Abstract

Tissue tracking in echocardiography is challenging due to the complex cardiac motion and the inherent nature of ultrasound acquisitions. Although optical flow methods are considered state-of-the-art (SOTA), they struggle with long-range tracking, noise occlusions, and drift throughout the cardiac cycle. Recently, novel learning-based point tracking techniques have been introduced to tackle some of these issues. In this paper, we build upon these techniques and introduce EchoTracker, a two-fold coarse-to-fine model that facilitates the tracking of queried points on a tissue surface across ultrasound image sequences. The architecture contains a preliminary coarse initialization of the trajectories, followed by reinforcement iterations based on fine-grained appearance changes. It is efficient, light, and can run on mid-range GPUs. Experiments demonstrate that the model outperforms SOTA methods, with an average position accuracy of 67% and a median trajectory error of 2.86 pixels. Furthermore, we show a relative improvement of 25% when using our model to calculate the global longitudinal strain (GLS) in a clinical test-retest dataset compared to other methods. This implies that learning-based point tracking can potentially improve performance and yield a higher diagnostic and prognostic value for clinical measurements than current techniques. Our source code is available at: https://github.com//.

Links to Paper and Supplementary Materials

Main Paper (Open Access Version): https://papers.miccai.org/miccai-2024/paper/1980_paper.pdf

SharedIt Link: pending

SpringerLink (DOI): pending

Supplementary Material: https://papers.miccai.org/miccai-2024/supp/1980_supp.zip

Link to the Code Repository

https://github.com/riponazad/echotracker/

Link to the Dataset(s)

N/A

BibTex

@InProceedings{Aza_EchoTracker_MICCAI2024,
        author = { Azad, Md Abulkalam and Chernyshov, Artem and Nyberg, John and Tveten, Ingrid and Lovstakken, Lasse and Dalen, Håvard and Grenne, Bjørnar and Østvik, Andreas},
        title = { { EchoTracker: Advancing Myocardial Point Tracking in Echocardiography } },
        booktitle = {proceedings of Medical Image Computing and Computer Assisted Intervention -- MICCAI 2024},
        year = {2024},
        publisher = {Springer Nature Switzerland},
        volume = {LNCS 15004},
        month = {October},
        page = {pending}
}


Reviews

Review #1

  • Please describe the contribution of the paper

    It presents a coarse-to-fine model that initializes trajectories coarsely, then refines them iteratively based on fine-grained appearance changes. This shows a significant improvement in clinical metrics like global longitudinal strain.

  • Please list the main strengths of the paper; you should write about 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.
    • introduces a novel two-fold coarse-to-fine model for point tracking in echocardiography, which is tailored to manage the complexities of cardiac motion and ultrasound image artifacts.

    • outperforms SOTA and offers an efficient computation that can run on mid-range GPUs, making it accessible for broader clinical use

    • related work section is succinct yet appropriate and provides an good review of relevant work.

  • Please list the main weaknesses of the paper. Please provide details, for instance, if you think a method is not novel, explain why and provide a reference to prior work.
    • The study depends exclusively on data from a single brand of ultrasound scanner, which could limit the applicability of the results across different ultrasound technologies.

    -he two-stage coarse-to-fine model is complex, raising questions about the necessity of such an approach when a simpler, integrated end-to-end model might achieve similar results with potentially greater ease of use and implementation.

    • The validation primarily uses internally generated data without external datasets. Also missing an ethics IRB/REB #.

    • The model’s performance appears heavily reliant on high-quality ultrasound data. There is scant discussion on how the model would perform with varied quality

    • Statistical, and clinical significance as well as interpretability remain in question

  • 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 authors claimed to release the source code and/or dataset upon acceptance of the submission.

  • Do you have any additional comments regarding the paper’s reproducibility?

    Adequately reproducible, pending the github link which they have appropriately anonymized.

  • Please provide detailed and constructive comments for the authors. Please also refer to our Reviewer’s guide on what makes a good review. Pay specific attention to the different assessment criteria for the different paper categories (MIC, CAI, Clinical Translation of Methodology, Health Equity): https://conferences.miccai.org/2024/en/REVIEWER-GUIDELINES.html

    Please refer to the strengths and weaknesses above.

