Intelligent VR Simulator for 30° Laparoscopic Camera Navigation Training: Construct Validation via Machine Learning and Latent State Models

Abstract This study presents the design, implementation, and construct validation of a novel virtual reality (VR) simulator - SECMA - for laparoscopic camera navigation using a 30-degree laparoscope. Central to this system is a custom-designed mechanical interface that transforms Oculus Quest 2 controllers into a physically constrained laparoscopic device, enabling realistic emulation of pivot-based instrument handling. The simulation architecture allows horizon stabilization during camera control and optics pointing by implementing independent rotational kinematics in Unity, accurately replicating real-world laparoscopic dynamics. To assess construct validity, a structured camera navigation task was developed, requiring users to sequentially center on six spatial targets while maintaining visual horizon alignment. Spatial trajectories, composed of 3D positional data and quaternion-based rotations, were captured during four practice sessions per participant. Interpolation and engineered features—such as displacement metrics, angular statistics, and dynamic motion indicators—were extracted from time-series data for classification and modeling. Multiple machine learning classifiers, including logistic regression, support vector machines, and random forests, were trained to distinguish between expert and novice performance, with the best models achieving robust discriminative accuracy. Hidden Markov Models (HMMs) revealed interpretable latent behavioral states, with state ``3'' showing higher visitation frequency among experts. Feature distributions within this state further delineated performance gaps. Additionally, linear mixed models (LMMs) demonstrated that experts benefited more from repeated practice, while novices exhibited stagnant progression over three sessions. These results highlight SECMA potential as a cost-effective and scalable tool for surgical training, integrating advanced simulation, trajectory analysis, and AI-driven feedback to support precision skill acquisition in minimally invasive procedures.