Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2026 Mar 24.
Published in final edited form as: Otolaryngol Head Neck Surg. 2025 Nov 12;174(1):219–229. doi: 10.1002/ohn.70062

Integration of Trackerless Surface Reconstruction-Based Surgical Navigation With Exoscopic Trans-Mastoid Surgery

Ryan A Bartholomew 1,2,†, Haoyin Zhou 3,†, Kyle Mital 2, Afash Haleem 1, Drew J Montigny 1, Timothy R Smith 4, Jeffrey P Guenette 3, Jayender Jagadeesan 1,‡, C Eduardo Corrales 1,2,‡
PMCID: PMC13006913  NIHMSID: NIHMS2147507  PMID: 41221830

Abstract

Objective.

A trackerless approach to surgical navigation employing simultaneous localization and mapping (SLAM) algorithms based on surface reconstruction from 3D endoscopy fused with preoperative imaging has been demonstrated to have approximately 1 mm mean surgical navigation errors in the anterior and lateral skull base. This study evaluates the feasibility of a SLAM-based surgical navigation system during combined exoscopic-endoscopic trans-mastoid surgery.

Study Design.

Interventional deceased donor cohort study.

Setting.

Tertiary care hospital.

Methods.

Trans-mastoid exoscopic dissections were performed in deceased donor temporal bones. At two operative steps, surface models were reconstructed from stereoscopic exoscopy or endoscopy video, and volumetric computed tomography (CT) models were segmented to generate anatomic models. The surface reconstructions and CT-generated models were coregistered using corresponding features, and accuracy was quantified using four metrics: reconstruction, registration, and exoscope-exoscope or endoscope-exoscope surface-surface (S-S) registration errors.

Results.

High-fidelity exoscopic surface model reconstruction with coregistration to CT or endoscope surface models was performed after mastoid bone exposure and mastoidectomy in 10 deceased donor temporal bones. The mean (SD) reconstruction, registration, exoscope-exoscope S-S registration, and endoscope-exoscope S-S registration errors in millimeters, respectively, were 0.72 (0.32), 1.43 (0.49), 1.72 (0.59), and 1.49 (0.61).

Conclusion.

A SLAM-based surgical navigation system can enable accurate localization of anatomic points of interest, which is maintained throughout exoscopic surgical dissection and between alternating use of exoscopy and endoscopy. With further development, this approach to navigation has the potential to provide accurate and continuous navigation data during lateral skull base surgery without compromising surgical efficiency and economy of motion.

Keywords: augmented reality, endoscopy, exoscope, image guidance, lateral skull base surgery, neurotology, otology, simultaneous localization and mapping (SLAM), surface reconstruction, surgical navigation

Surgery of the lateral skull base requires carefully drilling through opaque bone to expose pathology while avoiding fragile anatomic structures.1 Misnavigation of these millimeter-scale surgical corridors can result in iatrogenic hearing loss, dizziness, and facial paralysis.2–6 Surgical navigation systems are available as adjunct tools to supplement a surgeon’s anatomic knowledge and clinical experience. Current surgical navigation systems register intraoperative anatomy to medical imaging using either optical or electromagnetic tracking methodologies.7–11 However, these systems are infrequently utilized in the lateral skull base.12,13 Deterrents to their use are the additional monitors and external tracking equipment required, optical systems requiring maintenance of line of sight of tracked instruments,12,13 electromagnetic systems being subject to tracking errors from electromagnetic interference,10,14,15 and the high accuracy requirements of lateral skull base surgery.13,16–18

We recently developed a novel approach to surgical navigation, which may avoid many of these shortcomings.19,20 Three-dimensional (3D) models of the operative field surface are reconstructed in real time from 3D endoscopic video and registered to volumetric models segmented from preoperative imaging, all without requiring external tracking equipment. The surgeon can thereby be provided with continuous information about anatomic structures underlying the exposed tissue surface, even while the endoscope is in motion. The surface models are generated by stitching video frames captured with 3D endoscopy into a 3D mosaic using stereo matching and then aligning the resultant data using simultaneous localization and mapping (SLAM) algorithms. Validation in deceased donor models of both anterior20 and lateral skull base surgery (H. Zhou and R. Bartholomew, unpublished data under review, 2025) has demonstrated approximately 1 mm mean surgical navigation errors, with feasibility being supported by early evaluations using clinical data.

