AI-driven intelligent training enhances clinical competence in oncology residency: a randomized controlled trial

TL;DR

An AI-powered training system improved clinical skills and knowledge in oncology residents compared to traditional methods.

Contribution

A novel AI-driven educational model combining personalized learning, virtual mentorship, and collaborative platforms was developed and tested.

Findings

Residents using AI training showed better knowledge, procedural accuracy, and collaboration than controls.

AI training reduced cognitive workload and improved long-term knowledge retention and clinical outcomes.

The AI system enhanced adaptability and evidence-based practice in oncology residents.

Abstract

Rapid advances in artificial intelligence (AI) offer new opportunities to address persistent challenges in healthcare professions education, particularly in oncology residency training, where rapidly evolving knowledge, complex decision-making, and limited high-fidelity practice environments hinder competency development. However, evidence from rigorously evaluated educational interventions remains limited.

We conducted a randomized controlled trial involving 124 breast oncology residents from three tertiary hospitals. Participants were randomly assigned to an AI-empowered intelligent teaching (AIEIT) group (n = 62) or a control group receiving conventional training (n = 62). The AIEIT model integrated a dynamic knowledge graph for personalized learning, a virtual patient–AI mentor system for adaptive skills training, a mixed-reality multidisciplinary team platform for collaborative decision-making, and a learning analytics dashboard for continuous feedback. Outcomes included knowledge acquisition, clinical reasoning, procedural competence, collaborative performance, cognitive efficiency, and longitudinal clinical outcomes.

The AIEIT group outperformed the control group across all domains, demonstrating superior mastery of theoretical knowledge, higher procedural accuracy, and greater multidisciplinary collaboration (all P < 0.001). Cognitive workload and training time were significantly reduced, while technology adaptability and evidence-based practice utilization markedly improved. At 3-month follow-up, the AIEIT group maintained higher knowledge retention (91.2 ± 3.5% vs 76.8 ± 8.4%, P < 0.001) and better clinical outcomes, including fewer postoperative complications and higher patient satisfaction.

This study demonstrates that an AI-driven, closed-loop educational model can substantially enhance clinical competence formation in oncology residency training. By integrating data-driven personalization, human–AI collaboration, and virtual–real learning environments, the AIEIT framework offers a scalable and evidence-based approach for advancing healthcare professions education.

Figures