Stage-aware analyzability of perioperative EEG under real-world clinical conditions: a methodological feasibility study

Background Perioperative electroencephalography (EEG) is increasingly used in anesthesiology and perioperative neuroscience research, but its practical analyzability may differ across perioperative stages and across analysis families. We aimed to develop a stage-aware methodological framework for perioperative EEG under real-world clinical conditions. Methods This methodological feasibility study was based on a real-world perioperative EEG cohort of 61 surgical patients. Candidate EEG windows within each subject-stage underwent a uniform preprocessing pipeline including rereferencing, filtering, fixed 2-s epoching, and epoch-level artifact rejection. Analyzable duration after quality control was quantified as good_sec. One principal window per subject-stage was retained using a prespecified automated ranking rule based on highest good_sec, higher retention ratio, and longer raw window length. Two analysis families were prespecified: Level 1 spectral and aperiodic features and Level 2 connectivity-derived network metrics. Stage-specific operating thresholds were defined a priori. Supplementary analyses examined the stability of intraoperative higher-order metrics after fixed 120-s truncation and the role of continuity of retained clean data. Results Raw acquisition was available in 56 of 61 preoperative recordings, 43 of 61 intraoperative recordings, and 56 of 61 postoperative recordings. Median analyzable duration of retained principal windows was 153 s preoperatively, 678 s intraoperatively, and 112 s postoperatively. Level 1 analyses were broadly feasible across stages, with eligible retained windows in 45/56 preoperative, 43/43 intraoperative, and 45/56 postoperative recordings. Level 2 analyses were markedly more stage-dependent: under the main framework, 23/56 preoperative and 41/43 intraoperative windows met higher-order eligibility, whereas no postoperative window met the main threshold; 23/56 postoperative windows met a prespecified exploratory higher-order threshold. Supplementary analyses showed that the intraoperative advantage was not explained by analyzable duration alone and that higher-order feasibility outside the intraoperative stage was further constrained by fragmentation of retained clean data. Conclusions Under real-world clinical conditions, perioperative EEG analyzability was both stage-dependent and layered by analysis family. Spectral and aperiodic analyses were broadly feasible across perioperative stages, whereas connectivity-derived network analyses were most consistently supported by intraoperative data and were substantially more constrained outside the intraoperative stage. A stage-aware framework based on uniform preprocessing, automated principal-window selection, and analysis-specific operating thresholds may provide a practical foundation for future perioperative EEG studies and for matching acquisition design to the intended level of downstream inference.