Geometry-Aware Visual Odometry for Bronchoscopic Navigation via High-Gain Observer Fusion

Farshid Alambeigi; Feng Li; Francis Xiatian Zhang; Kevin Dhaliwal; Mohammadreza Kasaei; Mohsen Khadem

REVIEW 4 major objections 5 minor 37 references

Geometry-Aware Visual Odometry for Bronchoscopic Navigation via High-Gain Observer Fusion

T0 review · 4 major / 5 minor · reviewed 2026-07-11 · grok-4.5

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Vanishing-point cues from airway lumens, fused by a high-gain observer, cut bronchoscope trajectory error by more than half without CT or external sensors.

desk verdict Real EMT-validated gain on ventilated human lungs for vision-only bronchoscopy VO; the fusion idea is useful, but the table does not isolate what actually bought the 50% ATE drop. read the letter →

arxiv 2607.05162 v1 pith:ZMYBNGP5 submitted 2026-07-06 cs.RO

Mohammadreza Kasaei , Francis Xiatian Zhang , Feng Li , Farshid Alambeigi , Kevin Dhaliwal , Mohsen Khadem This is my paper

classification cs.RO

keywords visualodometrybronchoscopyvanishingpointhigh-gainobserverairwaynavigationmonocularSLAMendoscopy

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Bronchoscope navigation today usually needs preoperative CT or external trackers, which limits use in intensive care and low-resource settings. Pure vision fails because airway images are nearly featureless, full of specular glare, and geometrically singular: the camera points down a tube so ordinary parallax disappears. This paper shows that the dark openings of distal lumens act as reliable vanishing-point portals. Detecting those lumens, back-projecting them to 3D rays, and fusing the rays yields a stable forward heading even when classical visual odometry has nothing to track. That heading, together with a simple looming-based speed estimate, is fed into a high-gain observer that also sees the noisy raw VO output and enforces the anatomical prior that the scope must stay inside the airway. On mechanically ventilated ex-vivo human lungs with electromagnetic ground truth, the fused estimate halves absolute trajectory error and produces the lowest relative pose error among strong baselines. The result is a practical path to sensor-free, CT-free navigational bronchoscopy.

What carries the argument

The high-gain observer (Eqs. 18–21) that continuously corrects a nonholonomic tip model by three innovations: cross-track position residual, spherical heading misalignment from the lumen-derived vanishing direction, and scaled-speed residual from looming; large gains force the estimate onto the airway-following manifold and reject VO drift.

What would settle it

Run the same eight ex-vivo sequences while deliberately allowing substantial wall contact and non-zero yaw, then recompute ATE against electromagnetic tracking: if error rises above the best baseline (DPVO ~16.5 mm), the geometric prior is a bias rather than a corrective force.

Watch

Extended reading notes

Core claim

A geometry-aware visual-odometry pipeline that recovers forward heading from weighted vanishing-point rays of detected airway lumens, estimates insertion speed from looming, and fuses both with ordinary VO inside a high-gain observer built on a reduced-order nonholonomic airway model reduces absolute trajectory error by more than 50 percent and yields the lowest relative pose error on ex-vivo human-lung sequences with electromagnetic ground truth.

Load-bearing premise

The tip is assumed always to advance exactly along the airway centerline with zero yaw, so the observer’s cross-track correction recovers the true path rather than projecting real wall-contact or lateral slip onto an oversimplified tube model.

Editorial extensions

If this is right

Vision-only pose can replace CT-EM registration for many ICU procedures such as BAL and targeted biopsy when preoperative imaging is unavailable.
The same lumen-ray heading prior can be dropped into any monocular VO backend (feature-based or dense) to stabilize tubular endoscopy beyond bronchoscopy.
Real-time airway-consistent trajectories become available as input to downstream mapping, robotic tip control, or biopsy targeting without external sensors.
Scale-ambiguous monocular VO can be rescued by a single scalar looming cue plus geometric orientation, removing the need for stereo or depth sensors inside the scope.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

If the observer is relaxed to allow small yaw and soft lateral walls, the same architecture could handle more tortuous distal airways where the current hard nonholonomic constraint becomes unrealistic.
The lumen-detection + vanishing-ray module is modality-agnostic and could stabilize other endoluminal domains (ureteroscopy, sinus) that share the same tubular singularity.
Coupling the observer output to a lightweight online airway map would give a pure-vision SLAM system whose loop closures are anatomically constrained rather than purely photometric.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

