多序列MRI影像组学联合临床特征预测儿童LEAT术后癫痫复发
本研究针对接受大体全切除的低级别癫痫相关肿瘤（LEAT）患儿，开发一种整合临床特征与多序列MRI影像组学的预测模型，用于预测术后癫痫复发。传统影像组学模型中多序列联合优于单序列，XGBoost获得最高AUC 0.889和准确率0.816。整合术前癫痫病程可进一步提升模型效能。
要点
- 多序列联合影像组学模型优于单一序列模型，阅片时应注意综合多序列信息。
- 术前癫痫病程是提升预测效能的关键临床特征，报告中宜记录该信息。
- XGBoost模型在传统影像组学中表现最佳，可供后续模型选择参考。
核心结论
- 多模态MRI影像组学结合临床特征的联合模型对儿童LEAT患者大体全切除术后癫痫结局具有预测潜力。
研究设计与样本
- 研究类型为预测模型开发研究，纳入接受大体全切除的儿童低级别癫痫相关肿瘤患者。
- 采用多序列MRI影像组学方法，并与传统影像组学模型及深度学习衍生影像组学模型进行比较。
诊断标准与关键数值
- 传统影像组学模型中，多序列联合优于单序列，XGBoost获得最高AUC为0.889，准确率为0.816。
- 整合术前癫痫病程可显著提升模型效能。
临床应用与阅片要点
- 对于拟行大体全切除的儿童LEAT患者，可结合多序列MRI影像组学特征与术前癫痫病程评估术后癫痫复发风险。
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）。
英文原文（摘要）· 点击展开
OBJECTIVE: This study aimed to develop a predictive model integrating clinical features and multisequence MRI radiomics to forecast postoperative seizure outcomes in pediatric patients with low-grade epilepsy-associated tumors (LEATs) who underwent gross total resection (GTR).
METHODS: In this study, we propose a novel radiomics-based approach to predict seizure recurrence. The model was further optimized by integrating clinical features, and its performance was compared with traditional radiomics models and deep learning-derived radiomics models.
RESULTS: For traditional radiomics models, multi-sequence combination (Combined) outperformed single sequences, with XGBOOST achieving the highest AUC (0.889) and accuracy (0.816). Integrating preoperative epilepsy duration significantly improved model efficacy.
CONCLUSION: The combined model of multimodal MRI radiomics and clinical features demonstrates potential for predicting postoperative seizure outcomes in pediatric LEAT patients after GTR.
引用
- Tang, Matthew, Wu, et al. Integrating multisequence radiomics and clinical features to predict seizure recurrence after gross total resection of pediatric low-grade epilepsy-associated brain tumors. Neuroradiology. 2026. PMID: 42230409. doi:10.1007/s00234-026-04051-3
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。