Multimodal Detection of Early Partial Seizure Onsets

描述(由申请者提供)：对于几种抗癫痫药物失败且不适合手术的患者，闭环式治疗(神经刺激、降温和药物治疗)作为癫痫发作治疗的替代方法重新引起了人们的兴趣。响应性闭环治疗旨在癫痫发作后几秒钟内进行干预，以终止癫痫活动或防止演变为致残性癫痫。对闭环治疗的挑战是要求非常早的检测，同时仍然保持特异性，所以治疗可以足够早地给予，并且仍然是安全的。设计更具体的检测方法的一个限制是，部分癫痫发作的特征不像原发全身性癫痫发作的特征那样明确。尽管部分性发作的发作对同一患者来说可能有些刻板印象，但总体上部分发作的发作通常更微妙、更缓慢地演变，最重要的是，更不同的。考虑到发作模式的这种变异性，可以使用多模式方法来设计更适应的检测范式，该方法将评估部分癫痫发作时颅内脑电特性的变化，包括通常观察到的频率内容的变化，以及复杂性和熵的变化。然后，发作变化的特征可以是同一患者所有癫痫发作的典型发作成分，以及不同发作之间变化的不同发作成分。结合基于模式识别的方法评估刻板印象成分和基于新颖性检测的方法评估癫痫发作的变异成分，可以提高部分发作发作的早期检测的可靠性。改进的癫痫早期检测不仅有利于闭环治疗的应用，而且可以为患者提供可靠的早期预警系统。

DESCRIPTION (provided by applicant): There is renewed interest for closed-loop therapy (neurostimulation, cooling, and medication) as an alternative method of seizure treatment for patients who have failed several antiepileptic drugs and are not surgery candidates. Responsive closed-loop therapy is designed to intervene within seconds after seizure onset to terminate seizure activity or prevent evolution to a disabling seizure. The challenge to closed-loop therapy is the requirement of very early detection while still retaining specificity, so the therapy can be administered early enough and still remains safe. One limitation of designing more specific detection methods is that partial seizure onset features are not as clearly established as those of primary generalized seizures. Although onsets of partial seizures can be somewhat stereotypical for the same patient, overall partial seizure onsets are typically more subtle, more slowly evolving and above all, more heterogeneous. Taking into account this variability of onset patterns, the design of a more adaptive paradigm of detection may be achieved using a multimodal approach which will assess the changes of intracranial EEG properties at the onset of partial seizures including commonly observed changes in frequency content but also changes in complexity and entropy. The ictal changes can be then characterized by the stereotypical ictal components occurring for all seizures in the same patient and the variant ictal components that are variable from seizure to seizure. Combinations of pattern recognition based methods to assess the stereotypical components and novelty detection based methods to assess the variant components of a seizure can improve the reliability of the early detection of partial seizure onsets. Improved early seizure detection can benefit not only the application of closed loop therapy, but also could be used for a reliable early warning system for patients.