Utilizing 3D Slicer for pulmonary bronchovascular anatomy reconstruction: a practical workflow and case examples

Introduction

Background

The evolution of imaging technologies has transformed the way operations are planned and executed across all surgical specialties, offering surgeons unprecedented visual clarity and anatomical precision. This is particularly relevant to thoracic surgery, where visualization of the pulmonary vasculature and tracheobronchial tree is essential in achieving oncologic resection (1). Anatomical lung resections for pulmonary malignancies must ligate the feeding vessels and bronchial tree in order to obtain adequate margins and prevent future recurrence. Furthermore, the lung has multiple anatomical variations, and the path of vessels often differs greatly in individual patients (2). To avoid the intraoperative complications of ligating the wrong structures, detailed preoperative planning with review of the computed tomography (CT) images is a key aspect of preparation for surgery. Cutting the wrong vessels can lead to lobar congestion, ischemia, and inadequate tumour margins, which predisposes patients to a higher chance of local tumour recurrence.

Rationale and knowledge gap

Traditionally, thoracic surgical planning has relied heavily on two-dimensional (2D) imaging modalities. These methods are inherently limited in representing the true three-dimensional (3D) relationships between bronchovascular structures. Accurate appreciation of these spatial relationships is crucial for precise surgical planning and aids the conduct of surgery, yet 2D imaging often fails to capture this complexity. This limitation underscores the growing need for approaches that bridge the gap between conventional imaging and its surgical application.

To achieve reliable and clinically meaningful 3D reconstruction, accessible and user-friendly tools are essential. 3D Slicer, a novel open-source software platform, has emerged as a versatile and powerful option for clinicians and researchers (3). It enables high-resolution segmentation and reconstruction of pulmonary bronchovasculature and tumours from the conventional 2D CT datasets. Its intuitive interface allows users to interact with complex 3D models dynamically, facilitating detailed anatomical analysis and more informed surgical decision-making.

Open-source software such as 3D Slicer have been used in prior studies to generate 3D lung reconstruction, with reports demonstrating that these models are clinically usable and comparable to commercially available systems (4). However, the existing literature using 3D Slicer largely falls into two categories. First, studies that focus primarily on clinical feasibility or postoperative outcomes, with limited methodological detail regarding how the 3D reconstructions were generated (4,5). Second, reports that describe the segmentation methodology in isolation, without integration into surgical planning, discussion of factors influencing model quality and workflow efficiency, or guidance on troubleshooting common challenges and encountered during reconstruction (6). As a result, while feasibility has been established and workflows do exist, they may not be readily reproducible or efficient enough for routine clinical use, and there remains a lack of a clearly described end-to-end workflow that can be consistently adopted into routine thoracic surgical practice.

Objective

In this manuscript, we aim to address the gaps by presenting a practical, reproducible, and efficient workflow for generating patient-specific 3D lung models using 3D Slicer, specifically tailored to thoracic surgical planning. We describe our step-by-step approach to segmenting pulmonary arteries, veins, bronchi, lung parenchyma, and tumours, as well as delineating surgical resection margins.

In addition to outlining the technical steps, we highlight common pitfalls encountered during the reconstruction process, discuss factors that influence efficiency and model quality, and provide strategies to overcome these challenges. Two representative case examples are presented to demonstrate how this workflow can clarify ambiguous anatomy on conventional CT imaging and directly inform operative planning.

Through this imaging-based approach, we aim to show how 3D Slicer can enhance surgical planning, improve anatomical understanding, and ultimately contribute to safer, more efficient, and more personalized thoracic surgery. We present this article in accordance with the SUPER reporting checklist (available at https://jtd.amegroups.com/article/view/10.21037/jtd-2025-1-2492/rc).

Preoperative preparations and requirements

Ethical statement

The study was conducted by the Department of Cardiothoracic Surgery at University Hospital Geelong, a tertiary regional hospital in Victoria, Australia that forms part of Barwon Health. All procedures performed in this study were in accordance with the Declaration of Helsinki and its subsequent amendments. This study was approved by the institutional ethics committee of Barwon Health (No. 23.178) and individual consent, including consent for the publication of deidentified data such as imaging and operative video material, was waived. The waiver was granted on the basis that the study posed negligible risk and involved the use of data that was acquired and accessed as part of routine clinical care by members of the clinical team who had legitimate access to this information as part of their standard clinical duties.