Our initial SLAM-based navigation work employed endoscopy, whose wide-angle view and angled optics permit visualization and dissection down narrow surgical corridors that are inaccessible to direct line of site microscopy.21,22 However, the majority of trans-mastoid and lateral skull base dissection is performed in a two-handed manner using microscopy, so a surgical navigation system restricted to the use of endoscopy has limited accessibility. Exoscopes, or extracorporeal video microscopes, are an emerging alternative to traditional operating microscopes in the lateral skull base and other surgical fields.21,23–26 Exoscopes consist of a high-definition or 4K video camera and an LED or fiberoptically delivered light source that is suspended above the operative field using a robotic or manually actuated articulating holder. The image is displayed in 2D or 3D on a high-resolution monitor in the surgeon’s view at eye level. The benefits of exoscopy compared to microscopy include superior ergonomics with heads-up surgery body mechanics, compact size, decreased cost, and an equivalent or improved visual experience for both the surgeon and observers.27–30 Importantly, as a digital stereoscopic visualization modality, exoscopes should be compatible with SLAM-based surgical navigation. An exoscope-integrated SLAM-based navigation system would permit two-handed dissection with the availability of surgical navigation throughout a trans-mastoid or lateral skull base operation without requiring any additional equipment. Moreover, intermittent use of endoscopy is still possible with preservation of surgical navigation during switches in visualization modalities.

The proposed workflow of an exoscope integrated SLAM-based navigation system (Figure 1) begins with preoperative acquisition of patient imaging—submillimeter spatial resolution computed tomography (CT) or magnetic resonance imaging (MRI). The imaging is then used to generate a volumetric model with segmented anatomic structures (Figure 1A). Intraoperatively, and after dissection of overlying soft tissue prior to drilling, a 3D surface model of the operative field is reconstructed from a brief exoscopic or endoscopic video of the exposed bone (Figure 1B). The surface model is then aligned (ie, coregistered) with the volumetric imaging model using corresponding features present in both models. The result is an intuitive representation of the relationship between the exposed tissue surface and underlying anatomy (Figure 1C). Navigation data could be presented on a secondary monitor or on the primary monitor superimposed over the surgeon’s view of the operative field (ie, augmented reality navigation, Figures 1C and 2).20 As dissection progresses or as the surgeon switches between exoscopy and endoscopy, continuous localization of anatomic structures with preservation of coregistration to imaging data can be maintained using video-based features (Figure 1C, bottom panel).

Figure 1.

Figure 1.

Open in a new tab

Simultaneous localization and mapping (SLAM)-based surgical navigation workflow. A volumetric model is segmented from preoperative imaging (A) and coregistered to a surface model reconstructed from intraoperative video (B). Coregistration is maintained during dissection, allowing for continuous surgical navigation (C). CT, computed tomography.

Figure 2.

Figure 2.

Open in a new tab

Augmented exoscopic view for surgical navigation. Prototype augmented exoscopic view for surgical navigation with hologram reprojections of segmented anatomic structures from computed tomography (CT) imaging. Shown before (A) and after (B) right-sided cortical mastoidectomy.

As an initial assessment of clinical feasibility for combined exoscopic-endoscopic surgical navigation in the lateral skull base, we evaluated a SLAM-based surgical navigation system in deceased donor models of trans-mastoid exoscopic surgery.

Methods

The SLAM-based surgical navigation system was evaluated using a deceased donor temporal bone model of 3D-exoscopic mastoidectomy. An ORBEYE® 4K 3D exoscope (Olympus America) was used for surgical dissection and acquisition of stereoscopic video, which served as the input for reconstruction of 3D surface models of the operative field. Stereoscopic video for surface model generation was also acquired using a 0 degree 4 mm TIPCAM®1 S 3D endoscope (KARL STORZ United States). The study was reviewed and approved by the Mass General Brigham institutional review board.

The following experimental protocol was repeated at two operative steps (mastoid bone exposure and cortical mastoidectomy).

  1. Surgical dissection and placement of screw fiducials (Figure 3A)

  2. Acquisition of stereoscopic video of the operative field and surface model reconstruction (Figure 3B)

  3. CT imaging of the deceased donor specimen and volumetric CT model segmentation (Figure 3C)

  4. Coregistration of surface and CT models with assessment of accuracy and navigational fidelity (Figure 3D)

Figure 3.

Figure 3.

Open in a new tab

Evaluation of exoscopic simultaneous localization and mapping (SLAM)-based surgical navigation in deceased donor specimens. Following exposure of the mastoid bone and mastoidectomy (A), surface (B) and computed tomography (CT) (C) models were generated and coregistered (D) using natural landmarks (green spheres).

Surgical Dissection and Fiducial Placement

Temporal bones extracted from deceased donor skulls and preserved with formaldehyde were used. Two operative steps were completed. The first operative step was mastoid bone exposure with removal of overlying soft tissue and elevation of the external auditory canal (EAC) skin (when present) from the lateral bony EAC. To limit the availability of bony landmarks for surgical navigation to those present during typical clinical circumstances, the boundaries of exposed bone during standard surgical conditions were outlined with India ink and reviewed by a senior neurotologist (C.E.C.). Five titanium screws (1.2 mm) were fixed to the bone surrounding the anticipated mastoid cavity for quantitative validation. Of note, screw fiducials were used for quantitative validation purposes only and are not an expected component of the proposed clinical workflow. The second operative step was a cortical mastoidectomy using a drill to expose the aditus ad antrum. The original five titanium screws were left in place, and an additional three screws were fixed to bone within the mastoid cavity, with a resultant eight total screw fiducials.