    The paper presents an innovative approach to myocardial point tracking in echocardiography with a two-fold coarse-to-fine model, demonstrating significant improvements in global longitudinal strain measurements. However, to enhance the robustness and clinical applicability of the research, I would recommend addressing the following points in future revisions:

    • It would strengthen the findings if the model were tested across different brands and types of ultrasound scanners. This could help validate the model’s robustness and applicability in diverse clinical settings.

    • Further justification on the current approach’s two-stage complexity. I leave the paper still thinking, “why do we need an iterative refinement stage in this modern era?”

    • Discuss or test the model’s performance using ultrasound data of varying quality and from different clinical environments. This could involve simulations or real-world testing where data quality is intentionally varied, providing insights into the model’s reliability under less-than-ideal conditions.

    • Broaden the discussion on limitations to include potential biases arising from data diversity and model training scenarios. This could involve a deeper analysis of how different patient demographics or pathological conditions might affect the model’s efficacy.

  • 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

    Weak Reject — could be rejected, dependent on rebuttal (3)

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

    Model complexity, limited validation, reliance on high-quality, homogeneous data.

  • Reviewer confidence

    Confident but not absolutely certain (3)

  • [Post rebuttal] After reading the author’s rebuttal, state your overall opinion of the paper if it has been changed

    N/A

  • [Post rebuttal] Please justify your decision

    N/A



Review #2

  • Please describe the contribution of the paper

    The authors develop and validate an echocardiography myocardial point tracking system inspired by recent neural network-based advances in video tracking. Their synthesized architecture involving coarse to fine point trajectory refinement provides significant improvement over prior SOTA methods fine-tuned to the same private datasets.

  • Please list the main strengths of the paper; you should write about 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.

    -Extensive background and justification -Clear overview of the architecture and logistics of training. -Excellent results extending to clinical index (GLS), and provided codebase.

  • Please list the main weaknesses of the paper. Please provide details, for instance, if you think a method is not novel, explain why and provide a reference to prior work.

    -Some significant details in model training are left to citation [20]. -Multiple views are noted in the datasets but not separated in performance results.

  • Please rate the clarity and organization of this paper

    Excellent

  • 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.

  • Do you have any additional comments regarding the paper’s reproducibility?

    source code, but no discussion of dataset availability.

  • Please provide detailed and constructive comments for the authors. Please also refer to our Reviewer’s guide on what makes a good review. Pay specific attention to the different assessment criteria for the different paper categories (MIC, CAI, Clinical Translation of Methodology, Health Equity): https://conferences.miccai.org/2024/en/REVIEWER-GUIDELINES.html

    -the supplemental videos show excellent results, but would there be a way to visualize median and low results in the static paper. Similarly a description of limitations and failure cases is missing.

  • 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

    Accept — should be accepted, independent of rebuttal (5)

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

    Novel method synthesized for echocardiography providing SOTA results in validation compared against other fine-tuned prior works, with codebase.

  • Reviewer confidence

    Confident but not absolutely certain (3)

  • [Post rebuttal] After reading the author’s rebuttal, state your overall opinion of the paper if it has been changed

    N/A

  • [Post rebuttal] Please justify your decision

    N/A



Review #3

  • Please describe the contribution of the paper

    This paper provides a novel approach to myocardial speckle tracking, by tailoring recent computer vision advancements to ultrasound myocardial strain assessment of left ventricular wall function. The authors intended to improve on the limitations of currently available optical flow architectures which have problems with long-term temporal context, poor spatial resolution and inefficiencies. They proposed a 2-fold, coarse-to-fine model approach. As a result, they trained a 2-stage model in which the initial stage involves the initialization of trajectories feature maps using a coarse network and in the second stage the trajectories are iteratively refined using a fine network. The architecture thus constitutes a 2-fold coarse-to-fine approach while at the same time sharing the first frame projection coordinates and flow information giving the combined model a better spatial performance and long-term temporal context.

  • Please list the main strengths of the paper; you should write about 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.

    One of the main strengths of the paper is the innovative ways that the authors have applied EchoTracker to overcome the limitations of recent architectures. Cardiac ultrasound imaging involves visualization and measurements of rapidly changing and deformable tissues/ myocardium and having a model like EchoTracker, with long-term temporal awareness, and potentially superior spatial resolution, is necessary to obtain a reliable measurement that will be clinically significant.

    Another main strength of the paper is the low computational requirement for EchoTracker. Its 2-fold coarse-to-fine architecture, pruning and low number of training iterations makes it suitable for deployment in low-income areas where heavy computation power is often a limitation.

