Performance Analysis of Patient Specific Epilepsy Risk Level Classifications from EEG Signals Using Two Tier Hybrid (Fuzzy, Soft Decision Trees Models and MLP Neural Networks) Classifiers
R. Harikumar1, M. Balasubramani2, T. Vijayakumar3

1Dr. R. Harikumar, Professor, ECE Department,. Bannari amman Institute of Technology, Sathyamangalam, India.
2M. Balasubramani, Assistant Professor, ECE Department, .Info Institute of Engineering, Coimbatore, India.
3T. Vijayakumar, Assistant Professor,(Sr.Gr) Department of IT, Bannari Amman Institute of Technology, Sathyamangalam , India.

Manuscript received on April 11, 2012. | Revised Manuscript received on April 14, 2012. | Manuscript published on May 05, 2012. | PP: 541-549 | Volume-2 Issue-2, May 2012 . | Retrieval Number: B0649042212/2012©BEIESP
Open Access | Ethics and Policies | Cite 
© The Authors. Published By: Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC BY-NC-ND license (

Abstract: This paper compares the performance analysis of a two tier hybrid Fuzzy, Soft Decision Tree (SDT) models and Multi layer Perceptron (MLP) neural networks in optimization of patient specific epilepsy risk levels classifications from EEG (Electroencephalogram) signals. The fuzzy classifier (level one) is used to classify the risk levels of epilepsy based on extracted parameters like energy, variance, peaks, sharp and spike waves, duration, events and covariance from the EEG signals of the patient. Soft Decision Tree (post classifier with max-min and min-max criteria) of three models and MLP neural networks are applied on the classified data to identify the optimized risk level (singleton) which characterizes the patient’s state. The efficacies of these methods are compared with the bench mark parameters such as Performance Index (PI), Sensitivity, Specificity and Quality Value (QV). A group of twenty patients with known epilepsy findings are analyzed. High PI such as 95.88 % was obtained at QV’s of 22.43 in the SDT model of (16-4-2-1) with Method-II (min-max criteria) and for MLP (4-4-1) 99.9%and 24.43 when compared to the value of 40% and 6.25 through fuzzy classifier respectively. It was identified that the SDT models and MLP (4-4-1) are good post classifier in the optimization of epilepsy risk levels. SDT models were well accounted for low training cost over heads. A part from the training cost MLP neural networks outperformed SDT classifiers in classifying the epilepsy risk levels.

Keywords: EEG Signals, Epilepsy, Fuzzy Logic, Soft Decision Trees, Multi Layer Perceptron (MLP) neural networks, Risk Levels.