Ordinary-Kriging Based Real-Time Seizure Detection in an Edge Computing Paradigm

To best of the authors' knowledge, this is the first work that uses Kriging for early detection of seizure. The threat of an epileptic seizure to the life of a patient is both social and physical. Seizure detection research has been developing over the years. Most efforts in the literature concentrate largely on accuracy and would often take computation to the cloud. However, there is a very short window between the onset of an epileptic seizure and a potentially fatal incidence that could lead to injury or loss of life. Hence, there is need for a more time-sensitive approach to seizure detection. Here, we propose a real time seizure detection model in an edge computing paradigm using signals collected through the electroencephalogram (EEG) from the brain of both healthy and epileptic patients. The fractal dimensions of the signals were taken after de-noising with the Discrete Wavelet Transform (DWT), and then classified using the Ordinary Kriging method which gives a training accuracy of 99.4% and a perfect sensitivity. The proposed model was validated with a hardware implementation using an edge computing device and the results show a comparable classification accuracy and a lower mean detection latency of 0.85 sec.