Software description

3D Slicer is an open-source desktop platform for medical imaging visualization, processing, segmentation, and analysis. Widely used in research and clinical applications, it provides a user-friendly interface accessible to users of varying experience levels. The software supports multiple imaging modalities, including CT, and is freely available for Windows, macOS, and Linux (3). These features make 3D Slicer an ideal tool for reconstructing intricate anatomical structures from CT datasets. Importantly, the software does not require formal training to use; most functions can be learned intuitively. Familiarity with the relevant anatomy and basic imaging concepts greatly enhances the accuracy and efficiency of segmentation. As such, this workflow can be implemented by users at all levels.

Image preprocessing

Contrast-enhanced CT chest scans were obtained in Digital Imaging and Communications in Medicine (DICOM) format with a slice thickness of 1.0 mm. The images were transferred to a mac-based workstation and imported into 3D Slicer.

Modules and tools

Segment editor: used to create and modify 2D and 3D segmentations. Key tools and their configurations are summarized in Table 1 (7).

Table 1

Description of the ‘Segment editor’ module tools utilized to segment pulmonary bronchovascular anatomy (7)

Icon	Name	Utilization	Configuration
	Paint	Utilized to paint in areas in the CT images and the 3D screen, thereby adding them to the segmentation	Activate the “Sphere brush” and “Edit in 3D” functions. Adjust the “diameter” of the brush to your preference
	Erase	Utilized to erase areas in the CT images and the 3D screen, thereby removing them from the segmentation
	Grow from seeds	Utilized to grow segments to complete segmentation based on location, size, and shape of the initial segments	“Editable area” should be set to the appropriate template layer
	Smoothing	Utilized to fill in small holes in the segment	The “Smoothing” method be set to the “Closing” (fill holes) setting
	Islands	Utilized to isolate individual segments (‘Islands’) when multiple unconnected segments are present	Must be set to “Keep selected island” setting
	Threshold	Utilized to fill segments based on source volume intensity ranges	The threshold parameters should be adjusted to isolate the region of interest based on intensity ranges
	Scissors	Utilized to draw free-hand or shapes to select areas that are erased from or added to the segmentation	Must be set to “Erase inside” and “Free form” settings
	Logical operators	Utilized to apply logical operators to segments, such as copy, add, subtract, and intersect	The target layer should be selected in the first box and the source layer should be selected in the second box. Ensure to set the required operation
	Margin	Utilized to set desired margins	Must be set to “Grow” and the desired margin should be entered

3D, three-dimensional; CT, computed tomography.

Markups: used to generate geometric nodes such as lines, planes, and curves for defining tumour dimensions and margins. Tools are detailed in Table 2 (8).

Table 2

Description of the ‘Markups’ module tools utilized to illustrate tumour margins (8)

Icon	Name	Utilization
	Line	Utilize to measure tumour dimensions
	Plane	Utilized to indicate the boundary of the tumour margins
	Closed curve	Utilized to illustrate tumour margins

LungCTSegmenter: an extension available for download within 3D Slicer. This automated module performs threshold-based region-growing segmentation optimized for thoracic CT imaging.

Step-by-step description

Setup

After data import, all the required layers should be created within the “Segment editor” and assigned a distinct colour and label. Ensure that all necessary modules are installed.

Tracheobronchial segmentation

The tracheobronchial tree can be segmented using the “Threshold” tool to isolate the trachea and bronchi from the surrounding tissue based on voxel intensity (Hounsfield units). Threshold values should be adjusted to include the trachea and bronchial tree whilst excluding adjacent soft tissue (range, −1,000 to −500). Once optimal parameters are identified, segmentation can be applied. Residual unwanted regions can be removed using the “Islands” tool, and small discontinuities within the walls of the tracheobronchial tree can be corrected using the “Smoothing” tool (Figure 1A-1C).

