Li Adam, Sarma Sridevi V, Fitzgerald Zachary, Hopp Jennifer, Johnson Emily, Crone Nathan, Bulacio Juan, Martinez-Gonzalez Jorge, Inati Sara, Zaghloul Kareem
Annu Int Conf IEEE Eng Med Biol Soc. 2019 Jul;2019:2328-2331. doi: 10.1109/EMBC.2019.8856591.
Cortical stimulation mapping (CSM) is a common clinical procedure for mapping eloquent cortex in epilepsy patients. Electrical responses to the stimulation, or after-discharges (ADs), that occur in response to stimulation can point to unstable regions of cortex that are more prone to spontaneous seizures. Clinicians are interested in identifying regions that start seizures, i.e., the epileptogenic zone (EZ), so that they can target treatment. However, during CSM, not all regions are stimulated, as it would be time-consuming and potentially harmful to the patient. This limits the clinician's ability to fully explore ADs to reliably localize the EZ. In this paper, we develop a virtual CSM procedure that processes pre-seizure intracranial EEG recordings obtained from epilepsy patients being treated at three different epilepsy centers. First, we identify a linear time varying network (LTVN) model from electrocorticography (ECoG) and stereo-EEG (SEEG) data using sparse least squares estimation for each patient. We then construct an virtual CSM by applying impulse perturbations to each electrode contact in the LTVN model and then measuring the ADs of the network. We summarize the l2-norm of the responses in the form of a heatmap that shows the spatio-temporal evolution of the ADs before, during, and after seizures. Finally we compute an impulse response ratio (IRR) metric from each heatmap, that measures the ratio between the mean norm of ADs of clinically annotated EZ contacts and the mean norm of ADs of the remaining contacts. We find that the IRR is higher in maps derived from patients with successful surgical outcomes and lower in failed surgical outcomes. This suggests that virtual CSM may provide valuable information to clinicians regarding EZ location.
皮层刺激图谱（CSM）是癫痫患者中用于绘制明确皮层的常见临床程序。对刺激产生的电反应或放电后反应（ADs），可以指向更易发生自发性癫痫发作的不稳定皮层区域。临床医生有兴趣识别引发癫痫发作的区域，即癫痫病灶区（EZ），以便确定治疗靶点。然而，在CSM过程中，并非所有区域都受到刺激，因为这对患者来说既耗时又可能有害。这限制了临床医生充分探索ADs以可靠定位EZ的能力。在本文中，我们开发了一种虚拟CSM程序，该程序处理从三个不同癫痫中心接受治疗的癫痫患者发作前的颅内脑电图记录。首先，我们使用稀疏最小二乘法估计，从每个患者的皮层脑电图（ECoG）和立体脑电图（SEEG）数据中识别线性时变网络（LTVN）模型。然后，我们通过对LTVN模型中的每个电极触点施加脉冲扰动，然后测量网络的ADs，构建一个虚拟CSM。我们以热图的形式总结反应的l2范数，该热图显示了癫痫发作前、发作期间和发作后的ADs的时空演变。最后，我们从每个热图中计算一个脉冲反应比（IRR）指标，该指标测量临床标注的EZ触点的ADs平均范数与其余触点的ADs平均范数之间的比率。我们发现，在手术成功的患者所得到的图谱中IRR较高，而在手术失败的患者中IRR较低。这表明虚拟CSM可能为临床医生提供有关EZ位置的有价值信息。
