Paper Info Reviews Author Feedback Meta-Review Back to top
List of Papers Browse by Subject Areas Author List

Abstract

Cardiovascular Optical Coherence Tomography (OCT) is hindered by the brief imaging window provided by contrast agents, making it challenging to capture high-resolution images of multiple plaques over long vessel sections. Rapid catheter pullback and coarse spatial resolution increase the likelihood of missing subtle pathologies and critical plaque microstructures, compromising diagnostic accuracy. To address this, we introduce CardioInterp, the first generative interpolation model for cardiovascular OCT, designed to synthesize high-fidelity intermediate B-slices, enhancing structural continuity and spatial resolution. Our architecture integrates a latent diffusion framework with a novel Dual-Path Fusion Decoder designed to ensure inter-slice structural continuity while preserving microanatomical fidelity. Experiments on cardiovascular OCT datasets demonstrate that CardioInterp achieves superior interpolation quality (PSNR=28.59, SSIM=51.80%) at 6 times upscaling of B-slices and spatial resolution, surpassing traditional medical image interpolation methods and setting a new benchmark. This innovative computational approach enables high-resolution imaging of long vessel sections within a limited temporal window in cardiovascular OCT. The code is available at: https://github.com/Lee728243228/CardioInterp.

Links to Paper and Supplementary Materials

Main Paper (Open Access Version): https://papers.miccai.org/miccai-2025/paper/2405_paper.pdf

SharedIt Link: Not yet available

SpringerLink (DOI): Not yet available

Supplementary Material: Not Submitted

Link to the Code Repository

https://github.com/Lee728243228/CardioInterp

Link to the Dataset(s)

N/A

BibTex

@InProceedings{LiLin_CardioInterp_MICCAI2025,
        author = { Li, Linyuan and Yang, Bing and Zhang, Minqing and He, Mengxian and Yuan, Wu},
        title = { { CardioInterp: Generative Modeling for Cardiovascular OCT Interpolation with Anatomical Continuity and Fidelity } },
        booktitle = {proceedings of Medical Image Computing and Computer Assisted Intervention -- MICCAI 2025},
        year = {2025},
        publisher = {Springer Nature Switzerland},
        volume = {LNCS 15972},
        month = {September},
        page = {74 -- 83}
}

Reviews

Review #1

  • Please describe the contribution of the paper

    The paper proposed a generative model for intravascular OCT to synthesize intermediate B-slices, with the aim of enhancing structural continuity and spatial resolution. The topic of the study is of clinical importance yet the results of the study demonstrated low clinical feasibility.

  • 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.

    This study is a novel application of generative diffusion model on intravascular OCT. Some special designed modules were introduced to address specific challenges in OCT, such as feature flow to estimate deformation across slices, incoporating TSM to increase efficiency.

  • 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 model was trained using images acquired by a specific OCT system (Vivolight). The generability of the model in other common commercial OCT systems like Abbott, Volcano remains unknown.
    2. Only limited figures with very normal vessels are shown in the paper. It did not include representative cases, with diseased atherosclerotic plaques, i.e., calcium. lipid,etc, where the imaging features are complicated and the accurate reconstruction of such frames are critical.
    3. The paper claimed to generate images with high-fidelity and structural continuity. However, there is a lack of quantitative evaluation to support them.
    4. The clinical feasibility is very low, given that the current implementation is limited to a reconstruction scale of 256 × 256 pixels, which is below the standard OCT resolution.
    5. Voxel patch volume randomly selected from sub-volumn with a resolution of 7256256 is suboptimal because it cannot make sure critical positions near the lumen/disease were sampled.
    6. The paper did not demonstrate the clinical value of the study
  • 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 has provided an anonymized link to the source code, dataset, or any other dependencies.

  • 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/2025/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.

