Adeli H., Ghosh-Dastidar S. Automated EEG-Based Diagnosis of Neurological Disorders: Inventing the Future of Neurology

Boca Raton: CRC Press, 2010. 423 p.Based on the authors’ groundbreaking research, Automated EEG-Based Diagnosis of Neurological Disorders: Inventing the Future of Neurology presents a research ideology, a novel multi-paradigm methodology, and advanced computational models for the automated EEG-based diagnosis of neurological disorders. It is based on the ingenious integration of three different computing technologies and problem-solving paradigms: neural networks, wavelets, and chaos theory. The book also includes three introductory chapters that familiarize readers with these three distinct paradigms. After extensive research and the discovery of relevant mathematical markers, the authors present a methodology for epilepsy diagnosis and seizure detection that offers an exceptional accuracy rate of 96 percent. They examine technology that has the potential to impact and transform neurology practice in a significant way. They also include some preliminary results towards EEG-based diagnosis of Alzheimer’s disease. The methodology presented in the book is especially versatile and can be adapted and applied for the diagnosis of other brain disorders. The senior author is currently extending the new technology to diagnosis of ADHD and autism. A second contribution made by the book is its presentation and advancement of Spiking Neural Networks as the seminal foundation of a more realistic and plausible third generation neural network.Basic Concepts.Signal Digitization and Sampling Rate.
Time and Frequency Domain Analyses.
Short Time Fourier Transform (STFT).
Wavelet Transform.
Types of Wavelets.
Advantages of the Wavelet Transform.Attractors in Chaotic Systems.
Measures of Chaos.
Preliminary Chaos Analysis - Lagged Phase Space.
Final Chaos Analysis.
Data Classification.
Cluster Analysis.
k-Means Clustering.
Discriminant Analysis.
Principal Component Analysis.
Artificial Neural Networks.
Feed forward Neural Network and Error Backpropagation.
Radial Basis Function Neural Network.
Automated EEG-Based Diagnosis of Epilepsy.
Spatio-Temporal Activity in the Human Brain.
EEG: A Spatio-Temporal Data Mine.
Data Mining Techniques.
Feature Space Identification and Feature EnhancementUsing Wavelet-Chaos Methodology.
Development of Accurate and Robust Classifiers.
Epilepsy and Epileptic Seizures.Wavelet Analysis of a Normal EEG.
Daubechies Wavelets.
Harmonic Wavelet.
Characterization.
Concluding Remarks.Wavelet-Chaos Analysis of EEG Signals.
Description of the EEG Data Used in the Research.
Data Preprocessing and Wavelet Decomposition of EEG into Sub-Bands.
Results of Chaos Analysis for a Sample Set of Unfiltered EEGs.
Statistical Analysis of Results for All EEGs.
Concluding Remarks.Wavelet-Chaos Analysis: EEG Sub-Bands and Feature Space Design.
Data Analysis.
k-Means Clustering.
Discriminant Analysis.
RBFNN.
LMBPNN.
Mixed-Band Analysis: Wavelet-Chaos-Neural Network.
Concluding Remarks.Principal Component Analysis for Feature Enhancement.
Cosine Radial Basis Function Neural Network: EEG Classification.
Output Encoding Scheme.
Comparison of Classifiers.
Sensitivity to Number of Eigenvectors.
Sensitivity to Training Size.
Sensitivity to Spread.
Concluding Remarks and Clinical Significance.
Automated EEG-Based Diagnosis of Alzheimer's Disease.Neurological Markers of Alzheimer's Disease.
Anatomical Imaging versus Functional Imaging.
Identification of Region of Interest (ROI).
Image Registration Techniques.
Linear and Area Measures.
Volumetric Measures.
Classification Models.
Approaches to Neural Modeling.
Hippocampal Models of Associative Memory.
Neural Models of Progression of AD.
EEGs for Diagnosis and Detection of Alzheimer's Disease.
Time-Frequency Analysis.
Wavelet Analysis.
Chace Analysis.
Concluding Remarks.Description of the EEG Data.
Chaos Analysis and ANOVA Design.
Complexity and Chaoticity of the EEG: Results of the Three-Way Factorial ANOVA.
Local Complexity and Chaoticity.
Discussion.
Chaoticity versus Complexity.
Eyes Open versus Eyes Closed.
Concluding Remarks.
Third Generation Neural Networks: Spiking Neural Networks.Information Encoding and Evolution of Spiking Neurons.
Mechanism of Spike Generation in Biological Neurons.
Models of Spiking Neurons.
Spiking Neural Networks (SNNs).
Unsupervised Learning.
Supervised Learning.
Feedforward Stage: Computation of Spike Times and Network Error.
Backpropagation Stage: Learning Algorithms.
Number of Neurons in Each Layer.
Initialization of Weights.
Heuristic Rules for SNN Learning Algorithms.
Input and Output Encoding.
SNN Architecture.
Type of Neuron (Excitatory or Inhibitory).
Convergence Results for a Simulation Time of ms.
Convergence Results for a Simulation Time of ms.
Input Encoding.
Output Encoding.
Convergence Criteria: MSE and Training Accuracy.
Convergence Criteria: MSE and Training Accuracy.
Heuristic Rules for Adaptive Simulation Time and SpikeProp Learning Rate.
Classification Accuracy and Computational Efficiency versus Training Size.Input and Output Encoding.
Classification Accuracy versus Training Size and Number of Input Neurons.
Classification Accuracy versus Desired Training Accuracy.Concluding Remarks.MuSpiNN Architecture.
Multi-Spiking Neuron and the Spike Response Model.
MuSpi NN Error Function.
Error Backpropagation for Adjusting Synaptic Weights.
Gradient Computation for Synapses Between a Neuron in the Last Hidden Layer and a Neuron in the Output Layer.
Gradient Computation for Synapses Between a Neuronin the Input or Hidden Layer and a Neuron in the Hidden Layer.
Parameter Selection and Weight Initialization.
Heuristic Rules for Multi-Spike Prop.
XOR Problem.
Fisher Iris Classification Problem.
EEG Classification Problem.
Discussion and Concluding Remarks.
The Future.