Figure 1 Tracheobronchial & lung segmentation. (A1-A4) The initial threshold outcome for the tracheobronchial tree. (B1-B4) The isolated tracheobronchial tree with numerous small discontinuities. (C1-C4) The finalised tracheobronchial tree once all the discontinuities are filled. (D1-D4) The finalised lung model generated by the “LungCTSegmenter” module.

Lung segmentation

Lung parenchyma can be segmented manually using the same tools described in the previous step, which may take 5–10 minutes depending on user proficiency. Alternatively, the “LungCTSegmenter” module can be used to automate the step. The algorithm automatically identifies lung boundaries and excludes the extrapulmonary tracheobronchial tree and mediastinal structures (Figure 1D). This automated approach significantly reduces manual workload, providing an accurate foundation for further segmentation of intrapulmonary vessels and lesions within 1–2 minutes.

Vessel segmentation

A “vessel mask” layer is used as a template for pulmonary vessel isolation. The “Threshold” and “Islands” tools can be applied to extract the pulmonary vessels based on voxel intensity (range, 500 to 1,000). Optimal results are achieved when using contrast-enhanced CT datasets, particularly when contrast is timed to opacify the pulmonary arteries (e.g., CT pulmonary angiograms). Delayed scans can still be used but typically provide less distinct vessel boundaries, while non-contrast imaging poses the greatest challenge due to poor vessel-to-tissue differentiation.

In cases where the pulmonary vasculature cannot be optimally isolated through the above, the “Erase” or “Scissors” tools can be used to manually remove unwanted tissue, such as the aorta or cardiac chambers (Figure 2).

Figure 2 Vessel segmentation. (A1-A4) The initial threshold outcome for the pulmonary vessels. (B1-B4) The isolated vessel template utilized to extract the pulmonary vessels. (C1-C4) The finalised combined pulmonary vessel model.

Differentiation of the isolated vessels into arteries and veins can be achieved using a seed-based approach. Using the “Paint” tool, small sections of the proximal and distal portions of the lobar branches should be manually labelled within the respective “arteries” and “veins” layers to indicate vessel identity. These labelled regions serve as seeds for automated classification. Once sufficient seeds are defined, the “Grow from seeds” tool can be applied to generate complete arterial and venous segmentations (Figure 3A-3C).

Figure 3 Vessel differentiation. (A1-A4) The pulmonary vessel template generated in the previous step. (B1-B4) The pulmonary artery (red) and pulmonary vein (blue) ‘seeds’ populated into the pulmonary vessel template. (C1-C4) The differentiated pulmonary artery (red) and pulmonary vein (blue) models. (D1-D4) The distal pulmonary branches not captured in the initial vessel segmentation. (E1-E4) The finalized pulmonary artery (red) and vein (blue) models after expanding them to include distal branches.

Vessel expansion

Depending on the intended purpose, it may be advantageous to extend the pulmonary vessels to include segmental and subsegmental branches. This provides greater anatomical detail, enabling assessment of segmental anatomy and evaluation of resection feasibility for procedures such as segmentectomies. The extent of distal branching required will directly influence the time needed for this step, as incorporating smaller branches increases the likelihood of capturing unwanted surrounding tissue, which may require an additional 5–10 minutes of manual refinement.

The previously segmented lung parenchyma can be utilized to restrict the editable region, ensuring segmentation tools operate exclusively within the intrapulmonary space and minimizing interference from surrounding tissues.

To generate the distal branches, an additional layer—“distal branches”—should be created. The “Threshold” tool can be applied, with the parameters adjusted to isolate the intrapulmonary vasculature. Once an optimal threshold is achieved, the “Islands” tool can be used to remove any unwanted residual tissue. Finally, the “Grow from seeds” tool can be applied to expand the existing pulmonary artery and vein segmentations into their respective segmental branches, completing the peripheral vascular model (Figure 3D,3E).

Tumour segmentation

Similar to the previous steps, pulmonary tumours can also be isolated using a combination of all the tools, primarily the “Threshold” and “Islands” tools, with manual refinement where necessary.