The load-bearing result is concrete: on eight multi-lobe trajectories in mechanically ventilated ex-vivo human lungs with EMT ground truth, their lumen-ray heading + looming + high-gain airway observer beats ORB-SLAM2, LoFTR-VO, and DPVO, cutting ATE to 11.3 mm from the best baseline’s 16.5 mm and posting the lowest RPE. That is a real system number in a setting commercial platforms still treat as CT- or sensor-dependent, so the clinical motivation is not hand-waving.

What is actually new is the specific fusion, not the parts. YOLO lumen detection, monocular depth, LoFTR, vanishing points, and high-gain observers are all known. The useful move is back-projecting multiple lumen centers to weighted 3D rays for a stable forward heading when parallax collapses, then feeding that plus a looming speed cue into an observer that hard-enforces the reduced-order nonholonomic airway model (ψ ≡ 0, pure body-x advance). Simulation tunes the gains before transfer; the ex-vivo setup with a robot-mounted scope and distal EM sensor is carefully done. The paper is honest about texture poverty and tubular singularities and cites the right prior bronchoscopy VO work.

Soft spots, in proportion. Table I is a single aggregate after Sim(3); no per-sequence numbers, no variance, no success/failure rates, and no ablation of the observer against the same LoFTR backend. So you cannot yet tell how much of the gain is vanishing-point geometry versus simple cross-track rejection or low-pass filtering of intermittent VO. The kinematic prior is load-bearing in Eqs. 18–19; if distal runs include wall contact or lateral slip, α_p e_⊥ can inject bias that Sim(3) partially absorbs. Gains are hand-tuned, free parameters exist (depth percentile, scale bounds), and there is no code or data release. These are ordinary engineering gaps, not internal contradictions.

This is for people building endoscopic VO or ICU-capable bronchoscopy navigation. A serious referee should see it; the empirical claim is sharp enough and the design coherent enough to deserve that time, even if revision will demand variance, ablation, and clearer isolation of the prior. I would engage with the work and expect it to survive review with those additions.

Referee Report

4 major / 5 minor

Summary. The paper proposes a geometry-aware monocular visual odometry pipeline for bronchoscopic navigation that recovers a stable forward heading from multi-lumen vanishing-point rays (YOLO detections refined by monocular depth, back-projected and weighted) and a scaled insertion velocity from looming, then fuses these cues with a noisy VO backend (LoFTR) inside a high-gain observer whose dynamics enforce a reduced-order nonholonomic airway model (ψ ≡ 0, pure body-x advance). The observer (Eqs. 18–21) rejects cross-track drift and spherical orientation error while adapting an unknown scale factor. Validation uses eight multi-lobe trajectories (30 s–5 min) on ex-vivo mechanically ventilated human lungs with Aurora EMT ground truth; after Sim(3) alignment the method reports ATE 11.3 mm / RPE 8.6° versus ORB-SLAM2, LoFTR-VO and DPVO, claiming >50 % ATE reduction relative to the best baseline.

Significance. If the accuracy claims hold under broader conditions, the work supplies a practical route to CT-free, sensor-light navigational bronchoscopy usable in critical-care and resource-constrained settings where pre-operative imaging or EM infrastructure is unavailable. Strengths that raise the contribution above incremental VO engineering include: (i) real ventilated human-lung tissue rather than phantoms or short synthetic clips, (ii) multi-minute multi-lobe trajectories with external EMT ground truth, and (iii) an explicit geometric prior that directly targets the vanishing-point singularity of tubular airways. The combination of lumen-derived heading, looming speed and a lightweight high-gain observer is novel for this domain and could seed downstream mapping or robotic control modules.

major comments (4)

