List of Papers Browse by Subject Areas Author List
Abstract
Bronchoscopy is a minimally invasive procedure for diagnosing and treating lung conditions, but navigation remains challenging due to the reliance on preoperative imaging, sensor-based tracking, and the feature-poor visual environment of the airways. To address these limitations, we propose a novel Navigated Bronchoscopy (NB) framework that enables real-time guidance and repeatable interventions without requiring external sensors or CT scans, which makes it particularly suitable for mechanically ventilated patients in critical care units with limited access to preoperative images. We base our approach on deep learning, specifically by performing airway landmark recognition using deep visual features and using a Vision Transformer (ViT)-based pose regression network to track bronchoscope motion. Our framework is deployed on a commercially available bronchoscope and validated through bronchoscopy trials in a phantom lung model and a mechanically ventilated ex-vivo human lung, demonstrating its feasibility for real-world applications.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2025/paper/3921_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)
https://doi.org/10.1142/S2424905X24400099
BibTex
@InProceedings{MacEmi_Navigational_MICCAI2025,
author = { Mackute, Emile and Zhang, Francis Xiatian and Dhaliwal, Kevin and Khadem, Mohsen},
title = { { Navigational Bronchoscopy in Critical Care via End-to-End Pose Regression } },
booktitle = {proceedings of Medical Image Computing and Computer Assisted Intervention -- MICCAI 2025},
year = {2025},
publisher = {Springer Nature Switzerland},
volume = {LNCS 15968},
month = {September},
page = {405 -- 415}
}
Reviews
Review #1
- Please describe the contribution of the paper
The main contribution is a novel NB framework for mechanically ventilated patients in critical care that operates without the need for preoperative imaging.
- 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 clearly describes the target of the navigational bronchoscopy, which aims at bronchoscopy in critical care. Based on the clinical requirements, the framework is divided into two phases including exploratory phase and navigation phase. Moreover, several bronchoscopy trials are conducted to evaluate the vision-based navigation framework.
- 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.
- Important details of the proposed method are missing, leading to confusion in understanding the navigational pipeline. 1) Since the global map of the lung (CT) is absent, why is pose estimation necessary even with landmark recognition? Will 3D map reconstruction be helpful during navigation? 2) Given that the camera is manually controlled, what is the importance of pose estimation? 3) In section 2.1, how did the author train the pose estimation model? 4) In section 2.2, the author did not describe the pose refinement process. The paper states, ‘This similarity metric ensures robust landmark matching, facilitating loop closure and drift correction in pose estimation.’ How is this similarity used to correct pose estimation? 5) In Fig. 2, the description indicates two phases and two models, which makes it confusing when the author states that the pipeline consists of four stages. Additionally, stage 4 is missing from the figure.
- The speed of the navigational framework is critical, yet the author does not provide any insights from the experiments.
- In Figs. 3 and 4, why is the repeated path outside the lung? It is suggested to provide a detailed description of the bronchoscopy trail to enhance understanding of the results. Additionally, Fig. 3 is not cited in the main content of the paper.
- Figures 3 and 4 illustrate the lung’s complex structure with many airway branches. During real surgery, in the exploratory phase, the surgeon must manipulate the bronchoscope to reach three targets (A, B, C). In the navigation phase, the surgeon still needs to steer the scope to these targets. How long does the entire process take? Why doesn’t the surgeon administer the drug during the exploratory phase?
- 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.
- 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
No
- 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 — could be rejected, dependent on rebuttal
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
Important details of the proposed method are missing, leading to confusion in understanding the navigational pipeline. The speed of the navigational framework is critical, yet the author does not provide any insights from the experiments.
- Reviewer confidence
Confident but not absolutely certain (3)
- [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 has addressed all issues.
Review #2
- Please describe the contribution of the paper
This work presents a bronchoscope navigation framework, which enables repeatable interventions without requiring external sensors or CT scans.
- 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 work presents a pre-op CT free bronchoscope navigation method targeting at critical care. It addresses the commonly known lung motion issues to some extend. It enables repeatable access to target regions under the guidance of landmark recognition. 2) This work provides phantom experiments for quantitative evaluation of pose estimation accuracy and feasibility analysis for the system. It also provides cadaver study to further assess the feasibility of the framework.
- 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 ‘navigation’ components is not clearly described. The authors mentioned the navigation is consist of pose prediction, landmark recognition and the navigation score. However, it is confusing how each one of them contribute to the operator’s judgement during procedure, e.g. without a CT scan, how does relative camera pose prediction provide guidance to the target. Also, the feasibility of the framework is not supported with qualitative experiments. 2) The predicted poses are recorded for landmarks during exploratory phase, but it’s unclear how are the poses still useful when the anatomy deforms. 3) For the navigation score, the possible failure cases are not discussed. 4) The pose regression comparison experiment lacks information such as length of evaluation sequence (fps, seconds), the settings to the SOTA methods (are they used out-of-box, or fine-tuned). 5) Identifying the hyper-parameters with phantom model and apply them to the ex-vivo lung trials is questionable. 6) The authors mention ‘real-time’ property of the proposed framework several time, but it is not supported with run-time analysis. 7) Some writing issues still exist, e.g. Fig.2. caption, pose regressions network is shown in the bottom right; in experiments section, Fig. 4 is not referred. Please double check and refine the manuscript.