Tumour dimensions can be directly measured on the CT images using the “Line” tool in the “Markups” module to define length, width, and height. These measurements appear in both 2D and 3D views for reference. Next, planes should be created and resized to encompass the tumour and desired margin in each orthogonal axis. For instance, a tumour measuring 20 mm × 15 mm × 10 mm with a 2 cm margin would yield planes approximately 60 mm × 55 mm, 55 mm × 50 mm, and 60 mm × 50 mm along the corresponding axes. Circular margins can then be delineated using the “Closed curve” tool by connecting the endpoints of each of the three planes.

Alternatively, the “tumour” layer could be duplicated, and the “Margin” tool applied to automatically expand the segmentation by a specified distance—effectively generating a 3D margin around the tumour corresponding to the desired resection boundary (Figure 4).

Figure 4 Tumour segmentation & margins. (A1-A4) A left upper lobe pulmonary nodule. (B1,B2) The dimensions markers (width & height) drawn onto the pulmonary nodule. (C1,C2) The plane (axial) utilized to set margin size. (D1,D2) The circular margin drawn around the pulmonary nodule.

Postoperative considerations and tasks

As this workflow is designed to support pre-operative anatomical planning rather than alter intraoperative technique, it does not introduce specific postoperative management requirements. In the cases presented, postoperative care followed standard institutional pathways for lung resection.

From a pathological perspective, the 3D models were used pre-operatively to assess the feasibility of achieving adequate resection margins and to inform surgical planning. Final margin assessment was performed using routine histopathological methods, which demonstrated adequate oncological margins.

Formal comparison of postoperative outcomes and pathological margins between cases planned with 3D reconstruction and those planned solely using conventional imaging represents important directions for future research.

Tips and pearls

Challenges and troubleshooting

Several practical challenges may arise during the segmentation process, particularly when working with poor-quality or non-contrast enhanced datasets. These challenges are more pronounced in anatomically complex cases, including those with densely calcified hilar lymphadenopathy or distortion of bronchovascular anatomy due to prior inflammatory or surgery.

Threshold-based segmentation greatly improves efficiency, but it can also select unwanted adjacent tissues. The most common issue is inadvertent inclusion of soft tissue abutting the airways or mediastinal tissue around the pulmonary vessels. This issue is more common in the presence of calcification or distorted anatomy. In these situations, gradually narrowing the threshold range is the most effective way to confine the segmentation to the intended structures. The “Islands” tool can then be used to fully isolate the airway or vascular structures from residual tissue, provided the segmented structures are not physically connected. While this may reduce the detail captured by the model, the structures can be expanded after their initial isolation in a process similar to extending the pulmonary vessels.

For the pulmonary vessels, vessel differentiation may be challenging when arterial and venous branches lie in close proximity or when contrast timing is suboptimal. In cases where contrast is timed to the pulmonary arteries, the arteries can often be segmented using thresholding alone, with venous structures subsequently isolated by subtracting the arterial tree using the “Logical operators” tool. Otherwise, a seed-based approach is required, which can be optimized by adding more detailed seed regions to both proximal and distal branches. This is particularly important when arterial and venous branches are in direct contact.

Any remaining difficulties can be addressed through manual refinement using the “Paint”, “Erase”, and “Scissors” tools, which can be applied on the CT slices or directly onto the 3D model. These tools allow tissue to be added or removed with precision, though their use can be time-consuming. The workflow can be further optimized by adjusting layer visibility as needed and by cross-checking against the CT imaging to ensure anatomical accuracy throughout the process.

Overall, the workflow is applicable across all anatomical scenarios, including complex, calcified, or distorted anatomy, although these cases typically require greater manual input. Image quality and anatomical complexity are therefore key determinants of overall reconstruction time.

In practice, the overall workflow typically requires 20–60 minutes with pulmonary vessel differentiation representing the most time-consuming step. Image-related parameters are the primary determinant of efficiency. High-quality, thin-slice, contrast-enhanced CT datasets—particularly those with contrast optimally timed to opacify the pulmonary arteries—substantially improve threshold-based segmentation. This reduces the need for manual refinement and allows for rapid and accurate the vessel differentiation. In contrast, delayed-phase studies require additional manual input, while non-contrast studies are the least efficient, as poor vessel-to-parenchyma contrast limits automated segmentation and increases reliability on manual tools.