Table I reports only aggregate mean ATE/RPE over the eight trajectories after Sim(3) alignment; no per-sequence values, standard deviations, success/failure rates or confidence intervals are given. With n = 8 and trajectories that differ substantially in length and distal complexity, the claimed “>50 % ATE reduction” cannot be assessed for consistency or statistical reliability. A per-sequence breakdown (and ideally a paired statistical test) is required to support the central empirical claim.
The experimental design never ablates the high-gain observer (Eqs. 18–21) against the identical LoFTR-VO backend that supplies p_m. Consequently it is impossible to isolate whether the reported gain arises from vanishing-point heading + nonholonomic projection, from simple low-pass filtering of intermittent VO, or from the particular eight paths chosen. An ablation that freezes the observer gains to zero (or replaces the observer by a plain Kalman smoother) is load-bearing for attributing performance to the geometry-aware fusion.
Section II-A (Eqs. 2–3) and the observer innovation (Eq. 18) hard-enforce ψ ≡ 0 and pure body-x advance along the airway centerline via the cross-track rejection term α_p e_⊥. Real distal navigation frequently involves wall contact and lateral slip; if any of the eight runs contain such motion, the geometric prior injects systematic bias that Sim(3) can partially absorb, inflating apparent accuracy. The manuscript should either (a) quantify residual lateral error against EMT or (b) relax the nonholonomic constraint and re-evaluate.
Observer gains (α_o = 15, α_p = 8, au_∥ = 0.35, au_v = 6, au_v = 1.2, ℓ_κ = 0.2) are tuned exclusively on synthetic straight-plus-arc trajectories and transferred without sensitivity analysis or re-tuning on real data. Because the free-parameter set also includes the depth-percentile threshold and ray-weight ε, a brief sensitivity study (or leave-one-trajectory-out gain selection) is needed to show that the >50 % claim is not an artifact of the particular gain vector.

minor comments (5)

Figure 4 caption and text refer to “five exemplary trials” while the simulation description mentions ten trajectories; clarify the selection criterion.
Equation (13) states ˙ρ ∝ v_z / Z but the subsequent text treats the median optical-flow magnitude as a direct proxy for insertion velocity; a short derivation or calibration note would make the scale-ambiguity handling clearer.
The depth network is cited as [33] (BREA-Depth) without stating whether it was fine-tuned on the same five lungs used for YOLO training; domain-shift risk should be noted.
RPE is reported “over Δt = 10 frames” in the table caption but “60 frames at 10 fps” in the text; reconcile the two statements.
Minor typographical inconsistencies appear (e.g., “V anishing-Point”, mixed use of ˜v versus ˜v, and “typ. σ_p = 20,mm” with comma as decimal).

Circularity Check

0 steps flagged · score 1.0 of 10

No load-bearing circularity; empirical ATE/RPE gains measured against external EMT ground truth after independent synthetic gain tuning.

full rationale

The derivation chain is self-contained and non-circular. The reduced-order nonholonomic model (Sec. II-A, Eqs. 1–3) is an explicit modeling assumption (ψ≡0, pure body-x advance) used to construct the observer innovations (Eqs. 17–21); it is not fitted to the evaluation data nor defined in terms of the reported ATE/RPE. Vanishing-point heading (Eqs. 4–11) and looming velocity are computed from image detections and depth, then fused with an off-the-shelf VO backend; the fusion does not redefine the external electromagnetic-tracking ground truth. Observer gains were fixed on synthetic trajectories (Sec. III) before being applied unchanged to the eight ex-vivo sequences. Self-citations (BREA-Depth depth network, prior challenge papers) supply modular components or motivation but are not invoked as uniqueness theorems that force the central claim. The >50% ATE reduction (Table I) is an empirical comparison after Sim(3) alignment to independent EMT poses, not a quantity recovered by construction from the method’s own inputs. Minor author-overlap citations exist but are not load-bearing; score remains 1.

Assumptions & free parameters 3 free parameters · 4 assumptions · 2 invented entities

The central accuracy claim rests on a domain kinematic prior, several hand-chosen observer and detection thresholds, and a custom fusion architecture. External EMT provides independent evaluation, so the ledger is moderate rather than heavy; the free parameters mainly affect smoothness and scale recovery, not the existence of the heading cue.