- 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 provide sufficient information for reproducibility.
- 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?
The contributions are solid, however, further clarification and more complete experiments are expected to prove the framework feasibility.
- 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.
Issues have been addressed.
Review #3
- Please describe the contribution of the paper
The paper presents a novel Navigational Bronchoscopy (NB) framework tailored for mechanically ventilated patients in critical care, addressing the challenge of accurate navigation without reliance on preoperative imaging or external sensors. The proposed method leverages a Vision Transformer (ViT)-based pose regression network combined with deep visual feature extraction and landmark recognition to enable real-time bronchoscope tracking and repeatable interventions.
- 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.
Novelty: The paper introduces a pioneering NB framework that eliminates the need for preoperative CT scans, making it highly relevant for critical care settings without extra sensors.
Method: The methodology is well-described, with detailed explanations of the ViT-based pose regression, landmark recognition via Bag-of-Words (BoW), and pose refinement through loop closure.
Experiments: The comparison in Table 1 shows that the proposed method outperforms existing approaches in pose estimation accuracy (RPE and ATE metrics), with a notable reduction in error.
- 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.
Computational Efficiency: It would be better to discuss the framework’s runtime performance and explore model optimization techniques (e.g., pruning, quantization) to ensure feasibility for real-time clinical use.
- 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.
- 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
Good paper, and would like to know if the data and model can be open source.
- 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?
This paper utilize a ViT for pose tracking and navigation, the idea is novel and the experiments is sufficient. The discussion for computational efficiency and reproducibility could be refined.
- 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.
Despite there exist few issue regarding to the clarity and computational cost, the author claims that they will add details discussion. Therefore I retain my opinion towards acceptance.
Author Feedback
We thank all reviewers for their thoughtful feedback and the time invested in evaluating our submission amid a high volume of papers. We are encouraged by the recognition of our contributions, including the CT-free, sensorless navigational bronchoscopy framework for critical care (R1, R3, R4), its strong experimental validation on human models (R3), and clinical relevance (R1). Below, we address the main concerns raised. [1] Purpose and Value of Pose Estimation Without CT or External Sensors [R1, R3] Our method enables relative navigation via frame-to-frame 6D pose regression (outlined in Section 2.1) to return to previously visited regions based on internal visual memory. Although the bronchoscope is manually controlled, pose estimation provides real-time spatial feedback that helps the operator re-align with prior targets, especially in low-texture or repetitive regions. As the ViT-based model continuously tracks bronchoscope motion. Without pose estimation, landmark recognition alone lacks directional context, resulting in unreliable guidance. Pose prediction transforms recognition into actionable navigation, enabling repeatable targeting—critical for interventions such as BAL or localized drug delivery in ICU settings, where CT and external tracking are unavailable. While 3D mapping may aid navigation (R1), most bronchoscopy depth methods rely on CT-based supervision, which generalizes poorly in ICU contexts due to anatomical distortion and dynamic ventilation. Our self-contained approach avoids such static assumptions and is more robust for in situ use. We will highlight these points in revision. [2] Refinement via Similarity Scores [R1] As described in Section 2.2 (p.5) and Eq. 4, deep features from current frames are matched to stored landmark vectors. When similarity exceeds a threshold, a pose correction is computed and propagated backward using interpolation [17] to reduce drift. This loop closure enhances pose consistency without requiring a “global CT-based map”. We will clarify this process in revision. [3] Runtime and System Feasibility [R1, R3, R4] Our inference pipeline achieves ~20Hz on an RTX 3060 (Section 2.1, p.4), sufficient for real-time clinical guidance. Though not deployment-optimized, it supports current workflows. We will include runtime profiling and note future optimization strategies, as suggested by R4. [4] Model Training and Evaluation [R1, R3] As noted in Section 2.1, our ViT model was trained on frame pairs with SE(3) labels from electromagnetic tracking (EMT) in a public dataset [12]; EMT was only used during training as it is not available in critical care setting. Baselines were evaluated on a 10% validation split using default settings. We will clarify trial duration and dataset splits. [5] Anatomical Deformation and Repeatability [R3] Our method assumes intra-procedure anatomical stability under ventilation, as stated in Section 3 (p.7). Landmarks remain visually consistent in this setting. We agree this limits inter-session use and will emphasize this in the revision as future work. [6] Experimental Setup, Figures, and Interpretation [R1, R3] As described in Section 3 (pp.6–8), trials lasted ~12–15 minutes, reflecting the time needed for initial exploration, landmark acquisition, and repeat navigation. Drug delivery was deferred (R1) to isolate navigation performance. Apparent off-lung paths in Figs. 3–4 result from accumulated odometry error during pose integration and are visualization artifacts—not navigation failures. Our system corrects for such drift through landmark-based pose refinement, which provides robust re-alignment and avoids misleading trajectory interpretations. We thank R1 and R3 for noting the mislabelled Fig. 4 references on p.7; these will be corrected. In summary, our work presents the first deployable solution for CT-free bronchoscopy. The identified revisions will help clarify its contributions, acknowledge limitations, and guide future research directions.
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.
Accept
- Please justify your recommendation. You may optionally write justifications for ‘accepts’, but are expected to write a justification for ‘rejects’
N/A