CT Imaging and Volumetric Model Segmentation

Temporal bone CT images of deceased donor specimens were acquired using a SOMATOM Force or Biograph mCT CT system (Siemens) with 0.6 mm slice thickness, 120 kVp, modulated mA, and 0.8 pitch. Images were reconstructed with a bone kernel. Anatomic structures were segmented in 3D Slicer31 using semiautomatic image intensity-based methods and included bone, ossicles, inner ear labyrinth, otic capsule (defined as the inner ear labyrinth segment with 1 mm bony margins expanded in all directions), facial nerve, internal auditory canal, sigmoid sinus, and screw fiducials.

SLAM-Based 3D Reconstruction of the Operative Field

Stereoscopic video was acquired by the calibrated exoscope at a fixed magnification and focal length (250 mm) in 1080p. Video clips were approximately 30 to 60 seconds in duration and were acquired in standard otologic surgical orientation. To permit video acquisition of the operative field from additional angles, the temporal bone was gently rotated without leaving the confines of standard surgical orientation. Stereoscopic video acquisition in 720p by the calibrated endoscope was performed over 30 to 60 seconds and with panning of the endoscope over all surfaces of the operative field. Using our previously described methods,19,20 3D models of the operative field were reconstructed using stereoscopic videos as input. In brief, the SLAM algorithm employs a stereo matching module that estimates the physical depth of corresponding pixels between the left and right camera of a stereoscopic camera. A mosaicking algorithm then assembles these data to create a high-resolution, single-layer dense point cloud in real-time—that is, a 3D surface model of the operative field.

Assessment of Model Coregistration Accuracy and Navigational Fidelity

Within the 3D Slicer software, corresponding landmarks present in both the surface and segmented CT models are annotated as fiducials (green and red spheres in Figures 3 and 4, respectively). These landmarks are aligned using a 3D point cloud alignment algorithm32 to generate 6-degree-of-freedom (6-DoF) transformations. The resultant transform is applied, resulting in a coregistered surface and CT model, which permits 3D visualization of the segmented anatomy underlying the operative field surface. To account for differences in calibration between the exoscope and endoscope, similarity transforms were used for registration of endoscopy surface models, which preserved the shapes of the models but permitted changes in scale. Otherwise, rigid transformations were used.

Figure 4.

Figure 4.

Open in a new tab

Preservation of registration across time and between exoscopy and endoscopy. Following initial computed tomography (CT) and surface model coregistration after mastoid exposure (A), registration is preserved as dissection progresses (B) and when switching between visualization modalities (C). Naturals landmarks used for coregistration indicated by red spheres.

Four quantitative error metrics are reported:

  1. Reconstruction error: As described in our prior studies,20 this metric evaluates how accurately the surface model represents the volumetric CT model and provides an assessment of the fidelity of the SLAM-based surface reconstruction. The coordinates of screw fiducials are annotated on both the surface and CT models (blue spheres in Figures 3B and C). Using the coordinates of corresponding screw fiducials, a 6-DoF transformation between the CT model and surface model is calculated and applied. The reconstruction error is defined as the distance between corresponding screw fiducials following surface-CT model coregistration using the screw fiducials.

  2. Registration error: Also previously described,20 this metric evaluates how accurately the surgeon can locate points of interest from CT imaging using the coregistered models. Corresponding anatomic features, or “natural landmarks,” are annotated in both the surface and CT models (green spheres in Figures 3B and C). The models are then aligned using a 6-DoF transformation computed using the coordinates of corresponding natural landmarks (Figure 3D). Following this surface-CT model coregistration, the screw fiducials are then treated as points of interest, and the registration error is calculated as the distance between corresponding screw fiducials (five screw fiducials for mastoid bone, eight screw fiducials for mastoidectomy).

    To assess the ability of the system to maintain navigational accuracy across time as dissection progresses and between switches in visualization modalities (exoscopy to endoscopy), the following two error metrics are described:

  3. Exoscope-exoscope surface-surface (S-S) registration error: This metric provides an assessment of the system’s accuracy in clinical conditions as exoscopic dissection progresses. Following coregistration of the mastoid bone surface model and CT model using natural landmarks as described above (Figures 3D and 4A), the subsequent mastoidectomy surface model is then coregistered to the initial mastoid bone surface model (Figure 4B) using a 6-DoF transform computed from the coordinates of corresponding landmarks present in both surface models (red spheres in Figure 4B). The result is indirect coregistration of the mastoidectomy surface model to the mastoid bone CT model (Figure 4B, right panel), using the initial (mastoid bone) surface model as a bridge. This is representative of the preservation of registration of iterative surface models during surgical dissection to the patient’s preoperative imaging. This is quantified with the exoscope-exoscope S-S registration error, defined as the distance between corresponding screw fiducials in the mastoid bone CT and exoscope mastoidectomy surface model spaces following indirect coregistration (light and dark blue spheres, respectively, in Figure 4B, right panel).

