Electroencephalogram-based Objective Assessment of Surgical Expertise during Microvascular Anastomosis Training

抄録

Electroencephalogram-based objective assessment of surgical skill remains underexplored in neurosurgical training. We investigated whether electroencephalogram spectral features can discriminate surgical expertise during microvascular anastomosis training and identify neural markers associated with task performance. We conducted a prospective observational study in 29 right-handed male participants (neurosurgeons, residents, and medical students) performing a 10-minute interrupted anastomosis on a 1.0-mm hydrophilic tube under an operating microscope. Electroencephalogram under the 10-20 system was recorded during the task, segmented into 3 procedural phases (needle insertion, knot tying, thread cutting), and epoched into 2-second windows. Participants were stratified post hoc into Skilled (≥6 units) and Less-skilled (≤5 units) groups based on task throughput. Power spectral density from 4 to 22 Hz was summarized into 6 frequency bands across 7 regions of interest, yielding 42 features per epoch. Classification models were trained using a nested cross-validation framework with a leave-one-subject-out outer loop, and epoch-level predictions of Skilled-class probability were aggregated into subject-level scores, from which phase-specific area under the receiver operating characteristic curve values were calculated. A total of 7,201 artifact-free epochs (76.3%) were included. Subject-level discrimination performance varied by phase, with area under the receiver operating characteristic curve values of 0.736 for needle insertion, 0.678 for knot tying, and 0.788 for thread cutting. Feature attribution and group-level analyses consistently revealed increased frontal alpha power and reduced midline theta power in Skilled participants, particularly during needle insertion and thread cutting. These findings support the feasibility of electroencephalogram-based objective assessment of microsurgical expertise and identify candidate neural markers of proficiency.