Artificial Intelligence in Bronchoscopy for Pulmonary Disease: A Systematic Review of Randomized Controlled Trials

  • Eric Daniel Tenda, Dr Division of Pulmonology and Critical Medicine, Department of Internal Medicine, Faculty of Medicine, Universitas Indonesia, National General Hospital of Dr. Cipto Mangunkusumo, Indonesia
  • Almerveldy Azaria Dohong, Dr Divisi Respirologi dan Penyakit Kritis, Departmen Ilmu Penyakit Dalam, Fakultas Kedokteran, Universitas Indonesia
  • Nadia Iqbal, Dr Faculty of Medicine, Pembangunan Nasional Veteran Jakarta University, Jakarta, Indonesia
  • Aziza Harris, Dr Division of Pulmonology and Critical Medicine, Department of Internal Medicine, Faculty of Medicine, Universitas Indonesia, National General Hospital of Dr. Cipto Mangunkusumo, Indonesia
  • Steffi Cong Andi Nata, Dr Division of Pulmonology and Critical Medicine, Department of Internal Medicine, Faculty of Medicine, Universitas Indonesia, National General Hospital of Dr. Cipto Mangunkusumo, Indonesia
  • Muhammad Aziz Arridho, Dr Faculty of Medicine, Syiah Kuala University, Banda Aceh, Indonesia
  • Alfino Syahputra, Dr Division of Pulmonology and Critical Medicine, Department of Internal Medicine, Faculty of Medicine, Universitas Indonesia, National General Hospital of Dr. Cipto Mangunkusumo, Indonesia
  • Sahat Halim, Dr Faculty of Medicine, Universitas Sumatera Utara, North Sumatra, Indonesia
##semicolon## artificial intelligence, bronchoscopy, randomized controlled trial, simulation based training, procedural performance

Abstract

Background: Bronchoscopy is a core diagnostic and therapeutic procedure for pulmonary diseases, yet performance is highly operator dependent, contributing to variability in airway inspection completeness and procedural efficiency. Artificial intelligence (AI) systems have been developed to support real time airway anatomy recognition, navigation, and performance feedback. This systematic review summarizes randomized controlled trial evidence on the impact of AI guided bronchoscopy on procedural performance.

Methods: This review followed PRISMA guidance and was registered in PROSPERO. We searched PubMed/MEDLINE, ScienceDirect, and SAGE Journals from database inception to January 2026. Two reviewers independently screened studies and extracted data. Risk of bias was assessed using the RoB 2 tool. Given heterogeneity in interventions, populations, and outcome definitions, results were synthesized narratively.

Results: From 1,654 records, four randomized controlled trials met inclusion criteria. All studies were conducted in simulation settings and enrolled participants with varying experience (novice to experienced bronchoscopists, including critical care physicians). Overall, AI guided bronchoscopy favored AI supported groups compared with comparators (no AI, written instruction, directed self learning, or expert instruction), showing improvements in key performance domains such as inspection completeness and structured progression, as well as efficiency related metrics including intersegmental time and or total procedure time.

Conclusion: Simulation based RCT evidence supports AI as a promising approach to improve standardization and procedural performance in bronchoscopy training. Multicenter patient based randomized trials are warranted to confirm benefits on clinically meaningful outcomes.

Keywords: artificial intelligence; bronchoscopy; randomized controlled trial; simulation based training; procedural performance.