free parameters (3)

observer gains (α_o, α_p, k_∥, α_v, k_v, ℓ_κ) = α_o=15, α_p=8, k_∥=0.35, α_v=6, k_v=1.2, ℓ_κ=0.2
Fixed by trial-and-error on ten synthetic trajectories to minimize RMSE, then frozen for all real experiments (Sec. III). Values: α_o=15, α_p=8, k_∥=0.35, α_v=6, k_v=1.2, ℓ_κ=0.2.
depth percentile threshold for lumen center refinement = 80th percentile
Pixels deeper than the 80th percentile inside each YOLO box define the lumen center (Sec. II-C). Chosen experimentally; changes the ray directions that feed heading.
ray weight ε and scale bounds [κ_min, κ_max]
ε≪1 regularizes inverse-depth weights; κ is projected into unspecified bounds during online adaptation (Eqs. 8, 21). Affects heading fusion and scale recovery.

assumptions (4)

domain assumption Bronchoscope tip motion obeys the reduced-order nonholonomic model with ψ=0 and body-x velocity only (Eqs. 2–3).
Sec. II-A derives Pfaffian constraints from the claim that the tip follows the airway centerline; this null-space basis is the structure the observer enforces.
domain assumption Detected lumen entrances act as portals aligned with the true bronchial axis, so their weighted back-projected rays yield a usable forward heading.
Stated in the introduction and Sec. II-C / Fig. 1; load-bearing for the geometry-aware prior when parallax is absent.
ad hoc to paper A high-gain observer with algebraic cross-track and spherical innovations can reject systematic VO drift faster and more stably than EKF/UKF in this setting.
Sec. II-D argues classical stochastic filters are ill-suited because the dominant error is bias, not zero-mean noise; the specific innovation structure (Eqs. 17–21) is paper-specific.
standard math Standard pinhole projection and monocular depth/pose modules supply usable rays and VO position traces.
Camera model K, LoFTR-VO backend, and monocular depth network are taken as given inputs to the fusion.

invented entities (2)

Weighted multi-lumen vanishing-direction estimator for bronchoscope heading
purpose: Produce a stable forward axis (θ, ϕ) from YOLO+depth lumen centers when classical parallax cues vanish.
Eqs. 4–11 define the specific back-projection, inverse-depth weighting, and arctan2 extraction used as d_m.
Airway-constrained high-gain observer fusing VO, vanishing heading, and looming speed
purpose: Convert noisy scale-ambiguous VO into a smooth tip trajectory that rejects cross-track drift and adapts scale κ.
Dynamics (18)–(21) and error definitions (17) are bespoke; not a standard off-the-shelf filter.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Geometry-Aware Visual Odometry for Bronchoscopic Navigation via High-Gain Observer Fusion." pith.science (2026). https://pith.science/paper/ZMYBNGP5

@misc{pith2026260705162,
  author       = {Pith},
  title        = {Pith review of: Geometry-Aware Visual Odometry for Bronchoscopic Navigation via High-Gain Observer Fusion},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZMYBNGP5}},
  note         = {Machine review of arXiv:2607.05162}
}

read the original abstract

Navigational bronchoscopy is critical for pulmonary interventions, yet current platforms depend heavily on pre-operative CT or external sensors, limiting their use in critical care and resource-constrained settings. Vision-only navigation offers a scalable alternative, but conventional visual odometry (VO) struggles with texture-poor airway images, specularities, and the vanishing-point singularities of tubular anatomy, leading to frequent tracking failures and drift. We present a geometry-aware VO framework that explicitly leverages vanishing-point cues from airway lumens. Detected lumens are back-projected to 3D rays, whose weighted fusion yields a stable forward heading even when parallax cues are absent. This heading, together with looming-based velocity estimates, is fused with noisy VO outputs using a bespoke high-gain observer that enforces airway-following priors and rejects drift. We validate the method on ex-vivo mechanically ventilated human lungs with electromagnetic tracking ground truth. Compared to state-of-the-art pipelines (ORB-SLAM2, LoFTR-VO, DPVO), our approach reduces absolute trajectory error by more than 50% and achieves the lowest relative pose error across all test sequences.

Figures

Figures reproduced from arXiv: 2607.05162 by the authors.

