarce. We evaluated the three publicly available datasets that we could find at the time of submission. These datasets cover low-, mid-, and high-speckle regimes as well as spanning over multiple anatomical regions from forearms to brain. Regarding moving organs such as the heart, this would require 4D (dynamic) Gaussian Splatting, which is orthogonal to our contribution and constitutes a compelling direction for future work.
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 all agree that the paper makes a novel and sound methodological contribution to ultrasound simulation based on Gaussian splatting (GS). The proposed approach relies on adding physics-aware and ultrasound-specific modeling to the GS. Reviewers acknowledge the clarity of the method description and the superiority of its performance compared with the state of the art. Questions were raised regarding the sensitivity to inaccuracies in pose priors and the performance on artifact-rich and low-scattering media. Importantly, it was also noted that details should be included to ensure reproducibility and that the ablation study could be further expanded.
