List of Papers Browse by Subject Areas Author List
Abstract
Due to its historic limitations in imaging speed, for a long time, ophthalmic Optical Coherence Tomography (OCT) image analysis focused on features such as retinal layer thickness or fluids, which are visible in 2D B-scans. Although OCT angiography raised interest in en face (C-scan) images to visualize vascular layers, analysis of structural en face features like nerve fibers or photoreceptors was limited by transverse resolution, shot noise, and eye motion creating non-uniform displacements and distortion in raster-scanned volume images. Imaging of these structures is limited to much more complex adaptive optics imaging devices, which place a major hurdle for clinical translation, and complicate imaging by requiring pupil dilation. In this contribution, we integrate computational methods for motion, distortion and illumination compensation of repeated OCT volumes with orthogonally alternating fast scan directions, resulting in a consistent 3D point cloud, and apply the first fully-automatic inverse model for reconstructing 3D (multi-frame) super-resolved ophthalmic OCT images with higher sample density and reduced blur. In particular, we visualize reflections from individual cones in the outer segment with visible-light OCT across a 3 x 3 mm^2 field. Our method is unique in its comprehensive description of confounders of the imaging process, unlocks a novel paradigm of widefield OCT acquisition, and significantly broadens access to features of high clinical interest.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026/paper/4007_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)
N/A
BibTex
@InProceedings{PloSte_VolumeFused_MICCAI2026,
author = { Ploner, Stefan B. AND Fan, Weijia AND Zhang, Hao F. AND Fujimoto, James G. AND Maier, Andreas},
title = { { Volume-Fused Super-Resolution OCT Reveals Cellular-Level Structures in the Outer Segment Across Wider Fields Without Adaptive Optics } },
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
The central contribution is a comprehensive, model-based 3D super-resolution pipeline for ophthalmic OCT that unifies motion, distortion, illumination, and artifact compensation into a single inverse-problem framework. By formulating the image formation process explicitly and inverting it through non-iterative fusion followed by iterative deblurring with Tikhonov regularization, the method achieves cellular-level imaging of photoreceptor cone reflections in the outer segment over a widefield, without requiring adaptive optics hardware or pupil dilation. This broadens access to clinically relevant retinal microstructures that were previously observable only with significantly more complex and expensive AO systems.
- 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 imaging model in Eq. (1) is notably comprehensive: it jointly accounts for non-uniform motion (via spatiotemporal displacement parameterization), illumination bias (via per-A-scan coefficients with foreground masking), optical distortion (shearing and curvature), and noise. This level of completeness is uncommon in prior OCT super-resolution work and is well motivated by the practical confounders in ophthalmic imaging. 2.The method is purely computational and does not require hardware modifications, pupil dilation, or real-time eye tracking. This represents a significant practical advantage for clinical translation compared to AO-OCT, which demands expensive additional hardware and is restricted to small imaging fields. 3.The reproducibility experiment in Figure 3, where the super-resolution pipeline is applied independently to two disjoint subsets of five volumes each, is a convincing and well-designed validation strategy. It demonstrates that the observed cone reflections are consistent across independent data splits rather than artifacts introduced by the reconstruction. 4.The paper provides a clear and well-structured ablation across intermediate processing stages (single input, fused, supersampled, super-resolved) in Figure 2, which allows the reader to appreciate the incremental contribution of each component. The visual improvements in contrast, sharpness, and separability of cone reflections are substantial. 5.The clinical relevance is clearly articulated: S-cones, which are the likely source of the observed reflections, are known to be more vulnerable in diseases such as retinitis pigmentosa, diabetes, and glaucoma, making them valuable as early biomarkers.
- 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 evaluation is limited to eight healthy volunteers from a single imaging site using a single prototype device. No pathological data are included, which leaves unclear whether the pipeline generalizes to diseased retinas where tissue geometry, reflectivity, and motion patterns may differ substantially. Even a preliminary case or discussion of expected failure modes in pathology would strengthen the paper. 2.The paper lacks quantitative evaluation metrics. All results are presented qualitatively through visual inspection of en face images and B-scans. No resolution measurements (e. g. , Fourier ring correlation, point source analysis), no quantitative comparison against AO-OCT references, and no SNR or contrast-to-noise ratio analysis are reported. This makes it difficult to objectively assess the magnitude of the improvement or compare with other methods. 3.The choice of key hyperparameters, including the upsampling factor (3x), deconvolution sigma_tv = 2.9 um, regularizer weights lambda = 0.2 and gamma = 0.1, and step size delta = 0.12, appears to be fixed without justification or sensitivity analysis. The paper does not discuss how performance varies with these settings or how they should be adapted for different devices, wavelengths, or imaging fields.