  4. Endoscope-exoscope S-S registration error: This metric provides an assessment of the system’s preservation of registration accuracy as the surgeon switches between exoscopy and endoscopy. Indirect coregistration of the exoscope mastoidectomy surface model to the mastoid bone CT model using the exoscope mastoid bone surface model as a bridge is first performed as described in the paragraph above (Figure 4B). Following this, the endoscope mastoidectomy surface model is coregistered to the exoscope mastoidectomy surface model (Figure 4C) using a 6-DoF transform computed from the coordinates of corresponding landmarks present in both surface models (red spheres in Figure 4C). The result is indirect coregistration of the endoscope mastoidectomy surface model to the mastoid bone CT model (Figure 4C, right panel), using two iterative exoscope surface models as bridges. This is quantified with the endoscope-exoscope S-S registration error, defined as the distance between corresponding screw fiducials in the mastoid bone CT and endoscope mastoidectomy surface model spaces following indirect coregistration (light and dark blue spheres, respectively, in Figure 3C, right panel).

Results

Mastoidectomy was performed on ten deceased donor temporal bones (five right and five left ears) with evaluation at two operative steps: exposure of the mastoid bone and completion of cortical mastoidectomy (Figures 2–4). High-resolution surface models were generated using stereoscopic exoscopy and endoscopy video of approximately 30 to 60 seconds duration. The reconstruction, registration, and S-S registration errors for each ear and operative step are given in Table 1. The overall average and standard deviation values for the reconstruction error were 0.72 ± 0.32 mm (mastoid bone: 0.60 ± 0.30 mm, mastoidectomy: 0.83 ± 0.31 mm); registration error was 1.43 ± 0.49 mm (mastoid bone: 1.64 ± 0.50 mm, mastoidectomy: 1.21 ± 0.39 mm); exoscope-exoscope S-S registration error was 1.72 ± 0.49 mm, and endoscope-exoscope S-S registration error was 1.49 ± 0.61 mm.

Table 1.

Surgical Navigation Errorsa

Overall Ear 1 Ear 2 Ear 3 Ear 4 Ear 5 Ear 6 Ear 7 Ear 8 Ear 9 Ear 10
Reconstruction error
 Mastoid bone 0.60 (0.30) 0.33 (0.14) 0.19 (0.06) 0.52 (0.33) 0.43 (0.14) 0.62 (0.51) 0.29 (0.09) 0.70 (0.13) 0.97 (0.35) 0.88 (0.44) 1.09 (0.50)
 Mastoidectomy 0.83 (0.31) 0.78 (0.73) 0.76 (0.37) 0.62 (0.50) 0.96 (0.92) 0.77 (0.41) 0.29 (0.21) 0.62 (0.24) 1.35 (0.84) 0.91 (0.39) 1.26 (0.51)
Registration error
 Mastoid bone 1.64 (0.50) 2.16 (0.94) 0.73 (0.42) 1.04 (0.48) 1.26 (0.48) 1.92 (0.74) 1.43 (0.86) 1.78 (0.91) 1.96 (0.59) 1.91 (0.64) 2.23 (0.33)
 Mastoidectomy 1.21 (0.39) 1.36 (0.64) 1.13 (1.03) 0.70 (0.54) 1.17 (0.94) 1.43 (0.43) 0.49 (0.32) 1.19 (0.37) 1.75 (1.09) 1.27 (0.32) 1.64 (0.53)
Exoscope-exoscope surface-surface registration error 1.72 (0.59) 1.76 (0.56) 0.96 (0.52) 0.96 (0.50) 1.97 (0.58) 1.94 (0.65) 1.34 (0.79) 1.48 (0.53) 2.88 (2.05) 1.62 (0.35) 2.26 (0.43)
Endoscope-exoscope surface-surface registration error 1.49 (0.61) 2.06 (0.62) 0.61 (0.13) 0.58 (0.19) 2.44 (0.81) 1.92 (0.61) 1.37 (0.60) 1.14 (0.48) 1.63 (0.43) 1.36 (0.30) 1.82 (0.31)
Open in a new tab
a

Mean (standard deviation) values in millimeters.