**Figure 2.** Block diagram of the proposed framework. Input [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗

**Figure 3.** A schematic of reduced-order model of bronchoscopy. [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗

Figures from the paper (3 more)

**Figure 4.** Figure 4: Simulation results for the bronchoscope trajectory observer. (a) A representative sequence showing the airway [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]

**Figure 5.** Figure 5: RGB bronchoscopy frames with corresponding depth [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]

**Figure 7.** Figure 7: Exemplary results from three trajectories showing observer-based bronchoscope pose estimation compared with EMT [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

37 extracted references · 4 canonical work pages

[1]

Establishing the diagnosis of lung cancer,

M. P. Rivera, A. C. Mehta, and M. M. Wahidi, “Establishing the diagnosis of lung cancer,”Chest, vol. 143, no. 5, pp. e142S–e165S, May 2013

2013
[2]

Ion: Technology and techniques for shape-sensing robotic-assisted bronchoscopy,

J. Reisenauer, M. J. Simoff, M. A. Pritchett, D. E. Ost, A. Majid, C. Keyes, R. F. Casal, M. S. Parikh, J. Diaz-Mendoza, S. Fernandez- Bussy, and E. E. Folch, “Ion: Technology and techniques for shape-sensing robotic-assisted bronchoscopy,”The Annals of Thoracic Surgery, vol. 113, no. 1, pp. 308–315, Jan. 2022

2022
[3]

Robotic bronchoscopy for diagnosis of suspected lung cancer: A feasibility study,

J. R. Rojas-Solano, L. Ugalde-Gamboa, and M. Machuzak, “Robotic bronchoscopy for diagnosis of suspected lung cancer: A feasibility study,”Journal of Bronchology & Interventional Pulmonology, vol. 25, no. 3, pp. 168–175, 2018

2018
[4]

Saghaie, J

T. Saghaie, J. P. Williamson, M. Phillips, D. Kafili, S. Sundar, D. K. Hogarth, and A. Ing, “First-in-human use of a new robotic electromag- netic navigation bronchoscopic platform with integrated tool-in-lesion tomosynthesis (TiLT) technology for peripheral pulmonary lesions: The FRONTIER study,”Respirology, vol. 29, no. 11, pp. 969–975, 2024

2024
[5]

Electromagnetic navigation bronchoscopy to access lung lesions in 1,000 subjects: First results of the prospective, multicenter NA VIGATE study,

S. J. Khandhar, M. R. Bowling, J. Flandes, T. R. Gildea, K. L. Hood, W. S. Krimsky, D. J. Minnich, S. D. Murgu, M. Pritchett, E. M. Toloza, M. M. Wahidi, J. J. Wolvers, E. E. Folch, and NA VIGATE Study Investigators, “Electromagnetic navigation bronchoscopy to access lung lesions in 1,000 subjects: First results of the prospective, multicenter NA VIGATE s...

2017
[6]

Rickets, K

W. Rickets, K. K. W. Lau, V . Pollit, S. Mealing, C. Leonard, P. Mallender, N. Chaudhuri, P. L. Shah, and U. B. Naidu, “Exploratory cost-effectiveness model of electromagnetic navigation bronchoscopy (ENB) compared with CT-guided biopsy (TTNA) for diagnosis of malignant indeterminate peripheral pulmonary nodules,”BMJ Open Respiratory Research, vol. 7, no....

2020
[7]

Bronchoscopy in intubated and non-intubated intensive care unit patients with respiratory failure,

S. Patolia, R. Farhat, and R. Subramaniyam, “Bronchoscopy in intubated and non-intubated intensive care unit patients with respiratory failure,”Journal of Thoracic Disease, vol. 13, no. 8, pp. 5125–5134, 2021. [Online]. Available: https://pmc.ncbi.nlm.nih.gov/ articles/PMC8411155/

2021
[8]

Bronchoscopic diagnosis of severe respiratory infections,

M. R ¨oder, A. Y . K. C. Ng, and A. Conway Morris, “Bronchoscopic diagnosis of severe respiratory infections,”Journal of Clinical Medicine, vol. 13, no. 19, p. 6020, 2024. [Online]. Available: https://doi.org/10.3390/jcm13196020

work page doi:10.3390/jcm13196020 2024

Show all 37 references

[9]

A 4DCT imaging-based breathing lung model with relative hysteresis,

S. Miyawaki, S. Choi, E. A. Hoffman, and C.-L. Lin, “A 4DCT imaging-based breathing lung model with relative hysteresis,”Journal of Computational Physics, vol. 326, pp. 76–90, Dec. 2016

