Electrophysiological Source Imaging of Partial Epilepsy

项目摘要 该研究计划的长期目标是开发和建立一种新型的电生理源 用于定位致癫痫脑组织并对其成像的成像技术，有助于病灶（部分）的术前计划 癫痫癫痫是一种常见的神经系统疾病，在美国有340多万患者， 全球百万。标准的临床常规严重依赖于使用植入到脑内的颅内EEG（iEEG）。 大脑，以确定癫痫发作区，以帮助定位致痫区（EZ），尽管有限的 iEEG电极的覆盖范围和多日手术的侵入性。临床上需要 开发一种基于电生理记录的非侵入性神经成像方法， 并从相关的癫痫生物标志物中描绘出EZ。在这个研究项目中，我们建议建立一个新的 无监督的机器学习框架，将能够处理无标记的，连续的，长期的 局灶性癫痫患者的EEG记录，以检测高频振荡（HFO），自动识别 病理性HFO（pHFO，即，HFO骑在尖刺上），并定位和图像癫痫脑活动， 已识别的pHFO。所提出的技术将在120多名局灶性癫痫患者中进行严格验证 与iEEG记录和手术切除结果的临床结果进行对比。我们的具体目标是：目标1。 一种新的无监督机器学习技术的开发和评估，以识别病理性HFO 从连续的脑电图记录。这一目标将建立一种新的无监督机器学习方法， 通过结合数据驱动特征，自动识别源自癫痫活动的pHFO 学习方法目标2.一种新的脑张量分解成像方法的开发和验证 用于HFO源成像。我们将开发一种新的源成像框架成像癫痫源， 时间、光谱和空间域。这一目标将建立一种新的数据驱动的源成像方法， 从头皮记录的pHFO精确映射和定位EZ。目的3：头皮检测pHFO的验证 通过iEEG鉴定的pHFO，与局灶性癫痫患者中EZ的临床表现。我们将检验这个假设， 头皮EEG识别的pHFO事件反映了iEEG识别的pHFO的基本特征，并且两者都是 指示从临床iEEG确定并通过手术切除结果确认的EZ。这一目标将 建立头皮识别的pHFO与iEEG识别的pHFO之间的关系， 癫痫网络拟议研究的成功完成将建立一个新的无监督 机器学习技术从连续和长期头皮EEG中检测和识别病理性HFO 记录，并非侵入性地和准确地定位和成像潜在的致痫区。的 这种新技术的建立有望显著改善焦点药物的临床管理， 抗癫痫，这是目前有限的，受益于众多的患者和医疗保健系统。

Project Summary The long-term goal of the research program is to develop and establish a novel electrophysiological source imaging technology to localize and image epileptogenic brain tissues aiding pre-surgical planning in focal (partial) epilepsy. Epilepsy is a common neurological disease impacting more than 3.4 million patients in the US and 70 million globally. The standard clinical routine heavily relies on using intracranial EEG (iEEG) implanted into the brain to determine seizure onset zone, to aid in the localization of epileptogenic zone (EZ), despite the limited coverage of iEEG electrodes and invasive nature of the multiple-day procedure. There is a clinical need to develop a noninvasive neuroimaging approach based on electrophysiological recordings that can reliably image and delineate the EZ from relevant epilepsy biomarkers. In this research project, we propose to establish a novel unsupervised machine learning framework that will be able to process unmarked, continuous, and long-term EEG recordings of focal epilepsy patients to detect high frequency oscillations (HFOs), automatically identify the pathological HFOs (pHFOs, i.e., HFOs riding on spikes), and localize and image epileptogenic brain activity from the identified pHFOs. The proposed techniques will be rigorously validated in over 120 focal epilepsy patients against clinical findings from iEEG recordings and surgical resection outcomes. Our specific aims are: Aim 1. Development and evaluation of a novel unsupervised machine learning technique to identify pathological HFOs from continuous EEG recordings. This aim will establish a novel unsupervised machine learning approach for automatically identifying pHFOs originated from epileptogenic activity, by incorporating data-driven feature learning methods. Aim 2. Development and validation of a novel brain tensor decomposition imaging approach for HFO source imaging. We will develop a novel source imaging framework imaging epileptic sources in temporal, spectral, and spatial domains. This aim will establish a novel data-driven source imaging approach for accurate mapping and localization of EZ from scalp recorded pHFOs. Aim 3: Validation of scalp-detected pHFOs by iEEG-identified pHFOs, with clinical findings of EZ in focal epilepsy patients. We will test the hypothesis that the scalp-EEG identified pHFO events reflect the essential features of iEEG-identified pHFOs, and that both are indicative of the EZ determined from clinical iEEG and confirmed by surgical resection outcome. This aim will establish the relationship between scalp-identified pHFOs with iEEG-identified pHFOs and the underlying epileptogenic networks. The successful completion of the proposed research will establish a novel unsupervised machine learning technology to detect and identify pathological HFOs from continuous and long-term scalp EEG recordings, and localize and image the underlying epileptogenic zone noninvasively and accurately. The establishment of such novel technology promises to significantly improve the clinical management of focal drug- resistant epilepsy, which is currently limited, benefiting numerous patients and the healthcare system.