Cai Tianxin, Lin Yaoxin, Wang Guofu, Luo Jie
School of Biomedical Engineering, Shenzhen Campus of Sun Yat-sen University, Shenzhen, China.
Key Laboratory of Sensing Technology and Biomedical Instrument of Guangdong Province, Guangdong Provincial Engineering and Technology Center of Advanced and Portable Medical Devices, Sun Yat-sen University, Guangzhou, China.
Front Neurol. 2024 Aug 23;15:1402004. doi: 10.3389/fneur.2024.1402004. eCollection 2024.
The success rate of achieving seizure freedom after radiofrequency thermocoagulation surgery for patients with refractory focal epilepsy is about 20-40%. This study aims to enhance the prediction of surgical outcomes based on preoperative decisions through network model simulation, providing a reference for clinicians to validate and optimize surgical plans.
Twelve patients with epilepsy who underwent radiofrequency thermocoagulation were retrospectively reviewed in this study. A coupled model based on model subsets of the neural mass model was constructed by calculating partial directed coherence as the coupling matrix from stereoelectroencephalography (SEEG) signals. Multi-channel time-varying model parameters of excitation and inhibitions were identified by fitting the real SEEG signals with the coupled model. Further incorporating these model parameters, the coupled model virtually removed contacts destroyed in radiofrequency thermocoagulation or selected randomly. Subsequently, the coupled model after virtual surgery was simulated.
The identified excitatory and inhibitory parameters showed significant difference before and after seizure onset ( < 0.05), and the trends of parameter changes aligned with the seizure process. Additionally, excitatory parameters of epileptogenic contacts were higher than that of non-epileptogenic contacts, and opposite findings were noticed for inhibitory parameters. The simulated signals of postoperative models to predict surgical outcomes yielded an area under the curve (AUC) of 83.33% and an accuracy of 91.67%.
The multi-channel coupled model proposed in this study with physiological characteristics showed a desirable performance for preoperatively predicting patients' prognoses.
难治性局灶性癫痫患者接受射频热凝手术后实现无癫痫发作的成功率约为20%-40%。本研究旨在通过网络模型模拟加强基于术前决策的手术结果预测，为临床医生验证和优化手术方案提供参考。
本研究回顾性分析了12例接受射频热凝治疗的癫痫患者。通过将偏相干性计算为立体定向脑电图（SEEG）信号的耦合矩阵，构建了基于神经团块模型子集的耦合模型。通过将真实的SEEG信号与耦合模型拟合，识别出激发和抑制的多通道时变模型参数。进一步纳入这些模型参数，耦合模型虚拟去除了在射频热凝中被破坏或随机选择的触点。随后，对虚拟手术后的耦合模型进行了模拟。
识别出的兴奋和抑制参数在癫痫发作前后显示出显著差异（<0.05），参数变化趋势与癫痫发作过程一致。此外，致痫触点的兴奋参数高于非致痫触点，抑制参数则相反。用于预测手术结果的术后模型模拟信号的曲线下面积（AUC）为83.33%，准确率为91.67%。
本研究提出的具有生理特征的多通道耦合模型在术前预测患者预后方面表现出良好的性能。
