Development and evaluation of a virtual dissection environment for anatomical learning in the endoscopic endonasal approach.

Abstract

Objective: A clear 3D understanding of complex skull base structures, including the cavernous sinus (CS), is vital for the endoscopic endonasal approach. However, traditional learning (TL) methods using textbooks and static materials have limits in fostering spatial comprehension. This study developed an interactive virtual dissection (VD) environment based on a virtual endoscopic skull base anatomy 3D computer graphics (VESA-3DCG) model to enhance 3D understanding of the sellar and parasellar regions and evaluated its educational effectiveness against TL methods.

Methods: The VESA-3DCG model was constructed by modifying previously developed high-fidelity 3DCG models, which were designed with reference to the authors' previous anatomical studies of the sellar and parasellar regions, and integrated into a VD environment. Twenty-eight Japanese neurosurgical residents (postgraduate years 3-7) were randomly assigned to the VD or TL group. Both learning sessions were conducted remotely via a screen-sharing platform, allowing participants to view and interact with the presented materials in real time. A knowledge test covering four domains-bony landmarks, CS anatomy, microvascular anatomy, and neural anatomy-was administered before and after learning. Gain scores, defined as pre- to posttest improvement, were calculated per domain and overall. Group comparisons were performed to assess learning outcomes, and satisfaction and confidence were rated on a 5-point Likert scale.

Results: The final model, consisting of 304 components and about 18.6 million polygons, accurately depicted the microanatomy of the sellar and parasellar regions. The VD environment supported interactive manipulation, including transparency and translucency control, rotation, zooming, virtual drilling, and retraction. Gain score analysis showed that the VD group achieved greater overall improvement in anatomical learning relative to the TL group (p = 0.036), with the most robust difference observed in the CS anatomy domain (p = 0.001). Within-group analysis in the VD group confirmed notable posttest gains in bony landmarks, CS anatomy, and neural anatomy. Participants reported high satisfaction and confidence with the VD environment.

Conclusions: The VD environment based on the VESA-3DCG model offered an effective, interactive platform for anatomical learning. It demonstrated favorable educational effects, particularly for anatomically complex regions such as the CS, and showed feasibility as a complementary tool to TL, including in remote education settings.