(c) 2018 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other users, including reprinting/ republishing this material for advertising or promotional purposes, creating new collective works for resale or redistribution to servers or lists, or reuse of any copyrighted components of this work in other works.

The continuous adoption of fitness and medical smart sensors are boosting the development of Internet of Medical Things (IoMT), reshaping and revolutionizing Healthcare. This digital transformation is paving the way to new forms of care based on real-time analysis of huge amounts of data produced by sensors, which is seen as a basis for improving clinical efficiency and helping to save lives. A medical sensor typically produces several KBs of data per second so the collection and analysis of these data can be approached with Big Data technologies. The aim of this paper is to present and evaluate a hybrid architecture for real-time anomaly detection from data streams coming from sensors attached to patients. The architecture includes an edge computing data staging platform based on Raspberry Pi 3 for data logging, data transformation in RDF triple and data streaming towards a cluster computing running Apache Kafka for collecting RDFStreams, Apache Flink for running a parallel version of the Hierarchical Temporal Memory algorithm and Cassandra for data storing. The different layers of the architecture have been evaluated in terms of both CPU performance and memory usage using the REALDISP dataset.

Peer Reviewed