- 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.
(3) Weak Reject — marginally below the acceptance threshold, but would not mind if accepted, dependent on rebuttal
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
While the proposed pipeline is comprehensive in its modeling of imaging confounders and demonstrates visually compelling results, the lack of any quantitative evaluation metrics makes it difficult to objectively assess the claimed resolution improvement or compare with existing methods. In addition, the evaluation is restricted to eight healthy subjects from a single site with no sensitivity analysis on key hyperparameters, leaving the generalizability and robustness of the approach insufficiently demonstrated for a venue of this caliber.
- 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
The authors propose an approach to the reconstruction of super-resolution OCT scans using computational methods without the use of adaptive optics. The concept of replacing adaptive optics with computational aberration correction methods has been around for a long time. However, this work presents a more comprehensive approach by integrating all elements of the distortions (motion correction, distortions, and super-resolution) into a single mathematical model. This mathematical model contributes to the development of the super-resolution OCT field. Elaborating on all the details and components of this approach more thoroughly, and linking them to specific examples of algorithmic implementations, would make this a highly valuable contribution to the field.
- 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 presents a comprehensive approach to the reconstruction of super-resolution images in OCT. The strength of this work lies in the fact that motion artifacts, aberration correction, and resolution enhancement are all addressed within unified mathematical procedures. The authors attempt to provide a comprehensive mathematical description of these procedures. I consider this part to be important, albeit highly abstract.
- 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:
- The authors claim: “This work describes a computational method for reconstructing higher resolution 3D images from repeated (non-AO) retinal OCT volumes by inverting a uniquely comprehensive imaging model.” and describe some sophisticated math in Section 2.However, there is a lack of clarity on how this actually works. My experience with numerical aberration correction using numerical wavefront reforming (e.g., with Zernike polynomial expansion) without adaptive optics demonstrates that achieving super-resolution requires numerical aberration correction with digital beam refocusing to a spot tighter than the initial probing beam. I know this is possible if one locks onto reference points on the retina, corrects the distortions for them to make them sharp structures, and then refocuses more narrowly than the initial beam width. However, the algorithm proposed by the authors is described in highly dense mathematical language, without revealing the physical principles behind such correction and refocusing processes. I would like to see a more detailed explanation in the paper connecting the math to the physical processes, clarifying how and through which operations on physical fields and wavefronts it is possible to compensate, correct, and refocus the beams to achieve super-resolution.
- I have some doubts regarding reproducibility: there is no mention in the paper whether the data and code are either already hosted somewhere or will be released upon publication so that the community can verify their functionality as described. Most of the operations remain described too abstractly for me, and it is not obvious how they are actually implemented in the algorithms. Personally, I would prefer to recommend such a complex paper for publication only if the code implementation of this algorithm, along with the provided test data, is made openly available for verification, as I do not have full confidence in the reproducibility of these results.
Minor:
- In the mathematical representations, for example in formula (3), the operations should clearly have the meaning of spatial spectrum filtering, represented explicitly as a convolution, or in the Fourier domain as operations on the spectrum. One way or another, distortions and aberrations that occur in OCT arise in k-space, and are usually corrected in this domain, either explicitly (Fourier domain) or in the spatial representation via convolutional functions. This physical meaning should be explicitly stated and described to improve the understanding of the presented algorithm.
- I would extract the initial parameters of the OCT system and describe them in a separate Methods section. The Methods section currently starts immediately with the algorithm, although it would be great to indicate at the beginning what exactly this is being applied to. Perhaps it would be useful to clearly add this in your response in the rebuttal thread.
- The following comment is also related to the description of the device and the process: the scanning pattern should be described and illustrated more clearly. It is currently not very clear how the information was acquired and accumulated. The point cloud panel in Fig. 1 does not look informative to me.
- The citations are not in the correct numerical order based on their appearance in the text (e.g., they jump from 9 to 6, then 25, etc.). Please renumber them sequentially.
- Please rate the clarity and organization of this paper
Poor
- 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?
Lack of clarity. See comments in the list the major weaknesses of the paper. Major: The mathematical description of the processes, which fundamentally deal with physical distortions, is too abstract. There is a weak connection in the description to the underlying physics and aperture effects. I do not believe the super-resolution together with aberration correction are a purely algorithmic problem where physics can be ignored, especially considering that the proposed approach replaces adaptive optics with computational operations. Furthermore, there is no mention of whether the source code will be provided to clarify how this actually works explicitly in practice. Adding clearer descriptions to the some appendix will help me better understand the method during the rebuttal and allow me to raise my score from a “weak accept” to an “accept.”
