记录时长与警觉状态对癫痫高频振荡特征的影响
结果显示7天记录可在98%患者中完整刻画HFO分布,多日聚合的切除比例预测术后结局的AUC达0.86。
要点
- 仅分析NREM睡眠时,21%患者需超过2天记录才能捕获完整HFO分布。
- 所有警觉状态合并时,7天记录可在98%患者中完整刻画HFO分布。
- 基于多日聚合的关键切除百分比(CReP)预测术后结局的AUC为0.86。
核心结论
- 短采样期可能无法完整呈现HFO空间分布,多日记录可提高HFO特征刻画及术后结局预测的可靠性。
研究设计与样本
- 患者平均年龄32.8岁(范围6-66),平均记录时长8.40天(范围3-22),其中女性30例。
主要发现
- HFO与癫痫发作起始区(SOZ)的关联存在于所有警觉状态,但随时间波动,NREM睡眠期时间变异性最小。
- HFO-SOZ关联在颞叶癫痫中更强,在发作频繁患者中更稳定。
- 仅分析NREM数据时,21%患者需超过2天记录才能捕获完整HFO分布,但30%患者全部NREM数据仍不足以描述完整HFO分布。
- 所有警觉状态合并时,7天记录在98%患者中完整刻画HFO分布。
- 多日聚合CReP预测术后结局的AUC为0.86(95%CI 0.70-1.00),优于随机单次5分钟NREM片段(中位AUC 0.42,IQR 0.23)或单日分析(不同分析日AUC范围0.63-0.71)。
临床怎么用
- 癫痫术前评估中HFO定位应优先采用多日记录,避免仅用短时段或单日数据。
- 分析HFO空间分布时可结合NREM睡眠期数据,但需注意部分患者NREM数据仍不充分。
- 基于多日聚合的CReP可用于辅助预测术后结局。
局限性
- 原文未报告研究局限性。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
原文
BACKGROUND AND OBJECTIVES: High-frequency oscillations (HFOs) are a promising biomarker for localizing epileptogenic tissue, yet the extent to which recording duration and vigilance state influence their spatial distributions remains unclear. This study quantified the recording duration and vigilance state required to reliably capture HFO spatial distributions to guide surgical planning in epilepsy. METHODS: We retrospectively analyzed long-term, continuous iEEG recordings from patients with drug-resistant epilepsy undergoing presurgical evaluation at the University of Michigan. Sleep stages were manually annotated, HFOs (80-500 Hz) were detected using a validated algorithm, and the correlations between HFOs and the seizure-onset zone (SOZ) were assessed across vigilance states. A novel similarity-based temporal padding approach was developed to measure the similarity of HFO distributions derived from data available up to a given time point with those derived from the full recording. Postsurgical outcome prediction was evaluated using a decision tree classifier based on the proportion of resected, top-ranked HFO-rate channels (critical resection percentage, CReP). RESULTS: Fifty-four patients were analyzed (30 female patients, mean age 32.8 [range, 6-66] years, mean recording duration 8.40 days [range, 3-22]). HFO-SOZ associations were present across all states but fluctuated over time; non-rapid eye movement (NREM) sleep showed the least temporal variability. HFO-SOZ associations were stronger in temporal lobe epilepsy and more stable in patients with frequent seizures. When analyzing only the HFOs during NREM, 21% of patients required more than 2 days of recording to capture its full distribution; however, all NREM data were insufficient to describe the full HFO distribution in 30% of patients. Across all vigilance states, 7 days of recording fully characterized HFO distributions in 98% of patients. Postsurgical outcome prediction using multiday aggregate CReP achieved more robust and accurate performance (AUC 0.86, 95% CI [0.70-1.00]) than analyses based on a random selection of single 5-minute NREM epoch (median AUC 0.42, IQR 0.23) or a single day (AUCs ranged 0.63-0.71 for different analyzed days). DISCUSSION: Short sampling period risks incomplete representation of the full HFO spatial profile. Future studies should consider multiday recordings to enable more reliable HFO characterization and improve HFO-based surgical outcome prediction.
引用
- Chen Z, Gliske SV, Alsammani A, et al. The Influence of Recording Duration and Vigilance State on High-Frequency Oscillation Characterization in Epilepsy. Neurology. 2026. PMID: 42348802. doi:10.1212/WNL.0000000000218225
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。