Enhancing ultrasound training for breast cancer diagnosis: a controlled study of AI-assisted learning.

OBJECTIVES: This study aimed to develop and evaluate an AI-assisted teaching platform to enhance diagnostic competency in breast ultrasound. The goal was to assess whether AI integration improves diagnostic accuracy, learning efficiency, and participant satisfaction within a residency training program. METHODS: We conducted a cohort-based study at our hospital. Twelve junior residents (experimental group) underwent AI-assisted training via a newly implemented platform, while twelve senior residents (control group) completed conventional training. Diagnostic performance was evaluated before and after the one-month intervention using consistent assessments. Participant satisfaction was surveyed across domains including learning engagement, skill development, and confidence. RESULTS: In the experimental group, post-intervention diagnostic scores (90.50 ± 9.82) were significantly higher than pre-intervention diagnostic scores(70.00 ± 17.55, P = 0.003,95%CI[-32.54,-8.46], Cohen's d=-1.44). Survey results indicated high satisfaction: 83.33% strongly agreed the platform facilitated learning, 66.67% reported improved pattern recognition, and 66.67% noted increased engagement in self-learning. A majority also reported gains in clinical reasoning and confidence when facing a real patient. CONCLUSIONS: We integrated an AI-assisted platform into ultrasound residency training, creating an educational tool. In this single-center exploratory study, the AI-assisted platform shows potential to improve residents' diagnostic skills for breast ultrasound.