Automated class-based compression for real-time epileptic seizure detection

The emergence of next generation wireless networking technologies has motivated a paradigm shift in development of viable mobile-Health applications for ubiquitous real-time healthcare monitoring. However, remote healthcare monitoring requires continuous sensing for different biosignals and vital signs which results in generating large volumes of data that requires to be processed, recorded, and transmitted. In this paper, we propose our vision for the benefits of leveraging edge computing for enabling automated real-time epileptic seizure detection. In particular, we propose an adaptive classification and data reduction technique that reduces the amount of transmitted data, according to the class of patients, while enabling fast emergency notification for the patients with abnormality. Using such an approach, the patient data aggregator can automatically reconfigures its compression threshold based on the characteristics of the gathered data, while maintaining the required application distortion level. Our results show the excellent performance of the proposed scheme in terms of classification accuracy and data reduction gain, as well as the advantages that it exhibits with respect to state-of-the-art techniques.