Epilepsey and Insomnia Detection Using EEG signals
The aim of the project is to develop a real time hybrid ML model capable of taking patient’s real time EEG data files, and predict the presence of dangerous epileptic seizure waves and insomnia disease using real time and internet datasets.
There are nearly 10 lakh cases recorded globally every year for this disease. People who are suffering from seizures may cry out, fall, shake or jerk, and become unaware of what’s going on around them. Preventing from such condition is very important. Soft computing methods are used for predicting epileptic seizures from Electroencephalograms (EEG) signals so that appropriate medication can be suggested.
Insomnia is defined as difficulty in either falling or staying asleep that is accompanied by daytime impairments related to those sleep troubles, People of any age may experience insomnia. That said, infants less than 6 months old are rarely diagnosed with insomnia since it is common for them to not sleep through the night. The American Academy of Sleep Medicine categorizes insomnia into different types: chronic insomnia disorder, short-term insomnia disorder, and other insomnia disorder.
Presently, the signals are manually recognized and classified. The software tool through which this condition can be predicted and identified, the software tool basically provides an interface for doctors to pass the EEG Signals and simplify the overall seizure prediction process. The study proposes a method to predict epileptic seizure and insomnia by analyzing electroencephalogram(EEG) signals. These signals are recorded using a Ultracortex Mark IV headset. Machine learning methods like K-Nearest Neighbour(KNN), Support Vector Machine (SVM) and Random Forest and a Deep Learning(DL) model were employed. The highest accuracy obtained was >98% for epilepsy and >80% for insomnia.
We propose hybrid models with two techniques using stacking and ensemble vote classifier with different permutations of three algorithms which show good results with EEG data i.e. SVM, KNN, Random Forest and the hybrid model using Random Forest and KNN as sub-classifier and SVC as meta-classifier gives accuracy over 98% and for insomnia disease we use EEG data and additional features regarding human lifestyle and performing two step prediction gives accuracy over 81%.
UI Front End
