Invasive Procedures in Interventional Pulmonology: A Narrative Review of Educational Evidence-Based Step-By-Step Protocols

Diagnosing and staging lung cancer and lung diseases often necessitate invasive procedures that yield high-quality tissue samples while minimizing risk to the patient. With new minimally invasive techniques emerging¹, systematic and step-wise approaches are essential to minimize procedural risk and maximize diagnostic yield. Whether targeting the lung parenchyma, pleural space, or mediastinal lymph nodes, structured procedural strategies reduce variability, standardize training, and enhance patient outcomes^{2,3,4,5,6,7,8}.

However, despite international guidelines and growing consensus on best practices, invasive procedures are often learned and executed differently across institutions and doctors⁹. The traditional Halstedian principle of "see one, do one, teach one" remains central to medical training, but is increasingly supported by visual and structured learning tools. In this context, video-based, standardized step-wise protocols can enhance procedural understanding, especially for early-career clinicians or when adopting new techniques. By providing a visual and methodical "see one," such articles could allow practitioners to gain an understanding of a procedure before the first "do one" or compare and refine their practice at the point of "teach one" level.

This narrative review summarizes recent step-wise approaches in interventional pulmonology for flexible bronchoscopy, endobronchial and endoscopic ultrasound, radial EBUS, electromagnetic navigation bronchoscopy, local anesthetic thoracoscopy, endobronchial cryobiopsy, and transthoracic ultrasound-guided lung biopsy (Table 1), each contributing to a broader, evidence-based framework for invasive pulmonology.

Table 1: Step-by-step protocols.

Lung cancer accounts for the highest number of cancer deaths globally¹⁰, with the patient requiring an invasive pulmonology procedure for tissue sampling and staging.

Flexible bronchoscopy, EBUS, and endoscopic ultrasound (EUS)
Tissue sampling from central tumors can be done using flexible bronchoscopy³, with lymph node sampling using endobronchial ultrasound (EBUS)⁵ and endoscopic ultrasound (EUS) of the esophagus⁶. Cold et al.³ present a four-landmarks approach for investigating the 18 bronchial segments efficiently (Figure 1)³, based on these three outcome measures: Diagnostic Completeness (DC) (e.g., inspected segments), Structured Progress (SP)¹¹, and Mean Intersegmental Time (MIT) (e.g., procedure efficiency). Concerningly, several studies have shown experienced bronchoscopists did not visit all segments, although they were instructed to do so^{11,12,13,14,15,16,17}. Experience does not ensure expertise, justifying the need for systematic approaches³. This shift toward competency-based training has been strongly endorsed by the 2015 CHEST Expert Panel Report on Adult Bronchoscopy Training¹⁸, which recommends validated assessment frameworks rather than procedural volume as the foundation for certification and patient safety. Therefore, the four-landmarks approach has been compared to an artificial intelligence (AI) navigation system for novice bronchoscopists¹⁹, favoring AI as the novices outperformed the four-landmarks approach, even when the AI was disabled¹⁹. The same AI also improved the performance for intermediate and experienced bronchoscopists²⁰, emphasizing that experience does not ensure expertise and the "see one, do one, teach one" approach is an outdated and insufficient training method, even when the physicians have performed >250 bronchoscopies, defined as experienced in the study²⁰. The Halstedian method or general apprenticeship training in bronchoscopy might also be revised. The only randomized controlled trial directly comparing AI-assisted bronchoscopy training with expert-led instruction²¹, demonstrated that AI guidance significantly improved procedural efficiency while maintaining diagnostic completeness and cognitive load comparable to expert tutoring. This finding supports the integration of AI as a supplemental teaching tool rather than a replacement for human expertise, offering a structured and scalable approach to competency-based training. This could be especially important for resource scar settings, where expert-led instructions are scarce. However, the scalability of AI in bronchoscopy need to be evaluated in future studies²². Another study showed superiority with the AI plus mastery learning²³, where the trainees have to train towards proficiency targets according to the performance measures DC, SP, and MIT. Even though mastery learning has been suggested for learning flexible bronchoscopy by the same 2015 CHEST expert panel for a decade¹⁸, this was the first study using mastery learning²⁴, highlighting gaps between bronchoscopy training studies and medical educational literature²⁵.

Figure 1: The four landmark approach for inspecting the bronchial tree systematically and efficiently. This figure has been modified with permission from Cold et al.¹⁸. Please click here to view a larger version of this figure.

