High bandwidth EEG data hosting and analytics in cloud-based storage systems

近年来，随着大数据的出现，特别是在从可穿戴式监控设备收集的生物特征数据领域，对用于存储和高带宽动态数据分析的基于云的解决方案(包括大规模分布式存储系统)的需求越来越高。这种生物特征数据的例子是脑电(EEG)数据，这些数据可从EEG可穿戴设备获得，并通过提供对密集阵列、高质量、原始EEG数据的访问，为情景研究应用提供新的机会。因此，在基于头皮的设备(大脑的“动态脑电监测器”)中出现高频脑电采样将有助于医生诊断、评估和治疗各种神经疾病，并有助于实现新的发现。特别是，我们的行业合作伙伴Avertus开发了一款用于高频采样的耳机，该耳机在非医院环境下佩戴起来既舒适，又非常准确，具有预测问题和实时提供治疗的长期视野。然而，对于即时收集大量生物特征数据的IT公司来说，一个关键问题是他们需要实时解决收集到的数据中的异常检测问题。在我们的EEG头戴式耳机中，高频采样会生成大量需要大量处理的原始数据(例如，估计为5 GB/患者/天)，这就需要一个用于研究(发现和存储库)和临床实用的分布式处理平台--最初是批量处理，但也包括流分析功能。在这项工作中，我们建议开发基于声光的数据分析和存储平台，用于托管和处理来自Avertus头戴式耳机的流数据。重点放在两个方面：(A)对来自Avertus用户的流动数据进行即时生物测定数据分析和异常检测，以及(B)即时分析计算机系统数据，以检测与系统行为有关的异常情况。我们预计，拟议的解决方案将最大限度地减少存储数据的数据分析搜索空间。我们将与合作伙伴组织合作，针对各种分布式存储系统上的人体和系统生物识别数据类型对我们的解决方案进行评估。

In latest years, with the advent of Big Data, especially in the area of biometrics data collected from wearablemonitoring devices, there is a high need for Cloud-based solutions (include large scale distributed storagesystems) for storage and high bandwidth on-the-fly data analysis. Examples of such biometric data are theElectroencephalography (EEG) data that become available from EEG wearable devices and offer newopportunities for contextualized research applications by providing access to dense array, high quality, rawEEG data. Thus, the emergence of high frequency EEG sampling in a scalp based device (a "Holter monitor"for the brain) will assist physicians in diagnosing, assessing and treating a myriad of neurological conditionsand help enable new discovery. In particular, our industry partner, Avertus, has developed a headset for highfrequency sampling that is both comfortable to wear in non-hospital settings and also highly accurate, with along term vision to predict problems and deliver treatments in real-time. However, a key problem for ITcompanies that collect large amounts of biometrics data on-the-fly is their need for real-time solutions foranomaly detection in the collected data. In our case of the EEG headset, high frequency sampling generateslarge amounts of raw data (e.g., an estimate of 5GB/patient/day) that requires significant processing,necessitating a distributed processing platform for both research (discovery and repository) and clinical utility -initially on a batch basis, but also including streaming analytic capability. In this work we propose to develop acloud based data analytics and storage platform for hosting and processing streaming data from Avertusheadsets. The focus is on two aspects: (a) on-the-fly biometric data analysis and anomaly detection ofstreaming data from the Avertus users, and (b) on-the-fly analysis of computer system data for anomalydetection with regards to the system behaviour. We expect that the proposed solution will minimize the dataanalysis search space of the stored data. We will evaluate our solutions in collaboration with the partnerorganization for types of human and system biometrics data on various distributed storage systems.