Abstract
Dedicated training for [Formula: see text] laparoscopic camera navigation remains limited by the cost of high-fidelity simulators and the absence of continuous objective assessment. We present SECMA, a portable virtual reality platform combining a mechanically constrained interface that reproduces the trocar pivot, decoupled horizon stabilisation, and optical-redirection channels specific to [Formula: see text] camera handling with high-frequency 6-DoF telemetry. To establish construct-related validity evidence, a six-target navigation task was administered to 20 surgeons and 18 medical students across four practice sessions. Three analytical layers were applied. First, under repeated nested cross-validation ([Formula: see text] outer folds), Logistic Regression was the best classifier ([Formula: see text], [Formula: see text]); execution time, path length, and depth-axis velocity were primary discriminating features. Second, Hidden Markov Models ([Formula: see text]) identified an expert-enriched coordination regime occupied 12.9 percentage points more by experts (Cohen's [Formula: see text]), characterised by purposeful pitch-axis exploration versus diffuse yaw-roll noise in novices. Third, linear mixed-effects models confirmed robust expert-novice differences across four kinematic outcomes (all [Formula: see text]) and practice-related improvement in temporal and spatial efficiency, with no significant between-group difference in improvement rate. These convergent findings support SECMA as a scalable and analytically rigorous platform for [Formula: see text] camera navigation training and proficiency-based assessment.