AI might be the solution to learning and ensuring competency in flexible bronchoscopy²⁶. AI can successfully improve and ensure the performance in colonoscopy^27,28,29, with several AI systems being commercially available. Clinical implementation of AI in bronchoscopy remains in its infancy, with most studies still focusing on diagnostic applications in EBUS and EUS rather than real-world procedural guidance³⁰. Therefore, the structured approach to EBUS⁵ and EUS⁶ will one day be supported by AI but rely on the physician's expertise for now. An expertise, which is most efficiently acquired in a simulation-based setting, as simulation-based training is superior to classical clinical training in EBUS³¹. Simulation-based training spares the patient from the initial part of the clinician's learning curve, and not just novices can benefit from simulation-based training interventions³² (Figure 2). Structured procedural strategies reduce variability, standardize training, and enhance patient outcomes in EBUS^33,34 as presented with the six-landmark approach in EBUS by Nielsen et al.⁵ (Figure 3). Based on this approach, the European Respiratory Society (ERS) has launched a training program, ensuring competence can be achieved in a simulation-based setting³⁵. Unfortunately, no virtual reality (VR) training simulators are available in EUS^6,36, likely due to the technical complexity of replicating both ultrasound and endoscopic imaging in one system and the difficulty of achieving realistic haptic feedback. We urgently ask developers and researchers to develop such, as EUS has been identified as the fourth most important technical procedure for pulmonary residents to learn³⁷. EBUS and EUS should be combined for the diagnosis and staging of lung cancer³⁸, giving their complementary use of reaching different lymph nodes (Figure 4). These recommendations are in accordance with the 2015 ERS/ESGE/ESTS Combined Endosonography Guideline for lung cancer staging, which emphasizes systematic and combined EBUS-EUS assessment for optimal diagnostic accuracy³⁸. Pulmonary physicians should therefore not be limited to only entering their organ systems, but could use the same endoscope to enter the esophagus and even collect thyroid biopsies when needed³⁹, or use EBUS in diagnosing and staging of other diseases like sarcoidosis, where systematic sampling improves diagnostic yield and reduces the need for surgical biopsy⁴⁰.

Figure 2: Simulation-based training can be beneficial both to novice and experienced physicians. Please click here to view a larger version of this figure.

Figure 3: The six-landmarks approach in EBUS. This figure has been modified with permission from Nielsen et al.⁵. Please click here to view a larger version of this figure.

Figure 4: Lymph nodes accessible by EBUS and EUS for the staging and diagnosis of lung cancer. This figure has been modified with permission from Issa et al.⁶. Please click here to view a larger version of this figure.

Radial EBUS (rEBUS), electromagnetic navigation bronchoscopy (ENB), and local anesthetic thoracoscopy
When target lesions are located more peripherally, radial EBUS (rEBUS) in combination with electromagnetic navigation bronchoscopy (ENB) can be used⁸. A previous article described only ENB⁴¹, and Juul et al. should be applauded for adding new techniques, such as rEBUS⁸, to ENB. Few studies have reported an increase in diagnostic yield by adding rEBUS to ENB⁴². Unfortunately, the studies are not consequently reported in accordance with EQUATOR guidelines⁴². Again, this highlights a gap between research in interventional pulmonary procedures and current guidelines. This has especially been true for research in assessment tools of performance, with only 3 percent of studies adhering to contemporary frameworks for validity evidence⁴³. Applying Messick's validity framework is essential to standardize assessment research, minimize subjective bias, and ensure that competency evaluation tools genuinely reflect clinical proficiency. Thereby, frameworks help in strengthening both scientific rigor and patient safety. Similarly, adherence to the EQUATOR reporting guidelines ensures methodological transparency and reproducibility in procedural research, allowing results to be critically appraised and compared across studies. Both are essential for improving patient safety and advancing evidence-based training. Not adhering to frameworks may risk overlooking the accumulated scientific foundation upon which contemporary assessment research is built, resembling attempts to reinvent rather than refine established and validated approaches. Therefore, Messick's framework for validity evidence is recommended by the American Educational Research Association in their Standards for Education and Psychological Testing since 1999^44,45. Messick collects validity evidence from five sources to ensure a test measures what it intends to measure⁴⁶. Applying Messick's unified framework is essential to standardize assessment research, minimize subjective bias, and ensure that competency evaluation tools genuinely reflect clinical proficiency -- thereby strengthening both scientific rigor and patient safety. Bodtger et al.² present a step-wise approach for local anesthetic thoracoscopy for undiagnosed pleural effusion, relying on an assessment tool relying on Messick's validity framework to ensure competence in local anesthetic thoracoscopy⁴⁷. Unlike in EUS, Nayahangan et al.⁴⁷ also developed a simulator for training. Since local anesthetic thoracoscopy could also be used for managing other diseases, like pleural empyema⁴⁸, such training modalities and assessment tools could rely on testing in different scenarios, which would increase the discriminatory abilities of a test and have been developed for bronchoscopy¹³.