The required level of anatomical detail also influences workflow duration. Basic reconstruction of lobal anatomy is quicker than detailed segmental or subsegmental modelling, which usually requires additional refinement steps. That said, when image quality and contrast timing are optimized, even anatomically detailed reconstructions can be generated efficiently using a combination of thresholding and seed-based segmentation, with minimal manual correction.

User proficiency represents another major factor. During initial use, cases may take 45–60 minutes as users become familiar with the software interface and its tools. In our experience, proficiency was achieved within approximately five cases, without formal guidance or an established workflow at the time. As proficiency relates primarily to software operation rather than interpretation of surgical anatomy or operative planning, a basic understanding of pulmonary anatomy—even at a medical student level—is sufficient to perform the reconstruction, as we observed in practice. With increasing familiarity, reconstruction time can be reduced to 20 minutes for optimized contrast-enhanced datasets, 20–30 minutes for delayed contrast studies, and 30–45 minutes for non-contrast studies or studies with highly complex anatomy. Proficiency could be achieved more rapidly using a structured workflow such as the one described in this manuscript, supported by accompanying step-by-step instructional videos, which use illustrative cases for educational purposes (Videos 1-11).

Use of 3D model in pre-operative and intra-operative setting

The 3D reconstructions were used primarily during the pre-operative planning phase to review bronchovascular anatomy, tumour location and segmental relationships, allowing surgeons to perform mental simulation of the planned resection and adjust their operative strategy where indicated. In selected cases, the model was also available intra-operatively on a monitor for reference if required. The model was not used for real-time navigation or continuous intra-operative guidance and did not alter standard operative workflow, serving instead as an adjunct to anatomical knowledge and intraoperative judgement.

Case 1

A 65-year-old female with a left upper lobe pulmonary tumour underwent contrast-enhanced CT imaging for operative planning. The 17 mm tumour was then reconstructed using 3D Slicer from a contrast-enhanced CT dataset acquired with a slice thickness of 1.0 mm with standard systemic contrast timing. Generation of the complete 3D model required 32 minutes and 31 seconds, incorporating a planned 20 mm surgical margin to determine whether a segment 3 (S3) segmentectomy would achieve clear surgical margins.

Once it was confirmed that clear margins could be obtained by performing a S3 segmentectomy, the reconstructed model was then used to clarify the bronchovascular anatomy in detail. When assessing the branches of the S3 artery (A3), it was noted that the small first branch arising from the large first left pulmonary artery branch was in fact a segment 1 (S1) artery (A1) branch that will need to be preserved during the operation (Figure 5).

Figure 5 Model generated for case 1 (left lateral view). (A,B) Demonstrate the location of the tumour within S3 and the proximity of the A1 to the intersegmental plane. (C,D) Demonstrate the spatial relationship between the tumour and the pulmonary bronchovasculature. The pulmonary artery is depicted in red, pulmonary vein in blue, bronchi in green, and tumour in yellow with surrounding margins. A1, segment 1 artery; A3, segment 3 artery; S1, segment 1; S2, segment 2; S3, segment 3; S4, segment 4; S5, segment 5.

The model also demonstrated that the intersegmental plane between S3 and S1 lay very close to the desired tumour margin. As a result, extending the parenchymal division an additional 10–15 mm into S1 at the time of resection was deemed necessary to ensure adequate oncological margins. Intraoperatively, the bronchovascular anatomy and arterial branching pattern closely matched the reconstructed model, and the anticipated intersegmental relationship was confirmed during dissection.

The patient underwent an uneventful robotic-assisted left S3 extended pulmonary segmentectomy and was discharged home on the fourth post-operative day.

