EEG-Based Experiment Design for Major Depressive Disorder: Machine Learning and Psychiatric Diagnosis

Cover
EEG-Based Experiment Design for Major Depressive Disorder: Machine Learning and
Psychiatric Diagnosis
Copyright
Dedication
About the Authors
Preface
Acknowledgments
1 Introduction: Depression and Challenges
	1.1 Introduction
	1.2 Depression and Subtypes
	1.3 Signs and Symptoms of Depression
	1.4 Unipolar Depression and Challenges
	1.5 Electroencephalography as a Clinical Modality
	1.6 Electroencephalography-Based Machine Learning Methods for Depression
		1.6.1 Data Preprocessing
		1.6.2 Feature Extraction
		1.6.3 Feature Selection
		1.6.4 Classification
		1.6.5 The 10-Fold Cross Validation
	1.7 Electroencephalography-Based Diagnosis for Depression
	1.8 Electroencephalography-Based Treatment Selection for Depression
	1.9 Discussion
	1.10 Summary
	References
2 Electroencephalography Fundamentals
	2.1 What Is Electroencephalography?
	2.2 Electroencephalography Applications
	2.3 Electroencephalography Frequency Bands
		2.3.1 Delta Band
		2.3.2 Theta Band
		2.3.3 Alpha Band
		2.3.4 Low Beta Band
		2.3.5 Beta Band
		2.3.6 High Beta Band
		2.3.7 Gamma Band
	2.4 Electroencephalography Recording Techniques
		2.4.1 Electroencephalography Sensor’s Location
		2.4.2 Electroencephalography Sensors and Conductive Media
		2.4.3 Electroencephalography Amplification
	2.5 Electroencephalography Reference Choices
		2.5.1 Linked-Ear Reference
		2.5.2 Average Reference
		2.5.3 Reference Electrode Standardization Technique
	2.6 Electroencephalography Artifacts
		2.6.1 Ocular Artifacts
		2.6.2 Muscle Artifacts
		2.6.3 Line Noise Artifacts
		2.6.4 Gait-Related Motion Artifacts
	2.7 Electroencephalography-Based Method for Artifact Reduction (Electroencephalography Preprocessing)
		2.7.1 Analog Methods
		2.7.2 Linear Filtering Methods
		2.7.3 Regression-Based Methods
		2.7.4 Dipole Modeling-Based Methods
		2.7.5 Blind Source Separation
		2.7.6 Principal Component Analysis
		2.7.7 Independent Component Analysis
		2.7.8 Canonical Correlation Analysis
		2.7.9 Adaptive Noise Cancellation
		2.7.10 Wavelet-Based Artifact Reduction/Thresholding Methods
		2.7.11 Template Matching Methods (Dynamic Time Warping and Empirical Model Decomposition)
		2.7.12 Gait-Related Motion Artifacts
		2.7.13 Subspace-Based Methods
		2.7.14 Hybrid Methods for Electroencephalography-Based Artifact Reduction
		2.7.15 Single Channel-Based Separation
		2.7.16 Miscellaneous Methods
	2.8 Summary
	References
3 Electroencephalography-Based Brain Functional Connectivity and Clinical Implications
	3.1 Introduction
	3.2 Clinical Implications of Electroencephalography-Based Brain Connectivity Methods
		3.2.1 Alzheimer’s
			3.2.1.1 Interhemispheric Coherence
			3.2.1.2 Phase Lag Index
			3.2.1.3 Synchronization Likelihood
		3.2.2 Mild Cognitive Impairment
			3.2.2.1 Interhemispheric Coherence
			3.2.2.2 Synchronization Likelihood
		3.2.3 Major Depressive Disorder
			3.2.3.1 Interhemispheric Coherence
			3.2.3.2 Synchronization Likelihood
			3.2.3.3 Partial Directed Coherence
		3.2.4 Schizophrenia
			3.2.4.1 Interhemispheric Coherence and Imaginary Coherence
			3.2.4.2 Generalized Synchronization
			3.2.4.3 Correlation and Mutual Information
		3.2.5 Epilepsy
			3.2.5.1 Interhemispheric Coherence
			3.2.5.2 Synchronization Likelihood
			3.2.5.3 Correlation
		3.2.6 Alcoholism
			3.2.6.1 Interhemispheric Coherence
			3.2.6.2 Synchronization Likelihood
	3.3 Open-Source Toolboxes
		3.3.1 Extended Multivariate Autoregressive Toolbox
		3.3.2 The Granger Causality Connectivity Analysis Toolbox
		3.3.3 The Multivariate Granger Causality Toolbox
		3.3.4 HERMES Toolbox
		3.3.5 The eConnectome Toolbox
		3.3.6 The BSMART Toolbox
	3.4 Summary
	References
4 Pathophysiology of Depression
	4.1 Introduction
	4.2 Brain Volume Abnormalities during Depression
		4.2.1 Frontal Cortex
		4.2.2 Hippocampus
		4.2.3 Amygdala
		4.2.4 Basal Ganglia
		4.2.5 Temporal Lobe
	4.3 Mechanisms Underlying the Pathophysiology of Depression
		4.3.1 Genetic and Nongenetic Factors
		4.3.2 Hypothalamic-Pituitary-Adrenal Axis Dysfunction
		4.3.3 Neurotrophic Factors of Depression
		4.3.4 Defects in Intracellular Signaling Pathways
		4.3.5 Altered Glutamatergic and GABAergic Neurotransmission
		4.3.6 Circadian Rhythms
	4.4 Electroencephalography Correlates for Depression
	4.5 Electroencephalography-Based Diagnosis of Depression
		4.5.1 Electroencephalography Frequency Bands
		4.5.2 Event-Related Potentials Component: P300
		4.5.3 Machine Learning Methods to Diagnose Depression