    (2) Reject — should be rejected, independent of rebuttal

  • Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
    1. The clinical value and feasibility is low.
    2. The results did not demonstrate strong performance of the model, and there is a lack of more comprehensive evaluation and typical cases.
    3. The generability of the model remains unknown.
  • 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

    Rapid IV‑OCT pull‑back leaves 100–200 µm slice gaps that obscure plaque micro‑structures; the paper generatively inserts anatomically coherent intermediate B‑scans to restore volumetric continuity without re‑acquisition. CardioInterp compresses the first and last OCT slices into a latent space, sprinkles noise between them, and then runs a conditional diffusion process that “denoises” those latents into the missing slices. During decoding it follows two paths in parallel: Deep‑feature path — warped according to feature‑level correspondence so large‑scale vessel anatomy lines up; Shallow‑feature path — linearly blended to keep fine textures. A lightweight Temporal‑Shift block slips a few feature channels forward/backward in time, giving pseudo‑3‑D context without heavy 3‑D convolutions. While diffusion‑based frame interpolation is not new, this is the first application to cardiovascular OCT and the dual‑path fusion cleanly tackles the anisotropy issue. Innovation is incremental but practically valuable.

  • 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. First diffusion‑based framework tailored to bridge the 100–200 µm slice gaps in intravascular OCT, directly benefiting plaque assessment.
    2. Latent conditional DDPM + Dual‑Path Fusion decoder achieves temporal coherence and texture fidelity without heavy 3‑D convolutions; ablations show decisive gains.
    3. Solid empirical evidence
  • 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. Incremental originality
    2. No task‑level or radiomics metrics, no comparison to diffusion‑based VFI peers, and runtime/memory benchmarks are omitted
  • 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.

  • 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/2025/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 — could be accepted, dependent on rebuttal

  • Please justify your recommendation. What were the major factors that led you to your overall score for this paper?

    1  Clear problem framing & paper structure 2  Technically solid, well‑engineered framework 3  Reproducibility commitment 4  Strong empirical performance

  • Reviewer confidence

    Somewhat confident (2)

  • [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 author explained the questions I mentioned and others. But I do think the author should clarify the points in the camera-ready version to avoid misinformation.



Review #3

  • Please describe the contribution of the paper

    The authors introduce a novel approach for Cardiovascular Optical Coherence Tomography (OCT) interpolation that emphasizes anatomical continuity and fidelity. The method utilizes a diffusion-based model and a specially designed decoder to ensure inter-slice structural continuity and fidelity. Experimental results demonstrate the method’s effectiveness.

  • 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 proposed work is the first generative method focusing on cadiovascular OCT interpolation and the methodology design is technically sound and easy to follow.
    • The proposed method achieves good performance.
  • 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 approach to ensuring continuity, a critical aspect of the task, appears to lack significant innovation. The reliance on the decoder with Temporal Shift Module (TSM) for this purpose seems marginal and may not fully address the continuity challenge.

  • 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 has provided an anonymized link to the source code, dataset, or any other dependencies.

  • 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/2025/en/REVIEWER-GUIDELINES.html

    • The data used in this experiment consists of a limited number of slices, specifically 20 for the sub-volume and 7 for the patch volume. I am curious whether this dimensionality is typically small for this task. If the number of slices increases significantly, how might the method be adapted to accommodate this change? Would sampling methods, similar to those used in video generation and medical volume generation, be necessary to connect longer frames?

    • The term “diffusion-based priors” only happens in abstract and is not explained in main text. Also, I conjecture that the word “prior” is wrongly used.

  • 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 — could be accepted, dependent on rebuttal

  • Please justify your recommendation. What were the major factors that led you to your overall score for this paper?

    The methodology is technically sound and well-illustrated.

  • 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 all reviewers for their insightful comments on our paper.

Q1. Metrics for continuity evaluation@R1+R2+R3 Classic models like I3NET use SSIM/PSNR for reconstruction evaluation. Despite limited radiomic-specific metrics, we introduce Local Continuity Index (LCI), a mean SSIM between consecutive slices (0-1 scale). Our model achieves +30.89% (×2), +2.3% (×4), +1.06% (×6) LCI gains over SOTA. The Dual-fusion Strategy enhances continuity: flow estimation maintains deep semantic consistency, and linear interpolation preserves shallow texture continuity. And TSM is cost-free, with no additional computation. Ablation studies demonstrate TSM alone achieves 3.21% LCI improvement, and Dual-fusion strategy improves by 5.73%.

