Features and Futures: Seizure Detection in Partial Epilepsies

Many factors underlying basic epileptic conditions determine the characteristics of epileptic seizures and the therapeutic outcome. Diagnosis and treatment rely on the clinical manifestations as well as electroencephalographic (EEG) epileptic activities. Any EEG abnormality will increase the recurrence risk and indicate drug therapy in cases that are unprovoked. Ictal symptoms, which correlate with seizure EEG activity, can guide neurosurgeons on epilepsy surgery. Modern neuroengineers use engineering technology to analyze EEG signals in investigating and treating epilepsy. For example, a pivotal trial that used responsive brain neurostimulation (RNS System, NeuroPace, Inc, Mountain View, CA) achieved positive results by significantly reducing the frequency of seizures for epileptic patients with medical intractability. From the pathophysiologic point of view, it is feasible to achieve automatic detection of epileptic neurophysiology using an implanted device, and to concomitantly reset brain dynamics from the ictal to the interictal state by electrical stimulation over specific brain regions. However, to improve and strengthen the effectiveness of current treatments, more accurate detection of seizure activity over time and delineation of epileptogenic regions are critical for on-demand therapy.

In the early 1970s, Viglione and Walsh preceded seizure prediction by using linear analysis approaches to extract seizure precursors from surface EEG recordings of absence seizures. Thereafter, many algorithms were developed to quantify EEG or intracranial EEG for seizure prediction or detection. With the advent of the physical-mathematical theory, the fusion of sophisticated techniques based on nonlinear dynamic systems is superior to those based on linear analyses in dealing with the changes in complexity or energy preictally. However, the performance validation of nonlinear systems is still not statistically significant, and the present-day systems are still not practical enough to be embedded in the implantable devices that warn of impending seizures or trigger therapeutic impulse responsively. While many investigators focus on improving mathematics for seizure prediction rather than on realizing the pathophysiology of epileptogenesis, the performance of the current seizure-detection algorithms is not satisfactory. There is no doubt that improvement in performance can be obtained once the algorithm more closely refers to the underlying mechanisms.

This article briefly reviews the fundamentals of the EEG, interictal, and ictal electrical activities of both extracranial and intracranial EEG of partial epilepsies, based on the information obtained from epilepsy patients who have undergone epilepsy surgery. The authors also present the status of their current research, focusing on decomposed seizure sources and the rendered spatial-temporal transitions in focal seizure.

Generation of brainwaves

The neuronal networks that generate brainwaves and subcorticocortical loops are modeled to explain the rhythms of superficial field potentials. The near-surface apical dendrites of pyramidal neurons are primarily responsible for summating the afferent impulses into postsynaptic potentials and forming effective dipoles. The recording of the field potentials shows sinusoidal fluctuations when there is a periodic sequence of the afferent bursts. Once populations of neurons are synchronously activated as in the definition of epileptic activity, the activity can be recorded as a convolution of the unit dipoles. This convolved activity may also be represented as a dipole or sheet of dipoles along the cortex. The theory of brainwave formation thus supports the application of independent component analysis, which is discussed in the later section of this article.

Features of epileptic brainwaves

Since Frederic Gibbs started to use EEG to assess epilepsy, EEG has revolutionized the field of epileptology. Many EEG abnormalities including spikes, sharp waves, spike-and-slow-wave complexes, polyspikes, and hypsarrhythmia are interpreted as epileptic activities. In the widely used classification of epileptic seizures proposed by the Commission on Classification and Terminology (1981), EEG features have become an integral part. By definition, partial seizures or generalized seizures are those in which the first clinical changes indicate that the initial involvement starts at the unilateral or bilateral hemisphere. Clinicians usually cannot determine the types of epileptic seizures or syndrome when the clinical data are inadequate or incomplete, therefore the initial EEG might support the differentiation.

Electroencephalographic Features of Generalized Seizures

The generalized seizure types include absence, myoclonic, clonic, tonic, tonic-clonic, and atonic seizure. Regular and symmetric spike-and-slow-wave complexes at 3 Hz are easily recognized as the ictal EEG of typical absence seizures. The polyspike-and-slow-wave, spike-and-slow-wave, or sharp-and-slow-wave complexes concurrently appearing at brief shocklike contraction confined to the face, trunk, or extremities, support the myoclonic seizure diagnosis. A sudden lapse of muscle tone occurrence leading to a head drop or a slump, polyspikes-and-slow-wave, or flattening or low-voltage fast activity can also be found simultaneously. Usually known as grand mal, tonic-clonic seizures are the most frequently encountered of the generalized seizure type. The grand mal attack is initiated by an abrupt loss of voltage of a few seconds’ duration; there is evidence of very fast (20–40/s) activity in all leads. After the first transition phase of seizure onset, rhythmic activity at about 10 per second with rapidly increasing amplitude dominates the EEG. Subsequently the initial fast activity enters the theta range, followed by repetitive bursts of spikes during the clonic phase. All of the epileptiform discharges in different generalized seizures can be displayed on the interictal scalp EEG independently.

Electroencephalographic Features of Partial Seizures

As clinical features, the variety of the ictal EEG of partial seizures is remarkable. From the clinical perspective, partial seizure attacks may be undetermined because either the presence of change in mental status and inappropriate behaviors from focal epilepticus of frontal lobe origin, or the presence of visceral ictal symptoms from temporal lobe origin, is not easily noticed. Hence, epileptologists may not be able to identify the ictal EEGs if the subtle onsets on the EEG display are not carefully examined when reading the records from patients with partial epilepsies. From the EEG perspective, in the case of temporal lobe epilepsy the interictal scalp EEG may show no abnormality, slight asymmetry of the background activity, or even bilateral temporal spikes.

The initial ictal EEG change of temporal lobe seizure may consist of clear spikes, but often only theta activity is seen. In the early work of Gibbs and colleagues and Gibbs and Gibbs, the ictal EEG activity of the psychomotor seizure was described as a special type of seizure discharge characterized by bursts of serrated slow waves, 4 flat-topped waves per second, and 6 high-voltage rhythmic activity per second ( Fig. 1 ). Many investigators including Gastaut, Vigouroux, and Klass emphasized the variability of the ictal EEG.