A Comparative Bicentric Study on Ultrasound Education for Students: App- and AI-Supported Learning Versus Traditional Hands-on Instruction (AI-Teach Study)
Elena Höhne, E. Bauer, Claus Bauer, Valentin Sebastian Schäfer, Jennifer Gotta, Philipp Reschke, Thomas J. Vogl, İbrahim Yel, Johannes Weimer, Agnes Wittek, Florian Recker
Abstract
BACKGROUND: The integration of artificial intelligence (AI) into medical education presents significant opportunities for enhancing teaching methods and student learning outcomes. Despite its potential benefits, the implementation of AI in curricula remains limited and lacks standardized approaches. OBJECTIVE: This bicentric pilot study aims to examine the effectiveness of an innovative ultrasound course for medical students that combines AI-based teaching with blended e-learning, compared to traditional classroom lessons, to optimize educational practices. MATERIAL AND METHODS: This bicentric pilot study included medical students who were randomly assigned to an experimental group receiving AI-based blended e-learning for an ultrasound course or a control group receiving traditional classroom instruction. The curriculum consisted of two modules: lung ultrasound and Focused Assessment with Sonography for Trauma (FAST). The effectiveness of the interventions was evaluated using objective structured clinical examinations (OSCE) to assess ultrasound skills, administered as pre-tests and post-tests. Additionally, the quality of the ultrasound images obtained during the final assessment was rated using a standardized scoring system to further assess student competency. RESULTS: 50 clinical-phase medical students participated. OSCE results for both FAST and lung modules revealed no significant differences between groups at both pretest (pretest FAST p=0.722, pretest Lung p=0.062) and final exam (final exam FAST p=0.634, final exam lung p=0.843), with both cohorts achieving comparable improvements and nearly identical final scores, while ultrasound image evaluations confirmed similar outcomes (FAST images p=0.558 and lung images p=0.199) with excellent interrater reliability (ICC=0.993). CONCLUSION: AI- and app-based learning methods in ultrasound education showed to be equally effective as traditional hands-on teaching for medical students in this study. Incorporating the permanently growing innovations auf AI into curricula can provide valuable tools for educators and students alike.