- Reviewer confidence
Very confident (4)
- [Post rebuttal] After reading the authors’ rebuttal, please state your final opinion of the paper.
Accept
- [Post rebuttal] Please justify your final decision from above.
Although the topic could be explored from a physics perspective rather than solely a computer science one, the paper is interesting and merits acceptance. My remaining questions can be discussed with the authors in person at the conference or addressed in a future extended publication.
Review #3
- Please describe the contribution of the paper
The paper presents a computational framework for reconstructing super-resolved 3D retinal OCT volumes from repeated non-adaptive optics scans by combining motion correction, distortion compensation, illumination normalization, and inverse-model-based reconstruction. Its main contribution is enabling visualization of outer segment cellular structures over wider fields without adaptive optics hardware.
- 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.This paper achieves cellular-level OCT enhancement without adaptive optics, which improves clinical feasibility. The proposed pipeline is technically comprehensive and addresses multiple acquisition confounders in a unified framework. The qualitative results are also convincing, with clear improvements in sharpness and reproducibility shown in Figures 2 and 3.
- 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.Limited methodological novelty. Several major components in this paper are built directly on the prior work, especially motion correction, illumination compensation, and distortion modeling. 2.The super-resolution stage mainly relies on a classical weighted least-squares formulation related to prior multi-frame super-resolution methods. 3.Evaluation is mostly qualitative and limited to eight healthy-subject datasets.
- 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 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
It would strengthen the paper to include quantitative metrics for evaluating the reported improvements in image quality and structural visibility.
- 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 presents meaningful results with clear practical relevance. However, the methodological novelty should be stated more explicitly, and quantitative metrics should be added to better support the reported improvements.
- Reviewer confidence
Very confident (4)
- [Post rebuttal] After reading the authors’ rebuttal, please state your final opinion of the paper.
Accept
- [Post rebuttal] Please justify your final decision from above.
The paper quality is good. Just two concerns, as I mentioned: 1.data, and 2.quantitative metrics. The authors mentioned the size limitation of not having quantitative metrics. This does not sound reasonable to me. But I still lean towards accepting.
Author Feedback
We thank the reviewers for their careful feedback and positive comments, including “significant practical advantage for clinical translation”, “visual improvements in contrast, sharpness, and separability of cone reflections are substantial”, “a convincing and well-designed evaluation” (Fig. 3, qualitative), “clear and well-structured ablation”, “The clinical relevance is clearly articulated”, “cone reflections are consistent across independent data splits”, “well motivated by the practical confounders”. (Quotes from Reviewer #1.)
Subsequently, we focus on major critique, and Reviewer #1, because he provided the worst rating (weak reject).
Qualitative Evaluation [Meta,R1,R3]
- We agree that an extended journal article should include quantitative evaluation. Here, due to the page limit, we prioritized the in our view most tangible and convincing evaluation
- Our qualitative evaluation (Fig. 3) provides an immediate and intuitive impression of the improvement, and demonstrates its consistency in independent data splits. Multiple instances of each individual improvement (gaps/doubling, cone undistortion, contrast/separability of cone reflections) can be pointed out in a single, small crop.
- Quantitative evaluation would mix these effects into a single abstract number. It is often chosen to prove subtle/non-obvious improvements, and to compare to prior methods. However, our improvement is visually “substantial”, and since this is a fundamentally novel approach, no directly comparable method exists.
- The mentioned AO-OCT ground truth only exists in a handful places on earth; we do not have access to paired data. An unpaired comparison image is referenced on p. 7 [16, Fig. 5].
Hyperparameter selection [Meta,R1] 2D-upsampling-factor (3^2=9) should scale with volume repeats (10); sigma_tv was chosen as transverse resolution (9um) / 1D-upsampling-factor 3; regularizer weight controls denoising/blur and should be small; anisotropy factor gamma: small to not blur layers; step size: safe margin below divergence. Added.
Limitation to 8 datasets [R1,R3] Core processing steps (motion, illumination, distortion compensation) have been applied to >1000 clinical high-resolution 850nm SD-OCT datasets across diverse pathologies, incl Parkinson’s disease (severe motion) and elderly patients (severe opacities/vignetting). Failure modes are opacities (low illumination) or strong mirror artifacts, if a large portion of the dataset is affected. However, such data is not usable for clinical quantification; compensation is meaningless. Only a handful of the >1000 cases failed due to motion. Added. For our application (OS cones), we wanted to maximize resolution of the initial data, motivating our transition to vis-OCT despite limited data. We expect similar processing behavior in vis-OCT with somewhat increased illumination issues due to a higher sensitivity to opacity of shorter wavelength light.
