Augmented Reality–Based Liver Surgery Training: A Randomized Controlled Trial on Anatomical Comprehension, Efficiency, and Learner Engagement

Background Liver cancer remains one of the leading causes of cancer-related mortality worldwide, with surgical resection currently serving as the only curative option. The liver’s intricate three-dimensional anatomy, including complex vascular architecture and variable segmental organization, makes surgical training especially demanding. Conventional educational tools—such as two-dimensional atlases and cadaveric dissections—provide essential foundations but are limited in their ability to reproduce realistic intraoperative scenarios. Restricted cadaver availability and the absence of dynamic features, such as tissue elasticity and bleeding risk, further contribute to a steep learning curve, often delaying the development of operative proficiency. Advances in medical imaging and visualization technologies have introduced new possibilities for bridging these gaps, with augmented reality (AR) offering an immersive, interactive platform to enhance surgical education. Methods We designed an AR-based training system using a head-mounted display, which incorporated patient-specific three-dimensional liver models generated from CT/MRI imaging data. The system was further enhanced with real-time haptic simulation, and automated performance analytics. In a randomized controlled trial, surgical residents (n = 60) were assigned to either AR-based training (n = 30) or conventional instruction (n = 30). Primary outcomes included anatomical knowledge acquisition, training efficiency and knowledge retention. Secondary outcomes focused on learner engagement and usability. Engagement was evaluated through validated post-training surveys, including a 5-point Likert scale (e.g., self-reported interest and active questioning during sessions). Usability was assessed in the AR group using the System Usability Scale (SUS). In addition, participants were asked about their preference for AR training compared with conventional textbook-based learning. Results The AR group achieved higher anatomical test scores (4.5 ± 0.3 vs. 3.2 ± 0.5, p < 0.001), better operative planning accuracy (87% vs. 62%, p = 0.003), and greater procedural confidence (VAS 8.1 ± 0.9 vs. 5.4 ± 1.2, p < 0.001). Training efficiency was enhanced, with a 40% reduction in learning time (3.2 ± 0.8 vs. 5.4 ± 1.2 hours, p = 0.01). Knowledge retention after four weeks remained higher in the AR group (85% vs. 62%, p = 0.03). Learner engagement was also greater, with 90% reporting increased interest, more student-initiated questions, and high satisfaction (System Usability Scale: 82 ± 6). Technical evaluation demonstrated submillimeter model accuracy (0.4 ± 0.1 mm error) and rendering latency under 100 ms, with the inclusion of pathological variants in 15% of cases. Conclusion AR-based surgical training provides significant advantages in anatomical understanding, operative planning, and trainee confidence compared with conventional methods. The system improves efficiency, enhances learner engagement, and offers scalable opportunities to incorporate patient-specific variability. These findings support the integration of AR into surgical curricula as a means to accelerate competency development and improve preparedness for complex liver resections.