Case 2

A 70-year-old female with an incidental right pulmonary nodule underwent contrast-enhanced CT imaging for pre-operative assessment of the tumour. The lesion was located near the right pulmonary hilum and measured 15 mm in maximal diameter. The horizontal fissure was absent, making it impossible to determine whether the tumour was within the right upper or middle lobe.

The case was reconstructed using 3D Slicer from a contrast-enhanced CT dataset acquired with a slice thickness of 1.0 mm and pulmonary arterial phase contrast. Generation of the complete model required 19 minutes and 47 seconds. The reconstruction allowed detailed visualization of the tumour and its spatial relationship to the surrounding bronchovasculature despite the incomplete fissural anatomy (Figure 6).

Figure 6 Model generated for case 2 (right lateral view). (A,B) Demonstrate the location of the tumour within the right middle lobe and the extension of the surgical margins into S3. (C,D) Demonstrate the spatial relationship between the tumour and the pulmonary bronchovasculature, clearly showing that the nodule is supplied by the A4 and A5, drained by the V4 and V5 and associated with B4 and B5. The pulmonary artery is depicted in red, pulmonary vein in blue, bronchi in green, and tumour in yellow with surrounding margins. A1, segment 1 artery; A2, segment 2 artery; A3, segment 3 artery; A4, segment 4 artery; A5, segment 5 artery; B4, segment 4 bronchus; B5, segment 5 bronchus; S1, segment 1; S2, segment 2; S3, segment 3; S4, segment 4; S5, segment 5; V4, segment 4 vein; V5, segment 5 vein.

After review of the model by the primary surgeon, it was deemed that the lesion was supplied by the arteries of segments 4 and 5 (A4 and A5), drained by the veins of segments 4 and 5 (V4 and V5) and associated with the bronchi of segments 4 and 5 (B4 and B5), confirming that the tumour was within the middle lobe. It was appreciated that the superior lung parenchymal resection margin had to be extended into S3 of the upper lobe in order to obtain adequate tumour margins. These anatomical relationships were confirmed intraoperatively, with the tumour being supplied by the middle lobe bronchovascular structures as predicted by the 3D model.

The patient underwent robotic-assisted extended middle lobectomy. The tumour was resected with >2 cm margins and division of middle lobe vessels and bronchi. The patient made an uneventful recovery and was discharged home on the third post-operative day.

Discussion

Conventional medical imaging has improved modern surgery; however, its 2D nature continues to limit accurate appreciation of complex anatomical relationships. This is particularly evident in thoracic surgery, where the intricate and variable pulmonary bronchovascular anatomy is often difficult to visualize clearly on standard axial CT imaging alone. As thoracic surgical techniques evolve towards increasingly precise segmental and subsegmental resections, there is a growing need for tools that can represent patient-specific anatomy in 3D to better support surgical planning and execution.

There has been renewed interest in anatomical pulmonary segmentectomy for selected small primary lung cancers, following large, randomized trials showing comparable oncological outcomes to lobectomies with improved preservation of lung function (9,10). In such procedures, precise understanding of segmental anatomy and accurate assessment of resection margins are critical to ensure both adequate resection margins and technical safety. 3D reconstruction software provides a practical bridge between conventional 2D imaging and more advanced forms of surgical visualization by addressing these requirements. Retrospective studies suggest that 3D reconstructions reduce operative time, intraoperative blood loss, and conversion rates compared to conventional imaging alone, particularly in anatomically complex cases (11). Reviews further suggest that detailed 3D models improve tumour localization and enhance anatomical visualization, thereby enhancing surgical safety and precision (12).