Discussion

Building upon prior work employing 3D-endoscopy, we present the initial evaluation of the compatibility of SLAM-based surgical navigation with exoscopic transmastoid surgery. To our knowledge, this is the first reported approach to surgical navigation that takes unique advantage of the digital stereoscopic capabilities of an exoscope, allowing for a fully integrated navigation system that does not require any additional equipment. Using deceased donor models and stereoscopic exoscopy video, the system generated highly accurate surface models of the operative field with a mean reconstruction error of 0.72 mm. The quality of surface reconstruction was comparable to that using 3D endoscopy (mean reconstruction errors of the mastoid bone of 0.60 mm vs 0.41 mm, and of the mastoidectomy cavity of 0.83 mm vs 0.78 mm, for exoscopy vs endoscopy, respectively) (H. Zhou and R. Bartholomew, unpublished data under review, 2025). After coregistration of the exoscopic surface models to volumetric CT models using natural landmarks, a high accuracy in localizing points of interest was demonstrated with a mean registration error of 1.43 mm. Accuracy is preserved as dissection progresses, with a comparable mean exoscope-exoscope S-S registration error of 1.72 mm. The preservation of accuracy is similarly maintained should the surgeon switch visualization modalities and employ an endoscope, with an also comparable mean endoscope-exoscope S-S registration error of 1.49 mm. The registration error of exoscopic SLAM-based navigation was somewhat higher compared to those for endoscopy (mean registration errors of the mastoid bone of 1.64 mm vs 0.46 mm, and of the mastoidectomy cavity of 1.21 mm vs 0.85 mm, for exoscopy vs endoscopy, respectively) (H. Zhou and R. Bartholomew, unpublished data under review, 2025).

Exoscopy and endoscopy have key differences, with important implications for SLAM-based surgical navigation. Acquisition angles of stereoscopic video are more limited with the use of an exoscope due to a more fixed line of sight in comparison to an endoscope, which is frequently in motion. This results in comparatively less visual data as input for surface model generation. This is reflected in somewhat increased errors compared to the corresponding endoscopy data as described above. Fortunately, accuracy remains acceptable for exoscopy and may be further improved with further development. Taking advantage of the 4K video acquisition capabilities of exoscopes in the future may permit higher surface reconstruction fidelity, although with an increased computational processing cost.

Another emerging technique for surface reconstruction with potential applications to surgical navigation is the use of Neural Radiance Fields (NeRF), a deep-learning computer vision algorithm that can generate a 3D scene from a series of 2D images. Within otolaryngology, NeRF and 2D endoscopic video feed have been used to perform accurate 3D endonasal reconstructions following functional endoscopic sinus surgery in deceased donors.33 However, NeRF uses 2D video input, which is intrinsically less data-rich than the 3D input used by SLAM. It remains to be demonstrated whether NeRF can generate high-fidelity surface reconstruction in the lateral skull base or when using microscopy or exoscopy, which provide less dynamic video input than endoscopy.

Strengths and Limitations

Clinical readiness of an exoscopic SLAM-based surgical navigation system requires additional developments. Rigorous validation in clinical encounters and additional surgical approaches (middle fossa craniotomy, retrosigmoid, etc.) is required. Segmentation of patient imaging into volumetric models and identification of landmarks for registration will also need to be performed in an automated manner, likely using machine learning methodologies.34–38 Clinical compatibility likewise will require the system to use established machine learning approaches39–47 to compensate for real-world “visual contaminants” such as instruments, bone dust, and irrigation fluid. Additionally, exoscope magnification can change regularly throughout an operation. In this study, magnification was kept fixed across stereoscopic video acquisition, and future studies will need to account for varying magnification.

Submillimetric accuracy has been described as a goal for surgical navigation in the lateral skull base,8,13,16–18,48,49 which we approached but did not achieve. Submillimetric accuracy in investigational systems has required either invasive bone-anchored fiducials and intraoperative CT scans7,11,14,50,51 or screw fixation of a patient tracker to the skull with complete bony exposure of the EAC and middle ear for registration.52 These investigational surgical navigation systems also used conventional optical or electromagnetic registration and tracking methodologies, with the aforementioned associated limitations.

If the above challenges can be addressed, an exoscopic SLAM-based surgical navigation system could improve the efficiency and efficacy of lateral skull base surgery in multiple ways. A surgeon could elect to use navigation without the burden of space-occupying navigation equipment nor placement of a head-fixed intraoperative localization device on the patient. Surgical economy of motion could be preserved as the surgeon can have navigation data tastefully overlayed onto the display of the operative field (Figure 2) instead of requiring them to change their gaze and visuospatial reference frame to review the data in 2D on a secondary monitor. Artificial intelligence-mediated recognition of surgical tools39–47 could allow instruments already in use to double as navigation probes and also permit issuance of proximity warnings53 when a drill nears a critical structure. A surgeon can also readily switch between exoscopy and endoscopy with seamless preservation of surgical navigation accuracy across modalities.

Conclusion

In this study, we established the feasibility of a surgical navigation system for use in the lateral skull base, which does not need external tracking equipment, requiring only the digital stereoscopic capabilities of an exoscope and SLAM-based algorithms. In our deceased donor model of trans-mastoid surgery, the system reconstructed high-fidelity 3D models of the operative field. Coregistration of these surface models with patient imaging enables localization of anatomic points of interest with sufficient accuracy, which is maintained throughout surgical dissection and between changes in visualization modalities (exoscopy to endoscopy). Further development and validation to ensure full compatibility with “real-world” conditions is necessary before clinical introduction.