Transbronchial lung cryobiopsy (TBLC)
Davidsen et al.⁴ presented a structured, step-wise protocol for transbronchial lung cryobiopsy (TBLC). TBLC is increasingly used as a first step in diagnostic interstitial lung diseases (ILD), given its less invasive alternative to surgical lung biopsy (SLB)⁴⁹. Davidsen et al.⁴ emphasize critical procedural steps, including fluoroscopic guidance, precise biopsy site targeting, and prophylactic balloon placement to minimize bleeding risk -- one of the most significant complications associated with TBLC⁵⁰. While TBLC can deliver diagnostic results comparable to surgical biopsy in selected patients⁵¹, the authors underscore that successful outcomes depend on institutional experience, rigorous training, and multidisciplinary discussion of cases⁴⁹, and can even be combined with robot-assisted bronchoscopy (RAB)⁵². However, despite the growing interest in TBLC, procedural standardization and competency-based training remain underdeveloped in many settings. As Davidsen et al.⁴ argue, reproducibility, safety, and diagnostic yield are all improved by adhering to a clear, validated procedural framework, and their article serves as an important educational tool for centers adopting TBLC into clinical practice.

Ultrasound-guided transthoracic needle biopsy (US-TTNB)
Lung biopsies can also be obtained trans thoracically (TT). Laursen et al.⁷ present a step-wise approach for ultrasound-guided lung biopsies in patients with peripheral lung cancer. While techniques like computer tomography (CT)-guided and ultrasound-guided transthoracic needle biopsy (US-TTNB) offer a safe, cost-effective alternative with comparable diagnostic yields⁵³ and a less demanding setup than electromagnetic navigation TTNB⁵⁴. Its primary limitation is the requirement for lesions to be visualizable with ultrasound -- typically subpleural consolidations⁵⁵. Nonetheless, in appropriately selected patients, US-TTNB provides high diagnostic accuracy, low complication rates (e.g., pneumothorax ~4.6%)⁵⁵, and faster procedure times⁵³. The step-wise protocol⁷ integrates procedural, technical, and patient-centered considerations, reflecting current practice in outpatient settings⁷ with real-time image guidance, pre- and post-procedure safety steps, and options to optimize yield, such as the use of contrast-enhanced ultrasound (CEUS) and combining fine-needle aspiration with core biopsy^56,57. Despite widespread use of thoracic ultrasound, US-TTNB training is not standardized and lacks integration into formal curricula⁵⁸. Learning curve data suggest that even experienced physicians may require structured training to achieve procedural competency⁵⁹, again highlighting that experience does not ensure expertise and underlining the need for education enhancing research of medical procedures. AI is also here emerging, with a study showing that trained novices can produce expert-level lung US⁶⁰. Again, Laursen et al.⁷ should be applauded for relying on assessment tools based on Messick's validity framework^61,62,63.

Limitations and future perspectives
The future perspectives for invasive pulmonology procedures, like in the rest of society, will heavily rely on AI for training and assessment. A systematic review on AI in bronchoscopy still identified large gaps in clinical implementation³⁰. None of the developed AIs were open-source or used open datasets, which can enhance external use and validation^64,65,66. Despite its promise, clinical implementation of AI remains limited, along with the 'black box' challenge of algorithm interpretability. Overreliance on AI may risk skill degradation if AI systems are used as substitutes rather than complementing clinical judgment and performance. We therefore encourage future AI studies to limit AI to the training part of the intervention group, without the use of AI for testing, ensuring AI becomes a valuable training partner rather than a replacement for skill acquisition. As very few AI studies in bronchoscopy adhered to reporting guidelines²², we highly recommend future studies to do so, ensuring standardization and reproducibility with this emerging technology.

As a narrative review, this work does not possess the same methodological rigor as a systematic review, and other important aspects could have been addressed more comprehensively. Therefore, future research when implementing step-by-step protocols should explore the cost-effectiveness of implementing these along with their advanced technologies, such as electromagnetic navigation bronchoscopy and AI-driven training tools, particularly across different healthcare settings where resource disparities may limit access. Ethical considerations also require attention, including the risk of algorithmic bias and issues of liability in AI-assisted clinical decision-making. Moreover, most AI-driven approaches remain at the proof-of-concept stage, and large-scale, multicenter validation studies are essential to confirm their clinical reliability and generalizability²². As AI in bronchoscopy studies at large fail to adhere to recommended guidelines, we further emphasize that we recommend AI, training, and assessment studies to adhere to relevant guidelines and use contemporary frameworks for validity evidence to ensure robust scientific rigor.

With new techniques emerging in invasive pulmonology, visual articles allow the trainee to "see one" several times. The "do one" should be done through many simulation-based training and assessment sessions. Simulation-based training offers standardized environments to ensure procedural proficiency before exposing patients to procedural risks. AI is an emerging tool, which has already proven its usefulness in bronchoscopy, EBUS, and lung ultrasound. When "teaching one", it should be supported by AI to ensure that expertise, not just experience, is transferred to the trainee.