	4.6 Summary
	References
5 Using Electroencephalography for Diagnosing and Treating Depression
	5.1 Introduction
	5.2 Electroencephalography-Based Diagnosis for Depression
		5.2.1 Electroencephalography Frequency Bands
		5.2.2 Electroencephalography Alpha Interhemispheric Asymmetry
		5.2.3 Electroencephalography-Based Computer-Aided Diagnosis for Depression
			5.2.3.1 Year 2018
			5.2.3.2 Year 2017
			5.2.3.3 Year 2016
			5.2.3.4 Year 2015
			5.2.3.5 Brain Connectivity During Depression
	5.3 Electroencephalography-Based Antidepressant Treatment Selection
		5.3.1 Electroencephalography Frequency Bands
		5.3.2 Antidepressant Treatment Response Index
		5.3.3 QEEG Theta Cordance
		5.3.4 Referenced Electroencephalography
		5.3.5 Rostral Anterior Cingulate Cortex Activations
		5.3.6 Machine Learning Methods for Treatment Selection
	5.4 Event-Related Potential-Based Antidepressant Treatment Selection
		5.4.1 P200 and P300
		5.4.2 Loudness Dependence Auditory Evoked Potential
	5.5 Integrating Neurobiological and Electrophysiological Data
	5.6 Summary
	References
6 Neural Circuits and Electroencephalography-Based Neurobiology for Depression
	6.1 Introduction
	6.2 Neural Circuitry Implicated during Depression
		6.2.1 Clinical Phenomenology of Depression
		6.2.2 Neural Substrates of Mood Disorders
		6.2.3 Limbic Structures
		6.2.4 Prefrontal Cortex
		6.2.5 Cortical Projections to the Hypothalamus and Brainstem
		6.2.6 Cortico-Striatal-Thalamic Circuits Related to the Orbitomedial Prefrontal Cortex
	6.3 Neurobiology of Electroencephalography-Based Predictive Biomarker for Depression
		6.3.1 Changes in Alpha Band Activity
		6.3.2 Electroencephalography Theta Band Activity
		6.3.3 Alpha Interhemispheric Asymmetry
		6.3.4 Theta Cordance
		6.3.5 Antidepressant Treatment Response Index
		6.3.6 The P300
		6.3.7 LDEAP
	6.4 Summary
	References
7 Design of an Electroencephalography Experiment for Assessing Major Depressive Disorder
	7.1 Introduction
	7.2 Design of Study Protocol
		7.2.1 Sample Size Calculation
		7.2.2 Recruitment Criteria
		7.2.3 Clinical Questionnaires
		7.2.4 Experimental Setup for EEG/ERP Data Acquisition
	7.3 Study Participants Information
	7.4 Electroencephalography-Based Localization for Disease Pathology
		7.4.1 Topographic Maps of Activations
		7.4.2 sLORETA Analysis
	7.5 Low-Dimensional Representation
	7.6 EEG/ERP Differences Between Major Depressive Disorder Patients and Healthy Controls
		7.6.1 Event-Related Potential Component: P300
		7.6.2 Absolute Power and EEG Alpha Interhemispheric Asymmetry
	7.7 Summary
	References
8 Electroencephalography-Based Diagnosis of Depression
	8.1 Introduction
	8.2 Electroencephalography Preprocessing
	8.3 Feature Extraction
		8.3.1 ITMS Diagnosis
			8.3.1.1 Significance of the Features (ITMS Diagnosis)
			8.3.1.2 Computing Power for Different Electroencephalography Bands
			8.3.1.3 Computing Electroencephalography Alpha Interhemispheric Asymmetry
			8.3.1.4 Computing Functional Connectivity With Synchronization Likelihood
			8.3.1.5 Integration of Features
	8.4 Standardization
	8.5 Feature Selection
		8.5.1 Example 1
		8.5.2 Example 2
	8.6 Classification Models
		8.6.1 Logistic Regression Classification
		8.6.2 Support Vector Machine Classification
		8.6.3 Naïve Bayesian Classification
	8.7 Validation
	8.8 MDD Patients Versus Healthy Controls
		8.8.1 The sLORETA Analysis
		8.8.2 Electroencephalography Signal Power and Alpha Interhemispheric Asymmetry
		8.8.3 ERP Component: P300
		8.8.4 Classification Results (ITMS Diagnosis)
	8.9 Summary
	References
9 Electroencephalography-Based Treatment Efficacy Assessment Involving Depression
	9.1 Introduction
		9.1.1 ITMS-Treatment Selection
			9.1.1.1 Significance of Features (ITMS-Treatment Selection)
			9.1.1.2 Selection of an Appropriate Basis Function for Electroencephalography Analysis
			9.1.1.3 Computing Wavelet-Based Coefficients
			9.1.1.4 Computing Wavelet-Based Signal Energy
			9.1.1.5 Computing Wavelet-Based Sample Entropy
			9.1.1.6 Computing Wavelet-Based Composite Permutation Entropy Index
			9.1.1.7 Computing Wavelet-Based Fractal Dimension
			9.1.1.8 Integration of Features
		9.1.2 Finalizing the Electroencephalography Data Matrix
	9.2 Treatment Respondents Versus Nonrespondents
		9.2.1 sLORETA Analysis
		9.2.2 Topographic Maps
		9.2.3 Ranking Features Based on Receiver Operating Characteristic Criterion
		9.2.4 Low-Dimensional Representation
		9.2.5 Classification Results (ITMS-Treatment Selection)
	9.3 Discussion
	9.4 Summary
	References
Index
Back Cover