Acknowledgments

We would like to express thanks to the deceased donors whose anatomic gifts allowed the conduct of this research, as well as the generous equipment contributions of Olympus America and KARL STORZ United States. The cartoon of the surgeon operating and the blank operating room video monitor, but not the content placed on the screen of the monitor, featured in Figure 1, was generated using the ChatGPT image generation tools.

Funding source:

Jeffrey P. Guenette is funded by the National Institute of Biomedical Imaging and Bioengineering, K08EB034299, and the Agency for Healthcare Research and Quality, R18HS029839. This study was supported by National Institutes of Health grant R00EB027177 (Haoyin Zhou); National Institutes of Health grant R01EB036996 (Haoyin Zhou); National Institutes of Health grant R01EB025964 (Jayender Jagadeesan); National Institutes of Health grant P41EB028741 (Jayender Jagadeesan, Haoyin Zhou); KARL STORZ United States Unrestricted Investigator Initiated Research Grant (C. Eduardo Corrales, Ryan A. Bartholomew); New England Otolaryngological Society Resident Research Grant: “Lateral skull base surgical navigation using stereoscopic surface reconstruction” (Ryan A. Bartholomew); and Brigham Research Institute Microgrant: “Intraoperative image guidance using stereoscopic surface reconstruction in lateral skull base surgery” (Ryan A. Bartholomew).

Footnotes

Competing interests: Jayender Jagadeesan owns equity in Navigation Sciences, Inc. He is a co-inventor of a navigation device to assist surgeons in tumor excision that is licensed to Navigation Sciences. His interests were reviewed and are managed by BWH and Partners HealthCare in accordance with their conflict-of-interest policies. C. Eduardo Corrales is a senior clinical and translational advisor for the CHORD clinical trial at Regeneron Pharmaceuticals and owns equity in the company. His interests were reviewed and are managed by BWH and Partners HealthCare in accordance with their conflict-of-interest policies.