    A third strength of the paper is the relevance of the evaluation metrics. Positional accuracy and peak global longitudinal strain are appropriate metrics relevant to cardiac strain studies.

  • Please list the main weaknesses of the paper. Please provide details, for instance, if you think a method is not novel, explain why and provide a reference to prior work.

    One of the weaknesses of the paper is that it is unclear if the result obtained is statistically significant or obtained by chance without a comparative statistical test that reports the p-value of a test comparing EchoTracker to other architectures. Samples may have different measurements and without a statistical test, there is a possibility that the difference that is observed is a coincidence of random sampling (Parab and Bhalerao 2010).

    Little is reported on the distribution or characteristics of the reference data or the expertise/years of experience of operators despite Echo Tracker reportedly outperforming other SOTA models (PIPS++ and CoTracker) on the reference dataset. This data information represents the ground truth upon which the entire evaluation was based. Therefore, there is a need to demonstrate the robustness and diversity of the data, and the cases included (i.e. healthy adults, pathological cases or Pediatric cases). These variations are clinically relevant in cardiac ultrasound imaging interpretation. The comparison reported in this study, though better, may be biased to the nature of the data and how it was collected. In a study by Nyberg et al 2023, they reported that GLS measurements are affected by various factors including age groups, BMI, Mean Arterial Pressure (MAP) and heart rate.

    Clinical evaluation is another shortcoming of the paper. Data obtained by EchoTracker was compared to other models using different private data sets. Although the authors mentioned that a direct comparison could not be made, they went ahead to conclude that Echotracker may enhance GLS measurements in a test-retest approach. Clinical viability, validity and feasibility can only be ascertained by knowing the model’s performance across different clinical scenarios and cases and comparing models on the same data. Even though there are differences in measurements using different architectures including EchoTracker, it is difficult to demonstrate the clinical relevance of the differences without knowing a clinically significant margin to base such assumptions. The observation of differences in measurement does not automatically suggest clinical relevance. While clinical significance demonstrates dissimilarity between groups of measurements, statistical significance implies whether there is any mathematical significance to the analysis of the results or not (Ranganathan et al 2015).

  • Please rate the clarity and organization of this paper

    Very 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.

  • Do you have any additional comments regarding the paper’s reproducibility?

    The paper is reproducible to a large extent. The authors provided an anonymized link to their code. While the code may suffice to reproduce the architecture, little is known about the characteristics of the data and the operators.

  • Please provide detailed and constructive comments for the authors. Please also refer to our Reviewer’s guide on what makes a good review. Pay specific attention to the different assessment criteria for the different paper categories (MIC, CAI, Clinical Translation of Methodology, Health Equity): https://conferences.miccai.org/2024/en/REVIEWER-GUIDELINES.html

    This is a unique submission that aims to tackle deficiencies in the currently available point-tracking solutions for cardiac ultrasound imaging. It has great potential to achieve superior left ventricular cardiac function diagnosis and prognosis. This proposal can provide more efficient global ventricular health compared to the traditional approach of using ejection fraction.

    However, there should be more information on the data used in the evaluation of this architecture and the operators. A statistical test like the student-T test may be conducted between Echotracker and other architectures. This does suggest looking at clinical significance once a statistical difference is proven: https://www.ema.europa.eu/en/documents/scientific-guideline/points-consider-switching-between-superiority-and-non-inferiority_en.pdf.

  • 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

    Accept — should be accepted, independent of rebuttal (5)

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

    The paper identified a real problem in the currently available improvements to point-tracking systems in cardiac ultrasound imaging. The author’s approach to the problem demonstrated an understanding of the difficulties with these solutions. The proposed architecture, EchoTracker, explored a unique and innovative implementation to solve these limitations most efficiently in terms of better resolution, speed and lower computation requirements.

    Even though the data distribution and operators’ information including clinical relevance is not clear in this paper, it presents an interesting and unique approach that may benefit medical imaging and clinical communities. It may be explored to develop and improve patient diagnosis.

  • Reviewer confidence

    Somewhat confident (2)

  • [Post rebuttal] After reading the author’s rebuttal, state your overall opinion of the paper if it has been changed

    N/A

  • [Post rebuttal] Please justify your decision

    N/A




Author Feedback

N/A




Meta-Review

Meta-review not available, early accepted paper.



back to top