Artificial intelligence navigation is becoming more common in bronchoscopy, but hospitals still cannot determine how much of the clinical benefit comes from AI. Systems such as Johnson & Johnson’s MONARCH QUEST combine navigation software with robotic hardware, 3D imaging, and expanded computing power. That makes it difficult to credit better results to any one component.
This distinction matters to hospitals deciding whether the technology is worth the cost and changes to clinical workflows. Recent studies show AI can improve airway mapping and examination coverage. What they still do not show is that AI navigation improves lung-nodule biopsy yield when the robot, imaging system, and care team remain the same.
Why AI navigation was developed
Bronchoscopy is a procedure used to inspect the airways and diagnose lung conditions, including nodules that may be cancerous. Reaching peripheral lung nodules is challenging because airways branch in complex ways and respiratory motion can shift targets. Navigation systems were developed to help guide the bronchoscope to the lesion using CT-based maps. Early systems used electromagnetic tracking, and more recent robotic platforms added flexible catheters and improved stability. AI is now being layered onto those systems to interpret video, identify landmarks, and recommend next steps in real time.
The development of AI airway navigation builds on decades of work in computer-assisted bronchoscopy. Early navigation systems required registration between CT images and the live anatomy. Electromagnetic navigational bronchoscopy emerged in the early 2000s and used sensors at the tip of the catheter. Robotic bronchoscopy became more common in the late 2010s, with platforms such as MONARCH receiving FDA clearance. Adding AI to these platforms is a logical next step, but it also introduces the need for new validation methods. The medical device field is accustomed to hardware-based evidence, while AI software needs clinical studies that measure its contribution independently.
What current studies actually show
Johnson & Johnson introduced MONARCH QUEST after the FDA cleared an update to the MONARCH Platform in January 2025. J&J says the system’s NVIDIA RTX hardware increased real-time computing power by 260 percent, allowing it to run more complex navigation algorithms. In July 2026, the company reported the first clinical use of QUEST in the Middle East at Cleveland Clinic Abu Dhabi.
Those developments show that the technology is reaching clinical settings. They do not establish that AI improves biopsy results.
A 2026 study of AIBAN tested AI airway navigation more directly. The software identifies airway landmarks from bronchoscopy video, estimates the bronchoscope’s location, and recommends the next airway along a planned route. In simulation, it correctly localized the scope in 91.1 percent of validation frames and followed the correct anatomical path in 96.2 percent of test videos.
Clinical evidence is starting to emerge as well. A May 2026 multicenter randomized trial involving 204 patients tested AI assistance during conventional bronchoscopy. AI increased the proportion of bronchial anatomy examined from 82.70 percent to 92.09 percent, although average inspection time increased from 176.8 to 221.1 seconds.
The study measured examination coverage, not whether AI helped doctors reach peripheral lung nodules or improved biopsy yield. That is an important distinction. A 2025 systematic review had found that all four studies implementing AI airway navigation at that time were conducted in simulation.
Hospitals still cannot isolate AI’s effect on biopsy results
The question of attribution is central. If a robotic bronchoscopy platform with AI is used and yield improves, how much of that improvement is due to AI? The hardware may provide better reach, the imaging system may improve target localization, and the software may recommend a better path. The care team’s skill and technique also affect outcomes. To justify the cost of an AI upgrade, hospitals need evidence that the AI component adds value beyond the rest of the platform.
A 2026 Mayo Clinic study showed why separate evaluation matters. It compared 331 MONARCH procedures performed with and without mobile cone-beam CT. Diagnostic yield was almost identical: 70.9 percent with imaging and 71.7 percent without. The imaging-supported procedures were shorter but involved substantially more radiation. This did not test AI, but it showed that a high-tech add-on can change workflow and radiation without changing the outcome that matters most to patients.
Clinical AI also has to produce results that are reproducible across different teams and environments before hospitals can rely on performance reported in a limited study. The same caution applies when healthcare AI demonstrates a narrow technical capability before evidence supports broader clinical claims.
The software itself also requires monitoring after deployment. The FDA currently lists an open Class II recall covering certain MONARCH systems because restarting the software can reset patient-side positioning and potentially cause unexpected robotic-arm movement. The FDA attributes the problem to software design, not AI navigation, so the recall is not evidence that QUEST’s AI is unsafe.
FDA clearance also does not answer whether one component delivers better clinical outcomes than another. Regulatory approval is based on safety and intended use, not comparative effectiveness. Hospitals evaluating newer robotic surgery systems face a similar need to separate regulatory clearance from evidence about performance in practice.
Why biopsy yield matters
Biopsy yield is important because it determines whether the procedure can provide a definitive diagnosis. A lower yield means patients may need a second procedure, a more invasive test, or follow-up imaging. Improving yield is therefore one of the primary goals for any bronchoscopy technology. AI navigation could support this goal by helping doctors see more of the lung and avoid missing lesions. Yet seeing more airways is not the same as reaching a nodule. The AIBAN simulation and the 2026 trial both focused on navigation coverage, not on nodule targeting or tissue acquisition.
None of this means AI navigation lacks value. It could reduce variability among operators, help trainees locate difficult branches, and improve documentation of what was examined. It may also make robotic systems faster to learn. The AIBAN results suggest the software can correctly identify anatomy in most frames. If prospective studies confirm those results, AI could become a valuable adjunct. But the evidence has not yet established that AI alone improves the outcome that matters most: getting a reliable tissue sample from a suspicious nodule.
The next steps for evidence
For MONARCH QUEST, the most useful next study would compare the same bronchoscopy platform with AI navigation enabled and disabled while keeping the catheter, imaging system, operator, and clinical workflow unchanged. It should then measure outcomes hospitals actually care about, including biopsy yield and procedure performance across multiple clinical sites.
That kind of trial is difficult but not impossible. Randomized controlled studies have been done for other AI tools. The challenge in bronchoscopy is that AI is sold as part of a complex platform, and vendors may be reluctant to support studies that could show no added benefit. Nonetheless, payers and hospital committees increasingly demand such evidence before providing reimbursement or purchasing new technology.
Cost is also a concern. AI navigation upgrades add computing power, licensing fees, and training time. Hospitals must decide whether those costs translate into better patient outcomes or operational efficiency. If inspection time increases, as it did in the 2026 trial, that could slow down a busy bronchoscopy suite and reduce the number of procedures performed in a day. Shorter procedure times with imaging in the Mayo study did not improve yield, so faster is not always better. The tradeoffs are complex and need site-specific evaluation.
Hospitals evaluating these systems should ask specific questions. Which components changed compared with the prior platform? What do the clinical studies actually measure? Were the study patients similar to our patient population? How will outcomes be tracked after implementation? These questions can help clarify whether a claimed improvement is real and reproducible.
Until the evidence exists, hospitals have good reason to recognize the progress in AI-assisted bronchoscopy without assuming that every improvement from an AI-enabled platform was produced by the AI. A recent study found ChatGPT Health missed more than half of clinician-defined medical emergencies, providing another example of why clinical AI needs controlled evaluation before it guides care. The technology is advancing rapidly, but clinical proof is still catching up.
Source: eWeek News