Towards Personalized Neural Networks for Epileptic Seizure Prediction


- To develop a software tool that automatically classifies EEG (Electroencephalogram) data in one of four classes: inter-ictal (normal EEG pattern); pre-ictal (two minutes prior to the seizure onset); ictal (the seizure onset); pos-ictal (two minutes subsequent to seizure end). A typical epileptic seizure event is characterized by a cycle of those four stages.


- The data used in this investigation has been collected from the epilepsy database of Freiburg Center for Data Analysis and Modeling (FDM) of Albert Ludwigs University of Freiburg - EPILEPSIAE: Evolving Platform for Improving Living Expectations of Patients Suffering from IctAl Events, an FP7 ICT eHealth project.

- During an extensive experimentation, different neural network architectures have been applied to the datasets and compared. The best architectures of the neural networks have been selected and integrated in a MATLAB tool. The tool allows the users to select the EEG dataset, the neural network architecture and test the epileptic seizure events recognition performance.

- The source code and documentation of the application is available here.

- Software tool developed in collaboration with Joao Duarte during the Adaptive Computing and Diffuse Systems course.

- This work has been used in the publication - Antonio Dourado, Ricardo Martins, Joao Duarte, Bruno Direito - Towards Personalized Neural Networks for Epileptic Seizure Prediction, in Proc. of the ICANN 2008, pp. 479-487, Vol. 5164, Lecture Notes on Computer Science, International Conference on Neural Networks ICANN 2008, Prague, Czek Republic, September 2008

- This work has been used in the publication - Bruno Direito, Ricardo Martins, Rui Costa, Antonio Dourado, Francisco Sales, Marco Vieira - Computational Intelligence Algorithms for Seizure Prediction in Proc. of the 8th European Congress on Epileptology, 8th European Congress on Epileptology, Berlin, September 2008