.NET Project On  Automatic Detection Of Paroxysms In Eeg Signals Using Morphological Descriptors And artificial Neural Networks


The electroencephalogram (EEG) signal plays an important role in the diagnosis of epilepsy. The EEG recordings of the ambulatory recording system produce very lengthy information and the detection of the epileptic activity requires a time-consuming examination of the entire length of the EEG information by an expert. The traditional strategies for the investigation being tedious, many automated diagnostic systems for epilepsy has emerged in recent years.

This project proposes a neural-network-based automated epileptic EEG deduction system that utilizations approximate entropy (ApEn) as the input feature. ApEn is a statistical parameter that measures the predictability of the present amplitude values of a physiological signal based on its past amplitude values. It is realized that the estimation of the ApEn drops sharply during an epileptic seizure and this reality is utilized as a part of the proposed system. Two distinct kinds of neural networks, to be specific, Elman and probabilistic neural networks are considered. ApEn is used for the first time in the proposed system for the detection of epilepsy using neural networks. It is shown that the overall accuracy values as high can be achieved by using the proposed system.


The existing system was done just by manual to recognize epilepsy. The existing system of artificial neural network based detection system for epileptic analysis has been proposed by a few researchers. The technique proposed by Weng and Khorasani utilizes the highlights proposed by Gotman and Wang, specifically, normal EEG abundancy, normal EEG term, the coefficient of variation,  dominant frequency and average power range as contributions to an adaptive structured neural network. The strategy proposed by Pradhan et al. utilizes a crude EEG signal as a contribution to a learning vector quantization network. In 2004, Nigam and Graupe proposed another neural system show called LAMSTAR system, and two time-area properties of EEG, to be specific, relative spike adequacy and spike rhythmicity have been utilized as contributions with the end goal of the discovery of epilepsy. The method proposed by Kiymik et al. utilizes a back propagation neural system with periodogram and autoregressive highlights as the contribution to the robotized discovery of epilepsy.


Despite the fact that the utilization of Artificial Neural Networks increases the computational complexity, the high in general recognition exactnesses accomplished with this system outperforms its drawback as in any computerized seizure location system; the identification of the seizure with high precision is of primary importance. Approximate Entropy demonstrates clear segregation between the typical and epileptic EEG signals. The optimum Approximate Entropy got in light of this information may not hold good for a general case. Subsequently, utilizing a straight separator with known Approximate Entropy parameter esteems may not give good results in circumstances where countless subjects are included. This issue won’t emerge in the proposed ANN-based strategy as it has performed well regardless of the Approximate Entropy utilized. It is realized that Approximate Entropy has great qualities, for example, vigor in the portrayal of the epileptic examples and low computational weight. Henceforth, a mechanized system utilizing Approximate Entropy as the info highlight is most appropriate for the ongoing location of the epileptic seizures.

The proposed system depends on two kinds of EEG, to be specific, EEG signals of alert and epileptic subjects. It can be made more strong by acclimatizing it to alternate indications of EEG like rest EEG.


• Pre processing module

• Approximate Entropy

• Training Signals

• Classification module

• Output Module.


• ENVIRONMENT: Visual Studio .NET 2008

• .NET FRAMEWORK: Version 3.5



o System: Pentium IV 2.4 GHz.

o Hard Disk: 40 GB.

o Floppy Drive: 1.44 Mb.

o Monitor: 15 VGA Color.

o Mouse: Logitech.

o Ram : 512 Mb.

Leave a Reply

Your email address will not be published. Required fields are marked *