Reproducibility [R3,Meta] Our contribution is an extensive framework that unifies several prior works (motion, illumination & distortion compensation) and adds super-resolution processing. We carefully summarized key ideas of all prior work, but comprehensive reiteration is infeasible in an 8p paper (@R2: there is no appendix anymore); we have to refer to prior papers for implementation details.
Reproducibility concern on aberration correction (ACor) [R2] Digital ACor methods use large beam diameters to increase SNR & allow very fast scanning. This scales up aberration effects substantially, but temporal closeness makes them locally consistent, allowing estimation & compensation. Our novel approach: commercial beam diameter & scanning speed, SNR enhancement via averaging, digital super-resolution. Challenges: motion, illumination, distortion. We don’t describe ACor because it is less needed & we are not doing it. Our key message: We achieve substantial improvement without it.
Finally, we revised the manuscript for clarity regarding all related comments
Meta-Review
Meta-review #1
- Your recommendation
Invite for Rebuttal
- 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 authors are expected to provide reasonable explanations for the quantitative analysis and hyperparameter selection to address the reviewers’ concerns. In addition, some reviewers raised concerns regarding the reproducibility of the proposed method.
- After you have reviewed the rebuttal and updated reviews, please provide your recommendation based on all reviews and the authors’ rebuttal.
Accept
- Please justify your recommendation.
After reading the rebuttal and other reviewers’ comments, I think the authors have provided reasonable clarifications to the reviewers’ concerns.
Meta-review #2
- After you have reviewed the rebuttal and updated reviews, please provide your recommendation based on all reviews and the authors’ rebuttal.
Accept
- Please justify your recommendation.
This paper proposes a computational 3D super-resolution OCT reconstruction framework for repeated non-adaptive-optics retinal OCT volumes, aiming to reveal cellular-level outer-segment structures over wider fields without adaptive optics hardware or pupil dilation.
The reviewers recognized the practical and clinical significance of the work, especially its unified model-based treatment of motion correction, distortion compensation, illumination normalization, and inverse reconstruction. They also appreciated the visually compelling improvement in cone-reflection visibility and the reproducibility experiment using independent data splits.
The reviewers raised concerns about the limited quantitative evaluation, small healthy-subject cohort, lack of pathological cases, hyperparameter justification, and reproducibility. In the rebuttal, the authors clarified the rationale behind key hyperparameters, discussed prior experience on a larger set of clinical OCT data, and explained that the current work focuses on computational super-resolution rather than digital aberration correction. Reviewers were satisfied overall, and all post-rebuttal recommendations support acceptance.
Although quantitative metrics and broader pathological validation would strengthen an extended version, the paper presents a meaningful and practically relevant framework with convincing qualitative evidence. Therefore, I recommend acceptance.
Meta-review #3
- After you have reviewed the rebuttal and updated reviews, please provide your recommendation based on all reviews and the authors’ rebuttal.
Accept
- Please justify your recommendation.
The paper presents a computational framework for reconstructing super-resolved 3D retinal OCT volumes from repeated non-AO scans, aiming to reveal cellular-level outer segment structures over wider fields without adaptive optics. Reviewers agreed that the problem is important and that the practical clinical motivation is strong, especially given the potential to broaden access to microstructural retinal imaging without additional hardware. The unified treatment of motion correction, distortion compensation, illumination normalization, and inverse-model-based reconstruction was also viewed positively. Before rebuttal, the main concerns were the lack of quantitative evaluation, limited validation on only eight healthy subjects from a single site, and insufficient clarity on some methodological and physical aspects of the reconstruction process. The rebuttal addressed these points partially but meaningfully. In particular, the authors clarified the intended scope of the work, explained why paired AO-OCT ground truth and direct quantitative comparisons are not realistically available, provided additional justification for hyperparameter choices, and clarified that core preprocessing steps have already been applied more broadly in their larger clinical experience, even though the present study focuses on a limited vis-OCT dataset for the outer-segment cone application. They also clarified that the method does not rely on digital aberration correction, which helps sharpen the framing of the contribution. The paper still has limitations. Quantitative evaluation remains limited, pathology cases are not included in the current study, and reproducibility would be stronger with public code or more implementation detail. Some of the claims would also benefit from a clearer distinction between what is already validated in the manuscript and what is supported mainly by broader prior experience.
