Normative Brain Mapping of Scalp EEG to Localise Epileptic Foci for Epilepsy Surgery

引言和背景癫痫是一种相对常见的致残性神经系统疾病（de Boer et al. 2008），其特征为持续性和无端癫痫发作（Beghi，2020）。据估计，全世界有5000万患者患有各种形式的癫痫，约三分之一的癫痫患者无法通过药物控制癫痫（Kwan and Brodie et al. 2000）。现有药物在某些情况下有效。然而，对于耐药性局灶性癫痫患者，脑部手术是实现无癫痫发作和改善生活质量的最佳机会（Fitzgerald et al. 2021）。考虑手术是一个高度选择性的过程。它涉及临床评价和使用几种成像方式，以确定癫痫灶和确定癫痫类型。此外，很大一部分患者存在复杂病例或不同模式之间的矛盾结果，不考虑手术（Engel等，2003）。此外，尽管影像学和手术经验取得了进步，但许多接受癫痫手术的候选人无法实现完全的癫痫发作自由（Nowell等人，2014）。在这种情况下，大多数情况下，癫痫灶没有被准确识别，导致术后癫痫复发。总的来说，临床结果的矛盾和癫痫灶的定位不良构成了一个大问题。因此，提高不同成像模式之间的一致性以加强手术成功率至关重要。用于癫痫评估的非侵入性技术包括：头皮脑电图（EEG）、磁共振成像（MRI）、脑磁图（MEG）、正电子发射断层扫描（PET）等（Fitzgerald et al. 2021）。同样，一些候选人需要使用颅内脑电图（iEEG）等侵入性技术进行更详细的分析和计划，这可能会导致手术并发症（Arya等人，2013，Blauwblomme等人，2011）。由于可访问性和成本问题，这些技术中的许多在地理上是不受欢迎的。因此，癫痫手术在发展中国家的使用仍然有限（Wieser，1998）。或者，头皮EEG是一种低成本、多功能和便携式成像工具，可提供出色的时间分辨率和与其他成像方式的有效协作（Fitzgerald等人，2021; LaRocco等人，2020; Michel和Brunet，2019）。该项目的主要目标是研究癫痫病灶的定位使用头皮脑电图在癫痫手术，通过计算大脑建模，这有可能被用于worldwide.Data，方法和ProgressThe预先收集的数据将用于这个博士是由与UCL和UCLH的合作伙伴关系提供。EEG数据集包括62名患者和18名健康对照者，每个候选者还有5-10个片段。这些片段包括一系列静息状态，同时记录EEC和fMRI的任务，激光发射5.84-1887.86秒。大多数段包括32或64个通道，采样率为5000 Hz。

Introduction and BackgroundEpilepsy is a relatively common disabling neurological disorder (de Boer et al. 2008), characterised by persistent and unprovoked seizure generation (Beghi, 2020). An estimated 50 million patients worldwide suffer from all forms of epilepsy, and approximately one third of epilepsy patient population do not obtain seizure-control through medication (Kwan and Brodie et al. 2000).Available medication is effective in some cases. However, for patients with pharmacoresistant focal epilepsy, brain surgery holds the best chance of achieving seizure freedom and improving quality of life (Fitzgerald et al. 2021). Consideration for surgery is a highly selective process. It involves a clinical evaluation and the use of several imaging modalities in order to identify the epileptic foci and to determine epilepsy type. In addition, a significant portion of patients present with complex cases or contradicting results between different modalities, and are not considered for surgery (Engel et al. 2003). Furthermore, despite advances in imaging and surgical experience, many candidates who undergo epilepsy surgery do not achieve complete seizure freedom (Nowell et al. 2014). Most often in this case, epileptic foci were not accurately identified, leading to seizure recurrence postoperatively. Collectively, contradicting clinical results and poor localisation of the epileptic foci constitutes a big problem. It is therefore vital to improve the concordance between different imaging modalities, to strengthen surgical success.Non-invasive techniques used for epilepsy evaluation include; scalp electroencephalography (EEG), magnetic resonance imaging (MRI), magnetoencephalography (MEG), positron emission tomography (PET) and more (Fitzgerald et al. 2021). Similarly, some candidates require a more detailed analysis and planning using invasive techniques such as intracranial electroencephalography (iEEG), which may propose surgical complications (Arya et al. 2013, Blauwblomme et al, 2011). Many of these techniques are geographically undesirable due to accessibility and cost issues. Therefore, the use of epilepsy surgery in developing countries remains limited (Wieser, 1998). Alternatively, scalp EEG is a low-cost, versatile and portable imaging tool that offers excellent temporal resolution and effective collaboration with other imaging modalities (Fitzgerald et al. 2021, LaRocco et al, 2020, Michel and Brunet, 2019). The key goal of this project is to study localisation of epileptic foci using scalp EEG in epilepsy surgery, via computational brain modelling, which has potential to be used worldwide.Data, Methodology and ProgressThe pre-collected data to be used for this PhD is provided by partnership with UCL and UCLH. The EEG dataset clnsists of 62 patients and 18 healthy controls, with further 5-10 segments per each candidate. The segments include a range of resting-state, task of simultaneous EEC and fMRI recordings, lasing 5.84-1887.86 seconds. Most of the segments include 32 or 64 channels and a sampling rate of 5000Hz.