Many commercial software programs that semi-automate the 3D reconstruction process exist, but they are either not fit for the purpose, cumbersome to use for modelling lung anatomy or are limited by high software license or use costs (13,14). Open-source software, such as 3D Slicer, do not suffer from these limitations and provide comparable outputs to the proprietary systems (4). While studies have demonstrated the use of 3D Slicer in the realm of thoracic surgery, they are primarily focused on feasibility, clinical outcomes, or general usability, rather than providing a structured, end-to-end technical workflow that clinician can reliably replicate in routine practice (4,5). Reports that do describe their workflows are usually very limited, less efficient due to more reliability on manual refinement processes, and do not discuss key parameters influencing workflow efficiency nor provide strategies to overcome common pitfalls (6). The primary contribution of this manuscript lies in presenting a practical, reproducible, and clinician-focused workflow for generating patient-specific 3D reconstructions of pulmonary bronchovascular anatomy using 3D Slicer. This includes explicit methods of artery-vein differentiation, tumour margin delineation, discussion of factors affecting quality and proficiency, and common pitfalls. The included real-world case examples illustrate how this structured approach can resolve clinically relevant ambiguity on conventional imaging and directly inform surgical planning, distinguishing this work from previous reports that described isolated components or broad software feasibility without comprehensive methodological detail.

There are several limitations to this technique. First, the reliability and efficiency of the workflow are highly dependent on the quality of the CT dataset. Thin-slice datasets, preferably 1 mm or less, with adequate contrast enhancement timed to opacify the pulmonary arteries, will generate the most accurate and reproducible 3D reconstructions. Suboptimal contrast timing or insufficient differentiation between arteries and venous phases reduces the effectiveness of threshold-based segmentation and increases reliance on manual refinement, significantly prolonging reconstruction time and introducing greater operator dependency. Second, real-world use by co-authors would suggest that very small vascular branches—particularly those <1.5 mm in diameter—may not be captured in the 3D reconstruction, most commonly when the CT image quality or contrast resolution is suboptimal. This reflects the known spatial resolution limitations of CT imaging rather than the segmentation workflow itself (12). These tiny pulmonary arterial branches are especially problematic during fissure dissection, where they are prone to avulsion and bleeding. In practice, injury to vessels of this size is typically minor and can be readily managed using standard surgical techniques. In particular, small pulmonary arterials branch injuries are usually easily controlled with sustained compression, without clinical consequence. Importantly, the inability to visualize these vessels does not undermine the value of the model for surgical planning, as its primary purpose is to delineate major lobar and segmental bronchovascular anatomy rather than to identify all peripheral microvasculature. Surgeons should therefore continue to anticipate the presence of additional minor branches during dissection and manage them with meticulous surgical technique and standard hemostatic principles as is routine practice, using the 3D model as an adjunct to, rather than a substitute for, intraoperative judgement.

Finally, a major limitation of 3D imaging in thoracic surgery is the dynamic deformation of the lung intraoperatively. Once the lung is deflated and mobilized in multiple directions to access the hilum, discrepancies naturally arise between preoperative models and the operative anatomy. Although 3D reconstruction significantly enhances preoperative–intraoperative correlation, it cannot fully eliminate divergence caused by lung collapse and manipulation. Nevertheless, 3D reconstruction represents an important foundational step toward mixed-reality guidance systems, in which bronchovascular anatomy may eventually be superimposed onto the surgical field to facilitate precise and accurate dissection.

Integrating 3D reconstruction into routine clinical practice has the potential to improve both pre-operative and intra-operative decision making. Despite the inherent limitations of imaging resolution and intraoperative anatomical deformation, the ability to visualize patient-specific lobar and segmental anatomy pre-operatively provides clinically meaningful information that is not readily appreciated on conventional 2D imaging. The use of this workflow can assist in both patient selection and surgical planning by enabling accurate visualization of the tumour in relation to bronchovascular structures, as well as intersegmental and fissural planes. By providing surgeons with a clearer understanding of major bronchovascular relationships and anatomical variations, it can lead to safer and more tailored resections. A further advantage of this approach is that it is entirely cost-free due to the open-source nature of 3D Slicer. This enables centres to incorporate the 3D planning into their workflow without any financial barrier.

Beyond thoracic surgery, this approach could also be extended to other specialties, such as hepatic and renal surgery, where complex anatomy and individual variation has the potential to influence surgical outcomes. We hope our paper will allow clinicians to incorporate 3D planning into their surgical workflow and thus improve patient outcomes and operative safety.

