Analyzing dimensionality reduction with softmax discriminant classifier for epilepsy classification

摘要

Due to the unpredictable interruptions in the normal brain functions, recurrent seizures occur and this type of disorder is termed as epilepsy. The motor, sensory and other autonomic functions of the brain are severely affected by seizures. Also the memory, state of consciousness, emotional behaviour is equally affected by seizures. A most efficient and versatile equipment for the diagnosis of this syndrome is Electroencephalography (EEG). In the EEG signal, 2 kinds of abnormal activities can be sharply observed as ictal and interictal. Ictal activities happen when an epileptic seizure occurs and interictal activities happen in between the seizures. To avoid the misdiagnosis, the patient's ictal EEG is quite significant for the analysis. Due to the unforeseen and sudden occurrence of seizures, there is a huge difficulty in recording the EEG signals. The long term EEG recordings which are continuous in nature has a lot of data to be processed and so dimensionality reduction techniques like Factor Analysis and Singular Value Decomposition (SVD) are utilized to reduce the dimensions. The dimensionally reduced values are then classified with the help of Softmax Discriminant Classifier (SDC) for epilepsy classification from EEG signals. Results show that when Factor Analysis is classified with SDC, an average classification accuracy of 95.04% is obtained and when SVD is classified with SDC, an average classification accuracy of 96.42% is obtained.