2016
[10]

Harnessing foundation models for robust and generalizable 6-dof bronchoscopy localization,

Q. Tian, H. Liao, X. Huang, B. Yang, and H. Liu, “Harnessing foundation models for robust and generalizable 6-dof bronchoscopy localization,”arXiv preprint arXiv:2505.24249, 2025. [Online]. Available: https://doi.org/10.48550/arXiv.2505.24249

2025 doi arXiv
[11]

Visually navigated bronchoscopy using three cycle-consistent generative adversarial network for depth estimation,

A. Banach, F. King, F. Masaki, H. Tsukada, and N. Hata, “Visually navigated bronchoscopy using three cycle-consistent generative adversarial network for depth estimation,”Medical Image Analysis, vol. 73, p. 102164, 2021. [Online]. Available: https://doi.org/10.1016/j.media.2021.102164

2021 doi
[12]

Asano,Application and Limitations of Virtual Bronchoscopic Nav- igation

F. Asano,Application and Limitations of Virtual Bronchoscopic Nav- igation. In Flexible Bronchoscopy (eds K.-P . Wang, A.C. Mehta and J.F . Turner). Blackwell Publishing Ltd, 2012, pp. 266–269

2012
[13]

Interactive ct-video registration for the continuous guidance of bronchoscopy,

S. A. Merritt, R. Khare, andet al., “Interactive ct-video registration for the continuous guidance of bronchoscopy,”IEEE transactions on medical imaging, vol. 32, no. 8, pp. 1376–1396, 2013

2013
[14]

Development and comparison of new hybrid motion tracking for bronchoscopic navigation,

X. Lu ´o, M. Feuerstein, andet al., “Development and comparison of new hybrid motion tracking for bronchoscopic navigation,”Medical Image Analysis, vol. 16, no. 3, pp. 577–596, 2012

2012
[15]

Branch:bifurcation recognition for airway navigation based on structural characteristics,

M. Shen, S. Giannarou, andet al., “Branch:bifurcation recognition for airway navigation based on structural characteristics,” inMedical Image Computing and Computer-Assisted Intervention - MICCAI
[16]

Cham: Springer International Publishing, 2017, pp. 182–189

2017
[17]

Context-aware depth and pose estimation for bronchoscopic navigation,

M. Shen, Y . Gu, andet al., “Context-aware depth and pose estimation for bronchoscopic navigation,”IEEE Robotics and Automation Letters, vol. 4, no. 2, pp. 732–739, 2019

2019
[18]

Autonomous driving in the lung using deep learning for localization,

J. Sganga, D. Eng, andet al., “Autonomous driving in the lung using deep learning for localization,” 2019

2019
[19]

A bronchoscopic navigation method based on neural radiation fields,

L. Zhu, J. Zheng, C. Wang, J. Jiang, and A. Song, “A bronchoscopic navigation method based on neural radiation fields,” International Journal of Computer Assisted Radiology and Surgery, vol. 19, no. 10, pp. 2011–2021, Oct. 2024. [Online]. Available: https://doi.org/10.1007/s11...

2011 doi
[20]

Pans: Probabilistic airway navigation system for real-time robust broncho- scope localization,

Q. Tian, J. Luo, X. Huang, H. Liao, B. Yang, and H. Liu, “Pans: Probabilistic airway navigation system for real-time robust broncho- scope localization,” inInternational Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI). Springer, 2024, pp. 3–13

2024
[21]

Sage: Slam with appearance and geometry prior for endoscopy,

X. Liu, Z. Li, M. Ishii, G. D. Hager, R. H. Taylor, and M. Unberath, “Sage: Slam with appearance and geometry prior for endoscopy,” in 2022 International Conference on Robotics and Automation (ICRA), 2022, pp. 5587–5593

2022
[22]

Dynamic view expansion for minimally invasive surgery using simultaneous localization and mapping,

P. Mountney and G. Yang, “Dynamic view expansion for minimally invasive surgery using simultaneous localization and mapping,” in IEEE International Conference on Engineering in Medicine and Biology (EMBC), 2010, pp. 2107–2110

2010
[23]

