Segmentation Methods Overview 

Get Complete Project Material File(s) Now! »

Chapter 2 Current Clinical Practice for EEG Interpretation

EEG interpretation is a skill that requires years of training. Typically, clinicians learn by experience, gradually coming to
recognise the tell-tale signs of maturation and abnormality in EEG signals. There are also general guidelines on the different aspects of EEG signals that should be looked for, and how these aspects are impacted by maturation and brain injury. This chapter presents some of the EEG qualities that clinicians look for, summarises the way human experts approach the task of EEG interpretation, and explores how the qualitative approach can be translated into a quantitative approach which can be automated.

Neonatal EEG and Maturation

The behaviour of EEG signals for a premature infant changes very rapidly as the infant grows closer to term. This is due to the speed of development of neurons during these critical weeks. There are a few certain aspects of the EEG that clinicians use to judge how well the EEG reflects the conceptional age (CA) of an infant. Dysmature EEG traces, referring to EEG traces that appear to be recorded from a younger infant, may indicate neurological problems.

Continuity

Continuity, used in a clinical sense when describing EEG signal, refers to the variation of the EEG signal amplitude. A section of EEG signal where the envelope stays relatively constant is described as “continuous” signal. A signal consisting of periods
of high amplitude signal as well as low voltage activities is referred to as “discontinuous”. Continuity is usually described qualitatively. Figure 2.1 shows segments of EEG signal from the common categories of neonatal EEG continuity.Continuous normal voltage signal refers to continuous signal where the voltage remains within the normal range and maintains a relatively constant amplitude. This pattern is the normal behaviour for near term infants. Trac´e alternant, also referred to as discontinuous, is used to describe signal where regions of high and low amplitude can be easily identified. This pattern is normal for younger preterm infants. For very premature infants, EEG traces predominantly consist of the trac´e discontinu pattern,appearing as alternating periods of high amplitude bursts and very low amplitude inactivity. In older infants, this pattern is referred to as “burst suppression” and is considered a sign of abnormality. Continuous low voltage refers to EEG with a relatively constant amplitude with consistent abnormally low voltage; while flat lining means that the EEG signal is almost non-existent. As the infant grows closer to term, the amplitude of the bursts in the trac´e discontinu pattern will decrease, and the period of the low amplitude inactivity will shorten, until around 34 weeks CA, when the EEG of an infant during wakeful periods becomes relatively continuous. Figure 2.2 shows an example of progression of EEG continuity from 26 to 30 weeks CA. Because continuity is a subjective qualitative measurement, for this thesis, the different continuity labels need to be defined. Continuous EEG refers to EEG signal which shows relatively little variation in its amplitude, while discontinuous signal
refers to signal where the variation in amplitude is noticible. Burst suppression, in this thesis, refers to both the abnormal pattern that appears in term infants and the normal pattern that appears in preterm infants referred to as trac´e discontinu. Burst suppression can be seen as an extreme form of discontinuity, and the distinction the infant with its conceptional age. Tracings from infants that do not match the continuity criteria for their age are considered dysmature.

Sleep-Wake Cycle

Related to continuity is the sleep wake cycle. Starting from around 31 weeks CA,infants should display different patterns during sleep, that can be used to differentiate between the sleeping and wakeful periods. At around 33 weeks CA, rapid eye movement (REM) sleep and non-REM sleep should also be identifiable [11]. The absence of this sleep-wake cycle is considered an abnormality when the infant is over 33 weeks old. REM or active sleep refers to sleep periods where rapid eye movement can
be observed. EEG recordings in this period are generally continuous. Non-REM or quiet sleep refers to sleep periods not involving rapid eye movement, and is generally discontinuous in nature. The discontinuous EEG pattern displayed during non-REM
sleep is also referred to as trac´e alternant.A summary of the different behaviours of the sleep EEG patterns, from 24 weeks
to term (40 weeks), is shown in table 2.1 (adapted from [11]). The length of time an infant spends in the sleep or wakeful state is also of importance. However, this is often interrupted by caring procedures (e.g. feeding or changing) or medication (e.g. sedation given for seizures).

