The ultimate Analytics Framework for Integrated and Personalised Healthcare Services

AEGLE’s first prototype has been deployed on the oceanos IaaS (Infrastructure as a Service) public cloud infrastructure, which is used as the AEGLE’s cloud testing platform, where it is intended to work on sample dataset and not the overall data of AEGLE. Once this first prototype will complete internal validation, it will be re-deployed on a private IaaS cloud provider to enable SLA (Service Level Agreement) and more robust performance guarantees.

AEGLE system architecture enables aggregation of healtcare data from both the local and cloud domains. Local users are able to upload data to AEGLE cloud, passing through the anonymization modules or to directly perform complex cloud level analysis on either new data or data already available online.

Key features of AEGLE’s first prototype:

  • Scalable, anonymized and secure data storage: Data storage exploits FedEHR service and SQL cluster technology combining a anonymized uploads with high-level clinical modelling abstraction and with a big data technology to provide an intuitive data design and management interface to securely interoperate, integrate, store, manage and share medical information across the health care enterprise, while preserving patient confidentiality and data access rights at any time.
  • Scalable big data analytics: A big data analytics framework operates on top of the cloud infrastructure, providing interfaces and mechanisms for scalable analytic services and the lightweight resource management of the AEGLE cluster nodes. The implemented big data analytics framework supports dockerized instantiations for two state-of-art scalable execution engines, i.e. Hadoop MapReduce and SPARK. Integrated analytics have been developed for i) the Chronic Lymphocytic Leukemia (CLL)
    case concerning the analysis of immunogenetic data, and ii) the Intensive Care Unit (ICU) case concerning PVI specific analysis, general biosignal preprocessing for artifact rejection, feature extraction and dimensionality reduction tools for robust predictive modelling.
  • Web service oriented approach: A set of proper REST API intefaces have been provisioned to enable the integration between AEGLE’s user application and the big data framework at the backend. The REST APIs have been exposed at the big data framework level and exploited by the user application to enable navigation and selection from available online datasets as well as invocation of the used analytics.
  • Develop once deploy everywhere: All software components (user interfaces, big data analytics framework and AEGLE storage) captured within persistent Docker containers. This enables the implementation of a fully re-Locatable AEGLE infrastructure, e.g.~oceanos IaaS cloud or Amazon EC or even local clusters. In addition, it enables fast activation/deactivation of virtual cluster nodes, thus supporting effficient cluster resizening according to workload requirements.

Dr. Sotiris Xydis

Institute of Communication
and Computer Systems