Conclusions

3D Slicer provides an accessible and effective platform for reconstructing pulmonary bronchovascular anatomy from CT imaging. The workflow presented here offers an efficient and reproducible approach to reconstructing pulmonary anatomy and demonstrates its value in enhancing anatomical understanding and aiding surgical planning, particularly in anatomically challenging or ambiguous cases. Continued refinement and integration of 3D reconstruction techniques have the potential to further improve surgical precision and patient outcomes.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the SUPER reporting checklist. Available at https://jtd.amegroups.com/article/view/10.21037/jtd-2025-1-2492/rc

Peer Review File: Available at https://jtd.amegroups.com/article/view/10.21037/jtd-2025-1-2492/prf

Funding: None.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://jtd.amegroups.com/article/view/10.21037/jtd-2025-1-2492/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. All procedures performed in this study were in accordance with the Declaration of Helsinki and its subsequent amendments. This study was approved by the institutional ethics committee of Barwon Health (No. 23.178). Individual consent, including consent for the publication of de-identified data such as imaging and operative video material, was waived on the basis of the study posing negligible risk and all data being acquired and accessed as part of routine clinical practice.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Kandathil A, Chamarthy M. Pulmonary vascular anatomy & anatomical variants. Cardiovasc Diagn Ther 2018;8:201-7. [Crossref] [PubMed]
2. Nomori H, Okada M. Illustrated anatomical segmentectomy for lung cancer. Tokyo: Springer; 2012.
3. 3D Slicer image computing platform. [Cited 2025 Nov 17]. Available online: https://slicer.org/
4. Laven IEWG, Oosterhoff VPS, Franssen AJPM, et al. Evaluating three-dimensional lung reconstructions for thoracoscopic lung resections using open-source software: a pilot study. Transl Lung Cancer Res 2024;13:1595-608. [Crossref] [PubMed]
5. He H, Yu C, Yang Y, et al. Three-Dimensional Reconstruction and Virtual Simulation of Patient-Specific Anatomy for Procedural Planning in Thoracoscopic Segmentectomy: A Systematic Review and Meta-Analysis. Eur J Cardiothorac Surg 2025;67:ezaf283. [Crossref] [PubMed]
6. Rice D. Step-By-Step Guide to 3D Segmentation of Tumor, Airways, and Vessels for Preoperative Planning for Anatomic Segmentectomy Using 3D Slicer. CTSNet 2023. doi: 10.25373/ctsnet.21816309.v1
7. Pinter C, Lasso A, Fichtinger G. Polymorph segmentation representation for medical image computing. Comput Methods Programs Biomed 2019;171:19-26. [Crossref] [PubMed]
8. Markups—3D Slicer documentation. [Cited 2025 Nov 17]. Available online: https://slicer.readthedocs.io/en/latest/user_guide/modules/markups.html
9. Saji H, Okada M, Tsuboi M, et al. Segmentectomy versus lobectomy in small-sized peripheral non-small-cell lung cancer (JCOG0802/WJOG4607L): a multicentre, open-label, phase 3, randomised, controlled, non-inferiority trial. Lancet 2022;399:1607-17. [Crossref] [PubMed]
10. Altorki N, Wang X, Kozono D, et al. Lobar or Sublobar Resection for Peripheral Stage IA Non-Small-Cell Lung Cancer. N Engl J Med 2023;388:489-98. [Crossref] [PubMed]
11. Fortich S, Nguyen DM, Avella DM, et al. An update on three-dimensional computed tomography reconstruction and its applications to lung segmentectomy: a narrative review. AME Surg J 2025;5:44.
12. Lin G, Li R, Li X, et al. Advances in the Application of Three-Dimensional Reconstruction in Thoracic Surgery: A Comprehensive Review. Thorac Cancer 2025;16:e70159. [Crossref] [PubMed]
13. Lo P, van Ginneken B, Reinhardt JM, et al. Extraction of airways from CT (EXACT'09). IEEE Trans Med Imaging 2012;31:2093-107. [Crossref] [PubMed]
14. Stoll-Dannenhauer T, Schwab G, Zahn K, et al. Computed tomography based measurements to evaluate lung density and lung growth after congenital diaphragmatic hernia. Sci Rep 2021;11:5035. [Crossref] [PubMed]