Sd-defslam: Semi-direct monocular slam for deformable and dynamic environments,

J. Gomez Rodriguez, J. Lamarcaet al., “Sd-defslam: Semi-direct monocular slam for deformable and dynamic environments,” inIEEE International Conference on Robotics and Automation (ICRA), 2021, pp. 14 253–14 259

2021
[24]

Distinctive Image Features from Scale-Invariant Key- points,

D. G. Lowe, “Distinctive Image Features from Scale-Invariant Key- points,”International Journal of Computer Vision, vol. 60, no. 2, pp. 91–110, Nov. 2004

2004
[25]

Orb: An efficient alternative to sift or surf,

E. Rublee, V . Rabaud, andet al., “Orb: An efficient alternative to sift or surf,” in2011 International Conference on Computer Vision, 2011, pp. 2564–2571

2011
[26]

On challenges of monocular pose estimation for endoluminal navi- gation,

E. Mackut ˙e, A. Abdalla, S. Dickson, K. Dhaliwal, and M. Khadem, “On challenges of monocular pose estimation for endoluminal navi- gation,”Journal of Medical Robotics Research, vol. 09, no. 03n04, p. 2440009, 2024

2024
[27]

Feature-based visual odometry for bron- choscopy: A dataset and benchmark,

J. Deng, P. Li, andet al., “Feature-based visual odometry for bron- choscopy: A dataset and benchmark,” in2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). Detroit, Michigan, USA: IEEE, Oct. 2023

2023
[28]

Visual slam-based bronchoscope tracking with deformable lung phantom validation,

Y . Wang, L. Sun, J. Zhanget al., “Visual slam-based bronchoscope tracking with deformable lung phantom validation,”IEEE Access, vol. 8, pp. 107 907–107 918, 2020

2020
[29]

LoFTR: Detector-free local feature matching with transformers,

J. Sun, Z. Shen, Y . Wang, H. Bao, and X. Zhou, “LoFTR: Detector-free local feature matching with transformers,”CVPR, 2021

2021
[30]

Endoslam dataset and an unsupervised monocular visual odometry and depth estimation approach for endoscopic videos,

K. B. Ozyoruk, G. I. Gokceler, T. L. Bobrow, G. Coskun, K. Incetan, Y . Almalioglu, F. Mahmood, E. Curto, L. Perdigoto, M. Oliveira, H. Sahin, H. Araujo, H. Alexandrino, N. J. Durr, H. B. Gilbert, and M. Turan, “Endoslam dataset and an unsupervised monocular visual odometry an...

2021
[31]

Self-supervised direct pose estimation for vision-based tracking in navigated bron- choscopy,

P. Kalia, F. King, F. Masaki, H. Tsukada, and N. Hata, “Self-supervised direct pose estimation for vision-based tracking in navigated bron- choscopy,”Computers in Biology and Medicine, vol. 178, p. 110958, 2025

2025
[32]

Ultralytics yolov8,

G. Jocher, A. Chaurasia, and J. Qiu, “Ultralytics yolov8,” 2023. [Online]. Available: https://github.com/ultralytics/ultralytics

2023
[33]

Simple online and realtime tracking,

A. Bewley, Z. Ge, L. Ott, F. Ramos, and B. Upcroft, “Simple online and realtime tracking,” in2016 IEEE International Conference on Image Processing (ICIP). IEEE, 2016, pp. 3464–3468

2016
[34]

BREA-Depth: Bronchoscopy Realistic Airway- geometric Depth Estimation ,

F. X. Zhang, E. Mackute, M. Kasaei, K. Dhaliwal, R. Thomson, and M. Khadem, “ BREA-Depth: Bronchoscopy Realistic Airway- geometric Depth Estimation ,” inproceedings of Medical Image Computing and Computer Assisted Intervention – MICCAI 2025, vol. LNCS 15968. Springer Nature Sw...