READ  Chicken Digestive and Small Intestine Micro Anatomy

Abstract 
Acknowledgements 
Table of Contents 
List of Figures 
List of Tables 
List of Acronyms 
Glossary
1 Introduction
1.1 Objectives 
1.2 What is Electroencephalography?
1.3 Why Study Premature EEG?
1.3.1 Statistics of Preterm Neonatal Clinical Outcomes
1.3.2 Brain Monitoring and EEG
1.3.3 Bedside Monitoring
1.4 Contribution Summary 
1.5 Research Scope 
1.6 Overview of Proposed System
1.7 Thesis Structure 
2 Current Clinical Practice for EEG Interpretation 
2.1 Neonatal EEG and Maturation 
2.1.1 Continuity
2.1.2 Sleep-Wake Cycle
2.1.3 Synchrony
2.1.4 Other Landmark Patterns
2.2 Signs of Abnormal EEG 
2.2.1 Symmetry
2.2.2 Burst Suppression
2.2.3 Seizure
2.2.4 Dysmature EEG
2.3 Translating Clinical Knowledge into Mathematics 
2.3.1 Normal Range
2.3.2 Defining Continuity
2.3.3 Defining Synchrony
2.3.4 Defining Symmetry
2.3.5 EEG Pattern Recognition
2.3.6 Artifacts
2.3.7 The Big Picture Approach
2.4 Discussion 
2.5 Summary
3 Literature Review 
3.1 Medical EEG Studies
3.1.1 Signs of Maturation
3.1.2 Signs of Abnormal Development
3.2 Engineering Projects
3.2.1 Seizure Detection
3.2.2 Hypoxia Detection
3.2.3 EEG Summary by Segment Clustering
3.2.4 Background Continuity State Detection
3.3 Discussion 
3.3.1 The Gap between the Fields
3.3.2 Continuity as a Measure of Maturation
3.3.3 The Importance of Context
3.3.4 Using Engineering Methods to Help Medical Research
3.4 Summary 
4 Time Frequency Analysis of EEG Signals
4.1 Time Frequency Analysis 
4.1.1 Overview of Cohen’s Class Distributions
4.2 Kernel Decision 
4.2.1 Choi-Williams Distribution
4.2.2 Modified B-Distribution
4.2.3 Kernel Parameters and Smoothing Window Decisions
4.2.4 Comparison Between the Two Kernels
4.3 Time-Frequency Distributions of Different EEG Continuity Background States 
4.3.1 Normal Continuous Signals
4.3.2 Discontinuous Signals
4.3.3 Burst Suppression
4.3.4 Continuous Low Voltage
4.3.5 Seizure
4.3.6 Summary of Observations
4.4 Discussion 
4.5 Summary 
5 EEG Segmentation 
5.1 Segmentation Methods Overview 
5.1.1 Spectral Error Measurement (SEM)
5.1.2 Nonlinear Energy Operator (NLEO)
5.1.3 Generalized Likelihood Ratio (GLR)
5.2 Evaluation of the Methods
5.2.1 Method
5.2.2 Discussion
5.3 Autoregressive Model Order Optimisation 
5.4 Summary
6 Quantifying Continuity 
6.1 The Need to Quantify Continuity
6.2 Current Methods of Continuity Measurement
6.3 The Continuity Feature
6.3.1 Amplitude Vector
6.3.2 Modelling Amplitude Distribution
6.3.3 Distribution Parameters as a Continuity Feature
6.4 Displaying Continuity
6.4.1 Line Plot of the Continuity Feature
6.4.2 Scatter Plot of the Continuity Feature
6.5 Quantifying Continuity Using Principal Component Analysis 
6.6 Summary
7 The Continuity Feature and Maturation 
7.1 Maturation and Continuity 
7.2 Changes in the Continuity Feature Throughout Maturation 
7.3 Effects of Brain Injury on Maturation 
7.4 Discussion 
7.5 Summary 
8 Background State Classification 
8.1 Overview of Continuity Classification 
8.2 Evaluation of Classification Systems
8.3 Linear Discriminant Analysis
8.3.1 Results and Discussion
8.4 Self Organising Map 
8.4.1 Results and Discussion
8.5 Gaussian Mixture Model
8.5.1 Results and Discussion
8.6 Comparison of Algorithm Performance
8.7 Summary 
9 Discussion 
9.1 Quantitative Continuity Feature Versus Existing Continuity Measurements
9.2 Quantified Continuity Measure 
9.3 Monitoring Display 
9.4 Qualitative Labels
9.5 Continuity as the Context for EEG Analysis 
9.6 Maturation Index 
9.7 Summary
10 Conclusions and Future Works 
10.1 Conclusions 
10.2 Future Works
10.2.1 Automatic Sleep Pattern Detection
10.2.2 Further Feature Analysis
10.2.3 Burst Suppression Analysis
10.2.4 Maturation Index
List of References
A Database of Preterm Infant EEG 
B Cerebral Function Monitoring and Amplitude-Integrated EEG 
B.1 Matlab Code for TFD Calculation: with Time-Lag Kernels

GET THE COMPLETE PROJECT
Quantitative Continuity Feature for Preterm Neonatal EEG Signal Analysis

Related Posts