Artificial intelligence and digital health in bronchoscopy: advances in diagnostic accuracy, real-time navigation and procedural skills training.

Authors

Abstract

Artificial intelligence (AI) and procedure-linked digital health technologies are beginning to transform bronchoscopy across three interconnected domains. In terms of diagnostic accuracy, deep learning models applied to endobronchial ultrasound images and bronchoscopic videos can match or exceed human readers in differentiating benign from malignant lesions, although nearly all evidence remains at the level of test-accuracy metrics rather than clinical outcomes. In real-time navigation, AI-augmented fluoroscopy and tomosynthesis platforms represent the clearest example of clinically deployed AI that improves lesion localisation and diagnostic yield for peripheral pulmonary nodules, while camera-based localisation and shared-control robotic systems remain largely preclinical. In training, multiple randomised controlled trials demonstrate that AI-guided simulators significantly improve performance across experience levels, and that AI-derived metrics can objectively assess competence, although the first attempt to show impact on real-world clinical performance was inconclusive. This review synthesises the current literature across these three domains, distinguishing technologies that are currently used in clinical practice with supportive published evidence from those that remain investigational. It provides practising bronchoscopists and trainees with a balanced appraisal of what AI can realistically offer today, what limitations persist, and what developments are likely over the coming decade.

Tags

Topics