REAL- TIME MINING TECHNIQUES - LATENCY MITIGATION ALGORITHMS FOR SEIZURE DETECTION

Abstract

Speed is the name of the game in today’s world. Speed of change drives our everyday’s life, and hence in this
ever changing and agile environment it has become critical for systems to extract useful information in real
time. Techniques used in sifting through streaming data to evaluate and extract useful information are called
real time data analytics. The ability of these techniques to paint the picture of the changes as and when they
happen makes them a vital tool in analysis and visualization of medical data especially of the neurological
nature as it is very dynamic. The fact that all the evaluation and analysis needs to be done on the go make the
challenges enormous. However, this very challenge if successfully met can be used to analyze neurological data
related to disorders such as epilepsy, Alzheimer’s, strokes and even headaches. According to WHO these
ailments affects close to I billion people worldwide. However it is extremely difficult for human analysts to
collate, evaluate, manage and assimilate large volumes of data related to such alements. This is where real time
data analytic can step in to meet the challenges. Already we have CAD (Computer-Aided Diagnosis) that
automatically detects neurological abnormalities like seizures using medical big data. The next big step will be
to bring in real time analytics that have computational algorithms with the capability of predicting periods of
high seizure probability with reliability with minimum latency. This not only helps patients from enjoying life
threatening activities during such periods, but would also regulate medication to be used only during high
seizure probability windows. This paper attempts to highlight a big data processing framework under the cloud
computing platform for real time seizure prediction with the minimum of time delays.