2025
[35]

A benchmark for the evaluation of rgb-d slam systems,

J. Sturm, N. Engelhard, F. Endres, W. Burgard, and D. Cremers, “A benchmark for the evaluation of rgb-d slam systems,” inProceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Vilamoura, Portugal, 2012, pp. 573–580

2012
[36]

Orb–slam2: An open-source slam sys- tem for monocular, stereo, and rgb-d cameras,

R. Mur-Artal and J. D. Tard ´os, “Orb–slam2: An open-source slam sys- tem for monocular, stereo, and rgb-d cameras,” inIEEE Transactions on Robotics, vol. 33, no. 5, 2017, pp. 1255–1262

2017
[37]

Dpvo: Dense plane-based visual odometry,

Z. Teed and T. Zhao, “Dpvo: Dense plane-based visual odometry,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2023

2023

Pith tools

Reviewed July 11, 2026 · model on record in the stance chip above.

[1] [1]

Establishing the diagnosis of lung cancer,

M. P. Rivera, A. C. Mehta, and M. M. Wahidi, “Establishing the diagnosis of lung cancer,”Chest, vol. 143, no. 5, pp. e142S–e165S, May 2013

2013

[2] [2]

Ion: Technology and techniques for shape-sensing robotic-assisted bronchoscopy,

J. Reisenauer, M. J. Simoff, M. A. Pritchett, D. E. Ost, A. Majid, C. Keyes, R. F. Casal, M. S. Parikh, J. Diaz-Mendoza, S. Fernandez- Bussy, and E. E. Folch, “Ion: Technology and techniques for shape-sensing robotic-assisted bronchoscopy,”The Annals of Thoracic Surgery, vol. 113, no. 1, pp. 308–315, Jan. 2022

2022

[3] [3]

Robotic bronchoscopy for diagnosis of suspected lung cancer: A feasibility study,

J. R. Rojas-Solano, L. Ugalde-Gamboa, and M. Machuzak, “Robotic bronchoscopy for diagnosis of suspected lung cancer: A feasibility study,”Journal of Bronchology & Interventional Pulmonology, vol. 25, no. 3, pp. 168–175, 2018

2018

[4] [4]

Saghaie, J

T. Saghaie, J. P. Williamson, M. Phillips, D. Kafili, S. Sundar, D. K. Hogarth, and A. Ing, “First-in-human use of a new robotic electromag- netic navigation bronchoscopic platform with integrated tool-in-lesion tomosynthesis (TiLT) technology for peripheral pulmonary lesions: The FRONTIER study,”Respirology, vol. 29, no. 11, pp. 969–975, 2024

2024

[5] [5]

Electromagnetic navigation bronchoscopy to access lung lesions in 1,000 subjects: First results of the prospective, multicenter NA VIGATE study,

S. J. Khandhar, M. R. Bowling, J. Flandes, T. R. Gildea, K. L. Hood, W. S. Krimsky, D. J. Minnich, S. D. Murgu, M. Pritchett, E. M. Toloza, M. M. Wahidi, J. J. Wolvers, E. E. Folch, and NA VIGATE Study Investigators, “Electromagnetic navigation bronchoscopy to access lung lesions in 1,000 subjects: First results of the prospective, multicenter NA VIGATE s...

2017

[6] [6]

Rickets, K

W. Rickets, K. K. W. Lau, V . Pollit, S. Mealing, C. Leonard, P. Mallender, N. Chaudhuri, P. L. Shah, and U. B. Naidu, “Exploratory cost-effectiveness model of electromagnetic navigation bronchoscopy (ENB) compared with CT-guided biopsy (TTNA) for diagnosis of malignant indeterminate peripheral pulmonary nodules,”BMJ Open Respiratory Research, vol. 7, no....

2020

[7] [7]

Bronchoscopy in intubated and non-intubated intensive care unit patients with respiratory failure,

S. Patolia, R. Farhat, and R. Subramaniyam, “Bronchoscopy in intubated and non-intubated intensive care unit patients with respiratory failure,”Journal of Thoracic Disease, vol. 13, no. 8, pp. 5125–5134, 2021. [Online]. Available: https://pmc.ncbi.nlm.nih.gov/ articles/PMC8411155/

2021

[8] [8]

Bronchoscopic diagnosis of severe respiratory infections,

M. R ¨oder, A. Y . K. C. Ng, and A. Conway Morris, “Bronchoscopic diagnosis of severe respiratory infections,”Journal of Clinical Medicine, vol. 13, no. 19, p. 6020, 2024. [Online]. Available: https://doi.org/10.3390/jcm13196020

work page doi:10.3390/jcm13196020 2024