Q2. Data sampling & Pathological Performance@R2+R3 Our sampling strategy balances coverage and robustness: 1. Random sampling ensures comprehensive coverage of all tissue regions across training epochs. 2. Random interval of 1-3 slices makes the model adapt to sparser OCT reconstruction scenarios, enhancing its robustness for modeling more global deformations. 3. The 7-frame sampling is aligned with ×6 upsampling conventions of medical interpolation. If increasing the number of sampled frames, as cardiovascular OCT sequences are sparse with inter-frame deformations(200 um), it’s a challenge to directly apply natural video sampling strategies (e.g., sampling 16 frames for interpolation) due to the difficulty in modeling continuity. Integrating cardiovascular structural priors can aid in the modeling of continuity under such circumstance. To further enhance reconstruction quality and training stability, speckle noise in OCT images should be removed. In experiments on 20 pathological and 20 normal cases, our model achieves slightly lower but comparable performance on pathological regions (PSNR: 28.43, SSIM: 53.89%) vs. normal regions (PSNR: 28.87, SSIM: 54.06%), demonstrating strong generalization to complex anatomical structures.

Q3. Originality@R1 We justify the novelty from three perspectives: 1. Task originality: This is the first model dedicated to cardiovascular OCT interpolation; 2. Technical originality: The generative modeling framework outperforms traditional methods in continuity and details for long-range OCT reconstruction; 3. Architectural originality: Novel Dual-path decoder enhance OCT detail and continuity.

Q4. Comparison with diffusion-based VFI@R1 Ablations show significant gains over LVDM (diffusion-based VFI) baseline (Table 2). Other diffusion-based VFI methods rely on motion priors or image diffusion priors, which are inapplicable to OCT’s tissue deformation-based inter-frame variations.

Q5. OCT systems@R3 We appreciate the reviewer for pointing out this issue. Our images acquired by two OCT systems: 645 from Vivolight and 85 from Abbott . Q6. Reconstruction scale@R3 Generated 256 * 256 patches can be pixel-wise stitched into complete 1025×497-pixel slices, fully meeting clinical requirements.

Q7. Runtime/memory benchmarks@R1 Currently, this study focuses on validating the algorithm’s effectiveness, and runtime/memory optimization is not a core objective. These aspects are planned for future work.

Q8. Explanation on word “prior”@R2 We admit “prior” was unclear. To clarify, our intended meaning is that the model leverages the powerful generative capabilities of diffusion models.

Q9. Demonstration of clinical value@R3 Clinical significance can be demonstrated: 1.Our model achieves SOTA results in PSNR/SSIM (Table 1) + LCI metrics, generating sequences with continuity and fidelity, confirming the transitions critical for radiomic analysis. 2.Our method reconstructs anatomical details in diseased areas, with a negligible performance gap (PSNR: 28.43 vs. 28.87; SSIM: 53.89% vs. 54.06%) compared to normal tissues. 3.Generated patches can be seamlessly stitched into full-field OCT images at the pixel level, meeting resolution requirements of clinical diagnosis.



Meta-Review

Meta-review #1

  • Your recommendation

    Invite for Rebuttal

  • If your recommendation is “Provisional Reject”, then summarize the factors that went into this decision. In case you deviate from the reviewers’ recommendations, explain in detail the reasons why. You do not need to provide a justification for a recommendation of “Provisional Accept” or “Invite for Rebuttal”.

    N/A

  • 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. You may optionally write justifications for ‘accepts’, but are expected to write a justification for ‘rejects’

    N/A



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. You may optionally write justifications for ‘accepts’, but are expected to write a justification for ‘rejects’

    N/A



Meta-review #3

  • After you have reviewed the rebuttal and updated reviews, please provide your recommendation based on all reviews and the authors’ rebuttal.

    Reject

  • Please justify your recommendation. You may optionally write justifications for ‘accepts’, but are expected to write a justification for ‘rejects’

    The paper presents a method for cardiovascular OCT interpolation based on diffusion models with binary masks of the first and the last frame as conditions. Deep feature-flow warping and shallow feature interpolation were used to enhance the fidelity and continuity. The main concerns about this work includes methodology novelty, experimental evaluation (in terms of continuity) and comparison, generalization, clinical value and feasibility. While the authors add new results for continuity evaluation and explanations about the clinical value, The diffusion model-based interpolation with a two-path fusion strategy and TSM is of limited novelty. Besides, Inconsistent description about the data source and lack of detailed splitting reduce the credibility of the experiment results.



back to top