References

  • 1.Krombach GA, van den Boom M, Di Martino E, et al. Computed tomography of the inner ear: size of anatomical structures in the normal temporal bone and in the temporal bone of patients with Menière’s disease. Eur Radiol. 2005; 15(8):1505–1513. [DOI] [PubMed] [Google Scholar]
  • 2.Bartholomew RA, Poe D, Dunn IF, Smith TR, Corrales CE. Iatrogenic inner ear dehiscence after lateral skull base surgery: therapeutic dilemma and treatment options. Otol Neurotol. 2019;40(4):e399–e404. [DOI] [PubMed] [Google Scholar]
  • 3.Betka J, Zvěřina E, Balogová Z, et al. Complications of microsurgery of vestibular schwannoma. BioMed Res Int. 2014;2014:1–10. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Obaid S, Nikolaidis I, Alzahrani M, Moumdjian R, Saliba I. Morbidity rate of the retrosigmoid versus translabyrinthine approach for vestibular schwannoma resection. J Audiol Otol. 2018;22(4):236–243. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Ben-Shlomo N, Rahimi A, Abunimer AM, et al. Inner ear breaches from vestibular schwannoma surgery: revisiting the incidence of otologic injury from retrosigmoid and middle cranial fossa approaches. Otol Neurotol. 2024;45(3):311–318. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Deep NL, Kay-Rivest E, Roland JT Jr.. Iatrogenic third window after retrosigmoid approach to a vestibular schwannoma managed with cochlear implantation. Otol Neurotol. 2021;42(9):1355–1359. [DOI] [PubMed] [Google Scholar]
  • 7.Schneider D, Anschuetz L, Mueller F, et al. Freehand stereotactic image-guidance tailored to neurotologic surgery. Front Surg. 2021;8:742112. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Labadie RF, Majdani O, Fitzpatrick JM. Image-guided technique in neurotology. Otolaryngol Clin North Am. 2007; 40(3):611–624, x. [DOI] [PubMed] [Google Scholar]
  • 9.Kral F, Puschban EJ, Riechelmann H, Freysinger W. Comparison of optical and electromagnetic tracking for navigated lateral skull base surgery. Int J Med Robot Comput Assist Surg. 2013;9(2):247–252. [DOI] [PubMed] [Google Scholar]
  • 10.Franz AM, Haidegger T, Birkfellner W, Cleary K, Peters TM, Maier-Hein L. Electromagnetic tracking in medicine—a review of technology, validation, and applications. IEEE Trans Med Imaging. 2014;33(8):1702–1725. [DOI] [PubMed] [Google Scholar]
  • 11.Kohan D, Jethanamest D. Image-guided surgical navigation in otology. Laryngoscope. 2012;122(10):2291–2299. [DOI] [PubMed] [Google Scholar]
  • 12.Barber SR. New navigation approaches for endoscopic lateral skull base surgery. Otolaryngol Clin North Am. 2021; 54(1):175–187. [DOI] [PubMed] [Google Scholar]
  • 13.Schneider D, Hermann J, Mueller F, et al. Evolution and stagnation of image guidance for surgery in the lateral skull: a systematic review 1989–2020. Front Surg. 2021;7:604362. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Komune N, Matsushima K, Matsuo S, Safavi-Abbasi S, Matsumoto N, Rhoton AL Jr.. The accuracy of an electromagnetic navigation system in lateral skull base approaches. Laryngoscope. 2017;127(2):450–459. [DOI] [PubMed] [Google Scholar]
  • 15.Ecke U, Maurer J, Boor S, Khan M, Mann WJ. Common errors of intraoperative navigation in lateral skull base surgery. HNO. 2003;51(5):386–393. [DOI] [PubMed] [Google Scholar]
  • 16.Labadie RF, Shah RJ, Harris SS, et al. Submillimetric target-registration error using a novel, non-invasive fiducial system for image-guided otologic surgery. Comput Aided Surg. 2004;9(4):145–153. [DOI] [PubMed] [Google Scholar]
  • 17.Labadie RF, Davis BM, Fitzpatrick JM. Image-guided surgery: what is the accuracy? Curr Opin Otolaryngol Head Neck Surg. 2005;13(1):27–31. [DOI] [PubMed] [Google Scholar]
  • 18.Schipper J, Aschendorff A, Arapakis I, et al. Navigation as a quality management tool in cochlear implant surgery. J Laryngol Otol. 2004;118(10):764–770. [DOI] [PubMed] [Google Scholar]
  • 19.Zhou H, Jagadeesan J. Real-time dense reconstruction of tissue surface from stereo optical video. IEEE Trans Med Imaging. 2020;39(2):400–412. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Bartholomew RA, Zhou H, Boreel M, et al. Surgical navigation in the anterior skull base using 3-dimensional endoscopy and surface reconstruction. JAMA Otolaryngol Head Neck Surg. 2024;150(4):318–326. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Ridge SE, Shetty KR, Lee DJ. Heads-up surgery. Otolaryngol Clin North Am. 2021;54(1):11–23. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Kapadiya M, Tarabichi M. An overview of endoscopic ear surgery in 2018. Laryngoscope Investig Otolaryngol. 2019; 4(3):365–373. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Smith S, Kozin ED, Kanumuri VV, et al. Initial experience with 3-dimensional exoscope-assisted transmastoid and lateral skull base surgery. Otolaryngol Head Neck Surg. 2019;160(2):364–367. [DOI] [PubMed] [Google Scholar]
  • 24.Mattogno PP, Della Pepa GM, Menna G, et al. Posterior cranial fossa surgery with a 3 dimensional exoscope: a single-center survey-based analysis and a literature review. World Neurosurg. 2024;189:e15–e26. [DOI] [PubMed] [Google Scholar]
  • 25.Begagic E, Pugonja R, Beculic H, et al. The new era of spinal surgery: exploring the use of exoscopes as a viable alternative to operative microscopes—a systematic review and meta-analysis. World Neurosurg. 2024;182:144–158 e141. [DOI] [PubMed] [Google Scholar]
  • 26.Garcia JP, Avila FR, Torres RA, et al. Evaluating the exoscope as an alternative to the operating microscope in plastic surgery. J Plast Reconstr Aesthet Surg. 2023;85:376–386. [DOI] [PubMed] [Google Scholar]
  • 27.Ricciardi L, Chaichana KL, Cardia A, et al. The exoscope in neurosurgery: an innovative “point of view”. a systematic review of the technical, surgical, and educational aspects. World Neurosurg. 2019;124:136–144. [DOI] [PubMed] [Google Scholar]
