List of Papers Browse by Subject Areas Author List
Abstract
Clinicians lack a principled framework to quantify diagnostic utility in ultrasound reconstructions.
Existing standards like PSNR and VGG-LPIPS are inadequate, failing to account for modality-specific physics or the structural nuances of acoustic imaging.
We close this gap with a TinyUSFM-based evaluation framework featuring two distinct metrics: TinyUSFM-uLPIPS, a full-reference perceptual distance based on multi-layer token relations, and TinyUSFM-NRQ, a deployable no-reference quality score utilizing clean-manifold modeling and worst-region aggregation to detect localized harmful artifacts.
We demonstrate that the presented metrics have four unique advantages: 1) Task-linked quality, where TinyUSFM-uLPIPS achieves superior calibration with semantic task damage, accurately reflecting Dice-score drops in segmentation where VGG-based metrics fail; 2) Cross-organ comparability, maintaining stable scoring scales and consistent severity rankings across diverse anatomical sites and domain-shifted data; 3) PSNR-consistent sensitivity, with TinyUSFM-NRQ providing a reliable quality score without ground-truth images that remains consistent with traditional fidelity benchmarks (i.e. PSNR); and 4) Clinical utility, improving the prediction of expert preference from 47.2% to 72.8% accuracy and producing super-resolution reconstructions preferred by sonographers. By integrating these advantages into a unified assessment and optimization loop, this work establishes a modality-aligned standard that finally bridges the gap between algorithmic performance and diagnostic utility. Our code is available at https://github.com/sextant-fable/US-Metrics.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026/paper/2414_paper.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to the Code Repository
https://github.com/sextant-fable/US-Metrics
Link to the Dataset(s)
CUBS: https://data.mendeley.com/datasets/m7ndn58sv6/1
UF1990: https://data.mendeley.com/datasets/n2zcmcypgb/2
STMUS: https://data.mendeley.com/datasets/3jykz7wz8d/1
AUL: https://doi.org/10.5281/zenodo.7272660
BUSI: https://scholar.cu.edu.eg/?q=afahmy/pages/dataset
MMOTU: https://github.com/cv516Buaa/MMOTU_DS2Net
DDTI: http://cimalab.intec.co/applications/thyroid/
KidneyUS: https://rsingla.ca/kidneyUS/
CAMUS: https://www.creatis.insa-lyon.fr/Challenge/camus/databases.html
BibTex
@InProceedings{HuaZiy_Defining_MICCAI2026,
author = { Huang, Ziyang AND Li, Bingyan AND Ma, Chen AND Liu, Tianyi AND Zhai, Yihui AND Xu, Hong AND Guo, Yi AND Li, Zeju AND Wang, Yuanyuan},
title = { { Defining Robust Ultrasound Quality Metrics via an Ultrasound Foundation Model } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 16878},
month = {September},
page = {pending}
}
Reviews
Review #1
- Please describe the contribution of the paper
The paper proposes a TinyUSFM-based framework with full- and no-reference metrics that better capture clinically relevant degradation than standard perceptual measures. It demonstrates improved alignment with task performance and clinician preference.
- 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.
Well-motivated problem with clear clinical relevance. Strong use of domain-specific foundation model features, comprehensive evaluation (task-based, cross-organ, clinician study), and practical no-reference metric.
- 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.
Limited comparison with recent learned IQA methods and reliance on a specific foundation model. Insufficient ablations and limited detail on clinical study and computational cost. The tests were done at very low quality values (PSNR range 20-25).
- 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
This is globally a good and timely contribution. A few clarifications could further strengthen the paper: Could the authors clarify how their approach compares to more recent learned IQA methods in medical imaging or ultrasound specifically? To what extent do the gains come from the metric design versus the TinyUSFM backbone? Have alternative feature extractors been considered? Could the authors provide more details on the clinician study (e. g. , number of clinicians, inter-rater agreement, case diversity)? Additional ablations (e. g. , layer contributions in uLPIPS or aggregation in NRQ) would help better understand key design choices. Tests at higher PSNR values would help showing the validity of the metric at different quality ranges.
- 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 addresses an important problem in ultrasound imaging and proposes a technically sound and well-motivated solution using foundation model features. The introduction of both full-reference and no-reference metrics, along with strong validation through task-based evaluation and clinician studies, represents a meaningful contribution. While there are some limitations (e.g., limited comparison with recent methods, lack of ablations, and dependence on a specific model), the work is considered valuable and relevant to the MICCAI 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 #2
- Please describe the contribution of the paper
This paper introduces two novel ultrasound metrics: (1) TinyUSFM-uLPIPS: A full-reference perceptual distance metric; (2) TinyUSFM-NRQ: A no-reference quality score. The author systematically validates four key advantages of the framework: task-linked quality, cross-organ comparability, PSNR-consistent sensitivity, and clinical utility. And utilizes the TinyUSFM feature space as a perceptual loss for image super-resolution, producing reconstructions preferred by clinicians.
- 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.Methodological Novelty: The core innovation lies in leveraging the feature space of an ultrasound-specific foundation model (as opposed to natural-image models) to define quality metrics. This directly addresses the fundamental issue of domain mismatch. 2.Comprehensive and Convincing Evaluation: The validation is exceptionally thorough, demonstrating value from multiple complementary angles.
- Task-Linked Quality: Using the drop in segmentation Dice score as a “ground truth” to prove the proposed metrics better reflect semantic damage than VGG-LPIPS.
