儿童癫痫对患儿的生长发育及生命安全构成严重威胁，亟需寻找精确、无创且可纵向监测的生物标志物。既往研究证实，具有异常N-糖基化修饰的细胞外囊泡（extracellular vesicles，EVs）可调控多种神经系统疾病，其特征性N-聚糖可作为潜在的诊断标志物。本研究系统比较了三种不同方法分离的EVs特性。结果表明，外泌体纯化过滤柱（exosome purification filter column，EPF）联合超滤法是从大规模临床样本中分离EVs的最佳方法。继而对EVs及血清进行基于基质辅助激光解吸电离飞行时间质谱（matrix-assisted laser desorption ionization time-of-flight mass spectrometry，MALDI-TOF-MS）的糖组学分析，揭示了二者不同的N-聚糖特征。通过一种新颖的两步机器学习模型，我们鉴定出EVs中47种特征性N-聚糖可作为癫痫诊断与分类的生物标志物。这些标志物能有效区分正常对照、局灶性癫痫与全面性癫痫亚型，且其诊断性能优于血清N-聚糖。此外，我们构建了聚糖相关性网络图，揭示了癫痫发生过程中EVs聚糖表达模式的动态变化。综上所述，EVs N-聚糖作为癫痫检测生物标志物展现出良好潜力，为无创诊断与疾病监测提供了新思路。

Childhood epilepsy poses a serious threat to patients’ growth, development, and life safety, creating an urgent need for precise, non-invasive, and longitudinally monitorable biomarkers. Previous studies have confirmed that extracellular vesicles (EVs) with abnormal N-glycosylation modifications regulate various neurological disorders, where characteristic N-glycans serve as potential diagnostic markers. In this study, we systematically compared the properties of EVs isolated via three different methods. The results demonstrated that an exosome purification filter column (EPF) combined with ultrafiltration emerged as the optimal approach for isolating EVs from large-scale clinical samples. Subsequent matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOF-MS)-based glycomic profiling of EVs and serum revealed distinct N-glycan signatures. Utilizing a novel two-step machine learning model, we identified 47 characteristic N-glycans in EVs as biomarkers for epilepsy diagnosis and classification. These biomarkers effectively distinguished between normal, focal, and generalized epilepsy subtypes while also exhibiting superior diagnostic performance compared to serum N-glycan profiles. Furthermore, we constructed a correlation network map of glycans, which highlighted dynamic alterations in the expression patterns of EV glycans during epileptogenesis. Taken together, the N-glycans of EVs exhibit promising potential as biomarkers for epilepsy detection, offering new insights into non-invasive diagnosis and disease monitoring.