  • 28.Chiang H, Ledbetter L, Kaylie DM. Systematic review of otologic and neurotologic surgery using the 3-dimensional exoscope. Otol Neurotol Open. 2022;2(4):e024. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Tu NC, Doerfer K, Costeloe A, Sioshansi PC, Babu S. Educational benefit of the three-dimensional exoscope versus operating microscope in otologic surgery. Otol Neurotol. 2024;45(2):150–153. [DOI] [PubMed] [Google Scholar]
  • 30.Di Bari M, Colombo G. Exoscope-assisted surgery in otology and neurotology. Curr Opin Otolaryngol Head Neck Surg. 2024;32(5):301–305. [DOI] [PubMed] [Google Scholar]
  • 31.Fedorov A, Beichel R, Kalpathy-Cramer J, et al. 3D Slicer as an image computing platform for the Quantitative Imaging Network. Magn Reson Imaging. 2012;30(9):1323–1341. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Arun KS, Huang TS, Blostein SD. Least-squares fitting of two 3-D point sets. IEEE Trans Pattern Anal Mach Intell. 1987;PAMI-9(5):698–700. [DOI] [PubMed] [Google Scholar]
  • 33.Ruthberg JS, Bly R, Gunderson N, et al. Neural Radiance Fields (NeRF) for 3D reconstruction of monocular endoscopic video in sinus surgery. Otolaryngol Head Neck Surg. 2025;172(4):1435–1441. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34.Neves CA, Tran ED, Kessler IM, Blevins NH. Fully automated preoperative segmentation of temporal bone structures from clinical CT scans. Sci Rep. 2021;11(1):116. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Zhou L, Li Z. Automatic multi-label temporal bone computed tomography segmentation with deep learning. Int J Med Robot Comput Assist Surg. 2023;19(5):e2536. [DOI] [PubMed] [Google Scholar]
  • 36.Wang J, Lv Y, Wang J, et al. Fully automated segmentation in temporal bone CT with neural network: a preliminary assessment study. BMC Med Imaging. 2021;21:166. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.Ding AS, Lu A, Li Z, et al. A self-configuring deep learning network for segmentation of temporal bone anatomy in cone-beam CT imaging. Otolaryngol Head Neck Surg. 2023; 169(4):988–998. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Sahu M, Xiao Y, Porras JL, et al. A label-efficient framework for automated sinonasal CT segmentation in image-guided surgery. Otolaryngol Head Neck Surg. 2024; 171(4):1217–1225. [DOI] [PubMed] [Google Scholar]
  • 39.Zhang J, Gao X. Object extraction via deep learning-based marker-free tracking framework of surgical instruments for laparoscope-holder robots. Int J Comput Assist Radiol Surg. 2020;15(8):1335–1345. [DOI] [PubMed] [Google Scholar]
  • 40.Ruzicki J, Holden M, Cheon S, Ungi T, Egan R, Law C. Use of machine learning to assess cataract surgery skill level with tool detection. Ophthalmol Sci. 2023;3(1):100235. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41.Kamrul Hasan SM, Linte CA. U-NetPlus: a modified encoder-decoder U-Net architecture for semantic and instance segmentation of surgical instruments from laparoscopic images. Annu Int Conf IEEE Eng Med Biol Soc. 2019;2019:7205–7211. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42.Bamba Y, Ogawa S, Itabashi M, Kameoka S, Okamoto T, Yamamoto M. Automated recognition of objects and types of forceps in surgical images using deep learning. Sci Rep. 2021;11(1):22571. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 43.Matton N, Qalieh A, Zhang Y, et al. Analysis of cataract surgery instrument identification performance of convolutional and recurrent neural network ensembles leveraging BigCat. Transl Vis Sci Technol. 2022;11(4):1. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 44.Yeh HH, Jain AM, Fox O, Sebov K, Wang SY. PhacoTrainer: deep learning for cataract surgical videos to track surgical tools. Transl Vis Sci Technol. 2023;12(3):23. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 45.Mikada T, Kanno T, Kawase T, Miyazaki T, Kawashima K. Three-dimensional posture estimation of robot forceps using endoscope with convolutional neural network. Int J Med Robot Comput Assist Surg. 2020;16(2):e2062. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 46.Bareum C, Kyungmin J, Songe C, Jaesoon C. Surgical-tools detection based on Convolutional Neural Network in laparoscopic robot-assisted surgery. Annu Int Conf IEEE Eng Med Biol Soc. 2017;2017:1756–1759. [DOI] [PubMed] [Google Scholar]
  • 47.Li Z, Shu H, Liang R, et al. TAToo: vision-based joint tracking of anatomy and tool for skull-base surgery. Int J Comput Assist Radiol Surg. 2023;18(7):1303–1310. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 48.Williamson T, Gavaghan K, Gerber N, et al. Population statistics approach for safety assessment in robotic cochlear implantation. Otol Neurotol. 2017;38(5):759–764. [DOI] [PubMed] [Google Scholar]
  • 49.Chen JX, Yu SE, Ding AS, et al. Augmented reality in otology/neurotology: a scoping review with implications for practice and education. Laryngoscope. 2023;133(8):1786–1795. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 50.Rathgeb C, Anschuetz L, Schneider D, et al. Accuracy and feasibility of a dedicated image guidance solution for endoscopic lateral skull base surgery. Eur Arch Otrhinolaryngol. 2018;275(4):905–911. [DOI] [PubMed] [Google Scholar]
  • 51.Schwam ZG, Kaul VZ, Cosetti MK, Wanna GB. Accuracy of a modern intraoperative navigation system for temporal bone surgery in a cadaveric model. Otolaryngol Head Neck Surg. 2019;161(5):842–845. [DOI] [PubMed] [Google Scholar]
  • 52.Schneider D, Hermann J, Gerber KA, et al. Noninvasive registration strategies and advanced image guidance technology for submillimeter surgical navigation accuracy in the lateral skull base. Otol Neurotol. 2018;39(10):1326–1335. [DOI] [PubMed] [Google Scholar]
  • 53.Citardi MJ, Agbetoba A, Bigcas JL, Luong A. Augmented reality for endoscopic sinus surgery with surgical navigation: a cadaver study. Int Forum Allergy Rhinol. 2016;6(5):523–528. [DOI] [PMC free article] [PubMed] [Google Scholar]

RESOURCES