EEG Signal Processing and Machine Learning. Saeid Sanei
morphology very similar to ictal patterns. In old age patients other similar patterns such as subclinical rhythmic EEG discharges of adults (SREDA) over the 4–7 Hz frequency band around the centroparietal region, and a wide frequency range (2–120 Hz) temporal minor sharp transient and wicket spikes over anterior temporal and midtemporal lobes of the brain may occur. These waves are also nonepileptic but with seizure‐type waveform [47].
The epileptic seizure patterns, called ictal wave patterns, appear during the onset of epilepsy. Although the next chapters of this book focus on analysis of these waveforms from signal processing and machine learning points of view, here a brief explanation of morphology of these waveforms is given. Researchers in signal processing may exploit these concepts in the development of their algorithms. Although these waveform patterns are often highly obscured by the muscle movements, they normally maintain certain key characteristics.
Tonic–clonic seizure (also called grand mal) is the most common type of epileptic seizure. It appears in all electrodes but more towards the frontal electrodes (Figure 2.15). It has a rhythmic but spiky pattern in the EEG and occurs within the frequency range of 6–12 Hz. Petit mal is another interictal paroxysmal seizure pattern which occurs at approximately 3 Hz with a generalized synchronous spike wave complex of prolonged bursts. A temporal lobe seizure (also called a psychomotor seizure or complex partial seizure) is presented by bursts of serrated slow waves with relatively high amplitude of above 60 μV and frequencies of 4–6 Hz. Cortical (focal) seizures have contralateral distribution with rising amplitude and diminishing frequency during the ictal period. The attack is usually initiated by local desynchronization, i.e. very fast and very low voltage spiky activity, which gradually rises in amplitude with diminishing frequency. Myoclonic seizures have concomitant polyspikes seen clearly in the EEG signals. They can have generalized or bilateral spatial distribution more dominant in the frontal region [50]. Tonic seizures occur in patients with Lennox–Gastaut syndrome [51] and have spikes which repeat with frequency approximately 10 Hz. Atonic seizures may appear in the form of a few seconds drop attack or be inhibitory, lasting for a few minutes. They show a few polyspike waves or spike waves with generalized spatial distribution of approximately 10 Hz followed by large slow waves of 1.5–2 Hz [52]. Akinetic seizures are rare and characterized by arrest of all motion, which, however, is not caused by sudden loss of tone as in atonic seizure and the patient is in an absent‐like state. They are rhythmic with frequency of 1–2 Hz. Jackknife seizures also called salaam attacks, are common in children with hypsarrhythmia (infantile spasms, West syndrome) are either in the form of sudden generalized flattening desynchronization or have rapid spike discharges [51].
Figure 2.14 Bursts of 3–7 Hz seizure activity in a set of adult EEG signals.
Figure 2.15 Generalized tonic–clonic (grand mal) seizure. The seizure appears in almost all the electrodes.
There are generally several varieties of recurring or quasi‐recurring discharges, which may or may not be related to epileptic seizure. These abnormalities may be due to psychogenic changes, variation in body metabolism, circulatory insufficiency (which appears often as acute cerebral ischaemia). Of these, the most important ones are: periodic or quasiperiodic discharges related to severe CNS diseases; periodic complexes in subacute sclerosing panencephalitis (SSPE); periodic complexes in herpes simplex encephalitis; syncopal attacks; breath holding attacks; hypoglycaemia and hyperventilation syndrome due to sudden changes in blood chemistry [53]; and periodic discharges in CJD (mad cow disease) [54, 55]. The waveforms for this latter abnormality consist of a sharp wave or a sharp triphasic transient signal of 100–300 ms duration, with a frequency of 0.5–2 Hz. The periodic activity usually shows a maximum over the anterior region except for the Heidenhain form, which has a posterior maximum [47]. Other epileptic waveforms include periodic literalized epileptiform discharges (PLED), periodic discharges in acute cerebral anoxia, and periodic discharges of other etiologies.
Despite the above epileptiform signals there are spikes and other paroxysmal discharges in healthy nonepileptic persons. These discharges may be found in healthy individuals without any other symptoms of diseases. However, they are often signs of certain cerebral dysfunctions that may or may not develop into an abnormality. They may appear during periods of particular mental challenge on individuals, such as soldiers in the war front line, pilots, and prisoners.
Generation of epileptiform brain discharges from deeper brain layers such as the hippocampus during pre‐ictal or interictal periods is an indication of upcoming seizure. These discharges which are spike‐type and have particular morphology which can be seen by inserting electrodes such as multichannel foramen ovale electrodes deep into the hippocampus. More than 90% of these discharges cannot be seen over the scalp due to their attenuation and smearing. A comprehensive overview of epileptic seizure disorders and nonepileptic attacks can be found in many books and publications such as [53, 56]. In this book a chapter is dedicated to the methods for analyzing intracranial and scalp EEGs.
2.9.3 Psychiatric Disorders
Not only can functional and certain anatomical brain abnormalities be investigated using EEG signals, pathophysiological brain disorders can also be studied by analyzing such signals. According to the ‘Diagnostic and Statistical Manual (DSM) of Mental Disorders’ of the American Psychiatric Association, changes in psychiatric education have evolved considerably since the 1970s. These changes have mainly resulted from physical and neurological laboratory studies based upon EEG signals [57].
There have been evidences from EEG coherence measures suggesting differential patterns of maturation between normal and learning disabled children [58]. This finding can lead to the establishment of some methodology in monitoring learning disorders.
Several psychiatric disorders are diagnosed by analysis of EPs achieved by simply averaging a number of consecutive trails having the same stimuli.
Some pervasive mental disorders such as: dyslexia which is a developmental reading disorder; autistic disorder which is related to abnormal social interaction, communication, and restricted interests and activities, and starts appearing from the age of three; Rett's disorder, characterized by the development of multiple deficits following a period of normal postnatal functioning; and Asperger's disorder which leads to severe and sustained impairments in social interaction and restricted repetitive patterns of behaviour, interests, and activities; cause significant losses in multiple functioning areas [57].
ADHD and attention‐deficit disorder (ADD), conduct disorder, oppositional defiant disorder, and disruptive behaviour disorder have also been under investigation and considered within the DSM. Most of these abnormalities appear during childhood and often prevent children from learning and socializing well. The associated EEG features have been rarely analytically investigated, but the EEG observations are often reported in the literature [59–63]. However, most of such abnormalities tend to disappear with advancing age.
EEG has also been analyzed recently for the study of delirium [64, 65], dementia [66, 67], and many other cognitive disorders [68]. In EEGs, characteristics of delirium include slowing or dropout of the posterior dominant rhythm, generalized theta or delta slow‐wave activity, poor organization of the background rhythm, and loss of reactivity of the EEG to eye opening and closing. In parallel with that, the quantitative EEG (QEEG) shows increased absolute and relative slow‐wave (theta and delta) power, reduced ratio of fast‐to‐slow band power, reduced mean frequency, and reduced occipital peak frequency [65].
Dementia includes a group