- Cross-Organ Generalizability: Testing on a large-scale dataset covering multi-organs demonstrates the stability and comparability of the metrics.
- Clinical Alignment: The blinded clinician study increases the accuracy of predicting expert preference from 47.2% to 72.8%.
- Optimization Application: Serving as a perceptual loss, it guides the generation of super-resolution images superior in both fidelity and clinical acceptability.
- 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.
Major concerns: 1.Dependence on the Foundation Model: The framework’s performance and generality heavily rely on the quality and training data of the TinyUSFM model. The paper does not deeply discuss how potential biases in the foundation model itself (e.g., on rare pathologies or images from novel devices) might affect the reliability of the metrics. While inherent to foundation model-based approaches, this should be acknowledged as a limitation. 2.Limitation in Evaluation Scope: Experiments focus primarily on static images. Ultrasound is often used dynamically (e.g., echocardiography, fetal screening). The paper does not explore the metric’s capability to assess dynamic quality factors like temporal consistency or motion artifacts. This is an important future direction but can be noted as a current limitation. 3.Comparison with a Broader Range of NR Methods: While compared with NIQE, BRISQUE, and KDE-UQA, other learning-based methods in medical IQA exist. A more detailed discussion on the distinction and potential advantages of the proposed approach compared to these methods would be beneficial. Minor concerns: The presentation of the figures can be more visually appealing, for instance, by reducing the overlap between text and legends (Figures 2 and 3).
- 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 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?
This paper addresses the critical disconnect between algorithmic metrics and clinical utility in ultrasound imaging, holding high clinical and practical value. And presents the first unified quality assessment framework leveraging an ultrasound foundation model. The design of the two novel metrics is ingenious and tailored to ultrasound physics. The evaluation process is excellently designed, providing a complete evidence chain from numerical simulation, downstream tasks, and cross-domain generalization to blinded clinician studies, making the claims highly convincing.
- 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
Review #3
- Please describe the contribution of the paper
This paper introduces two new quality metrics for ultrasound images - the USFM-NRQ (no reference) and the USFM-LPIPS (full reference). The tiny USFM is used to reduce compute.
- 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 problem is introduced neatly, and the literature on the state-of-the-art nicely summarises the context of the problem.
The mathematics is explained at a very good level.
The experiments thoroughly show several downstream tasks, and how the created metrics influence or affect the tasks.
Open source datasets are used, which improves replicability.
Comparisons with SOTA methods are clearly made in both full reference and no reference cases.
A clinical study is also used to put these metrics into context with expert ratings.
- 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 diagrams are hard to read an need to be made more clear.
While explanatory, the writing in this paper is very dense, which affects readability.
To make this work self contained, a basic overview of USFM should also be included. I understand this is omitted due to space concerns.
- 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.
(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?
This paper addresses an impoortant problem - quality analysis in medical imaging. It also shows solutions, which are based on leveraging existing models. Unlike most IQA papers, provisions are made for both the full reference and no reference cases. Experimentally, many uses of the metrics have been shown.
- 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 sincerely thank the reviewers, Area Chairs, and Program Chairs for their careful reading, constructive comments, and positive assessment of our work. We are grateful that the reviewers recognized the significance of addressing the disconnect between conventional algorithmic image-quality metrics and clinical utility in ultrasound imaging, as well as the novelty of using an ultrasound foundation model to define both full-reference and no-reference quality metrics.
The reviewers’ comments provided useful guidance on how to make the paper clearer and more balanced. In the camera-ready version, we will more explicitly discuss the limitations of relying on a TinyUSFM backbone, including potential issues related to foundation-model bias, rare pathologies, unseen devices, and domain shift. We will also sharpen the discussion of related work by comparing our framework more clearly with recent learned IQA methods in medical imaging and ultrasound, and by distinguishing the contributions of the metric design from those of the ultrasound-specific representation.
We will revise the camera-ready manuscript to improve clarity and readability without changing the fundamental content of the work. Specifically, we will add a more explicit limitation discussion on the reliance on TinyUSFM and on the current focus on static ultrasound images, noting dynamic quality factors such as temporal consistency and motion artifacts as important future directions. We will also clarify the clinical study protocol, including reader evaluation details and case diversity, where space permits. In addition, we will improve the presentation quality of the figures and reduce visual clutter, as suggested by the reviewers.
We also appreciate the comments regarding ablations, higher-PSNR settings, and alternative feature extractors. While the final manuscript must remain within the scope of the submitted work, we will make the existing design choices and evaluation scope more transparent, and we will discuss these points as useful directions for future work. Overall, the feedback has helped us better understand how to present the contribution as a clinically motivated, modality-aligned framework for ultrasound image quality assessment, and we thank the reviewers and meta-reviewer again for their constructive evaluation.
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.
All three reviewers recommend Acceptance. The consensus is on the significance and novelty of this work, which addresses the critical disconnect between algorithmic metrics and clinical utility in ultrasound imaging. The paper presents the first unified quality assessment framework built upon an ultrasound foundation model, introducing two well-motivated metrics tailored to ultrasound physics that cover both full-reference and no-reference scenarios. The evaluation is comprehensive and convincing, spanning numerical simulations, downstream task performance, cross-domain generalization, and blinded clinician studies, forming a complete evidence chain that strongly supports the claims. For all these reasons, recommendation is towards Acceptance.