Application of an artificial-intelligence-based transesophageal echocardiography simulation system in residency...

Transesophageal echocardiography (TEE) is a key perioperative monitoring modality in anesthesiology, but traditional training is constrained by limited case exposure, patient safety concerns, and heterogeneous learning curves. Artificial-intelligence (AI)-based TEE simulators that integrate 3-dimensional (3D) anatomy visualization, interactive image interpretation, and virtual probe manipulation may improve learning efficiency and confidence in residents.

In this single-center randomized controlled study, sixty anesthesiology residents in standardized training at Sichuan Academy of Medical Sciences & Sichuan Provincial People's Hospital (December 2024-December 2025) were randomly allocated to an AI group (n = 30) or a control group (n = 30). Both groups received the same instructor-led TEE curriculum. The control group underwent traditional teaching (lectures, static images, and video demonstrations), whereas the AI group additionally trained on an AI-based TEE simulation system that provided 3D anatomy visualization, interactive image-reading exercises, and virtual probe operation with real-time feedback. Outcomes included a standardized image-interpretation test (0-100), Objective Structured Assessment of Technical Skills for TEE (OSATS-TEE; 0-150), key-view acquisition rate, procedure time, self-efficacy score, and course satisfaction (all 0-100).

Baseline demographic characteristics did not differ between groups (all p > 0.05). After the course, the AI group achieved higher image-interpretation test scores than the control group (84.3 ± 6.9 vs. 78.1 ± 7.8, p = 0.002) and higher scores on the training-context-specific OSATS-TEE assessment (121.5 ± 9.2 vs. 112.4 ± 10.1, p < 0.001). The key-view acquisition rate was significantly higher in the AI group (88.2% ± 8.1% vs. 76.5% ± 9.7%, p < 0.001), while mean operation time was shorter (27.1 ± 3.9 min vs. 32.4 ± 4.6 min, p < 0.001). Self-efficacy (87.2 ± 7.5 vs. 78.3 ± 8.4, p < 0.001) and course satisfaction (91.0 ± 6.8 vs. 80.5 ± 7.6, p < 0.001) were also higher in the AI group.

An AI-based TEE simulation training system was associated with improved residents' simulation-based TEE learning outcomes, including image-interpretation performance, procedural assessment scores, efficiency, and perceived self-efficacy, compared with traditional teaching. However, further studies are needed to determine whether these gains translate into clinical performance in perioperative practice.