Digital twin-based bronchoscopy simulator improves training performance and skill retention of novices: a randomised controlled study.

Abstract

Conventional bronchoscopy training often does not ensure lasting skill retention or adaptability to different anatomies, limiting real-world impact. This study used a digital-twin bronchoscopy simulator with various CT-derived bronchial tree models to better train novices. To explore training with various anatomically diverse bronchial tree models in novices' bronchoscopy performance. 60 bronchoscopy-naive participants were randomly assigned to three groups (n=20 each): control (written instruction only), anatomic-uniformity (trained on one standard bronchial model) and anatomic-variety (trained on multiple patient-derived bronchial models). All participants performed two tests: test 1 on a standard model and test 2 on a new CT-derived model. Both tests were repeated 3 months later to assess skill retention. The primary comparison was between the anatomic-variety and anatomic-uniformity groups. 60 participants completed tests 1 and 2. 55 returned at 3 months. In test 1, there were no significant differences between the anatomic-variety and anatomic-uniformity groups in diagnostic completeness (DC, 0 segments, p=0.576), structured progress (SP, 1 correct progression, p=0.091) and procedure time (31 s, p=0.831). In test 2, the anatomic-variety group had significantly higher DC (2.5 segments, p<0.001) and SP (9 progression, p<0.001) than the anatomic-uniformity group. At 3 months, the anatomic-variety group retained superior DC and SP scores in both tests despite slight declines. Training with diverse anatomical models significantly enhanced bronchoscopy performance compared with repetitive practice on a single standardised model with partially maintained learning gains at 3 months.