Smartphone based Early Detection of Epileptic Seizures Using Machine Learning

This research paper proposes early detection of epileptic signal using wearable sensor and signal is transmitted to smartphone. In smartphone, by using machine learning algorithm which works on EEG signal and after processing, it with machine learning model there is detection of preictal state and immediately emergency SMS is sent to nurse/doctor mobile so that drug can be injected in the body of the patient before the actual attack happens. Tremors of brain diseases like Epilepsy, Parkinson's and depression can be identified and diagnosed using machine learning methods. Due to deficiency of dopamine chemical in Substantia Nigra region of the brain, it results in epileptic seizures, which if predicted before the onset can be controlled through medication. Thus, Computational methods and machine learning methods involves prediction of epileptic seizures from Electroencephalograms (EEG) signals. But removal of noise and extracting features are two major challenges for prediction of epileptic seizures. Thus, in the upcoming research paper we propose a model which helps in epileptic seizures prediction sufficient time before seizures start. Support Vector Machine algorithm has been proposed as suitable machine learning model which finds it's application in preprocessing and train the model by extracting frequency and time domain features. Preictal state is detected through proposed model, which starts couple of minutes before the seizure onset, and accuracy is 92.2% and average time taken for prediction is 92.23 minutes by taking samples from 22 subjects using wearable sensors.