Nested Phase-Specific Speed-Deficit Viability for Computer-Assisted Robotic Surgical Training Assessment

Abstract Objective motion-quality assessment in computer-assisted robotic surgical training requires metrics that preserve task-phase structure rather than reducing complete trajectories to global kinematic summaries. This paper introduces Nested Phase-Specific Speed-Deficit Viability (Nested PS-SDVI), an interpretable and leakage-controlled trajectory-level scoring method for dVRK kinematics. For each held-out operator, task-specific good-quality speed reference profiles are estimated from training operators only. Positive speed-deficit exposure is then measured within normalized task phases, while phase-bin resolution, reference quantile, and deficit summary are selected through inner user-level validation. The evaluated operator is excluded from reference construction, thresholding, score standardization, and candidate selection. Technical validation used 46 ROSMA dVRK trials from three operators across Pea-on-a-Peg, Post-and-Sleeve, and Wire-Chaser I tasks. Nested PS-SDVI achieved macro-AUC of 0.762 for poor-quality trial detection and macro-Spearman association of 0.570 with train-standardized within-task ROSMA score, outperforming fixed PS-SDVI variants and raw kinematic baselines. These results support phase-conditioned speed-deficit scoring as a compact assessment layer for computer-assisted robotic surgical training analysis.