Epileptic Seizures Can be Anticipated by Non-Linear Analysis
abstract & commentary
Source: Martinerie J, et al. Epileptic seizures can be anticipated by non-linear analysis. Nature Medicine 1998;4: 1173-1176.
No present electrophysiologic technique is able to detect or abort a seizure after it begins. Even intracranial electrodes provide only a few seconds warning before clinical onset.
Martinerie and colleagues studied 19 seizures in 11 patients using depth electrodes through the mesial temporal lobe to record and localize attacks. Using non-linear analysis, Martinerie et al explored the phase space, a concept used in non-linear dynamics, comprised by all physiological and pathological states of the recorded region of brain. The phase space describing a two electrode system, for example, consists of an X-Y graph in which each axis represents the voltage recorded at one or the other electrode. The "state" of the system (i.e., the set of voltages recorded at each of the two electrodes) is represented as a point on the X-Y graph. Dynamic electrode voltages are represented as points in phase space lying in a path or line that represents the time evolution of the system. Additional information can be obtained by including past electrode voltages in addition to the current ones.
Martinerie et al examined the phase space comprised by the current and preceding voltages in four electrodes closest to the site of seizure onset. The voltages recorded at these electrodes traced a trajectory through phase space corresponding to the present and past behavior of the neurons at the seizure focus. If the focal seizure neurons remained in the same physiological state, the corresponding points in phase space lay close together and yielded a high correlation density, a term defined as the distance between points along a path through phase space. A high correlation density reflected the tendency of the system to remain in the same region of phase space, whereas a low correlation density identified neurons at the seizure focus evolving from one state to another. As neuronal activity in the seizure focus shifted from physiological to pre-ictal activity, the distance between increased and produced a fall in the correlation density. Non-linear analysis detected the abrupt decline in correlation density and so detected the transition to a seizure state.
In 17 of 19 seizures, the correlation density abruptly declined between two and six minutes before seizure onset. One of the two causes may explain the failure of the algorithm in two seizures. First, interictal signals may have contained subthreshold seizure activity. If interictal activity closely resembled seizure activity, the transition would not cause a fall in the correlation density; or, second, the transition from the normal to the seizure state may have occurred too slowly to produce a fall in the correlation density.
Commentary
Martinerie et al provide a novel approach to the identification of seizure onset. Rather than simply recording the epileptic discharge, it identifies the instant that the state of neurons at the seizure focus begin to evolve.
The clinical use of this method will depend on at least two additional factors, namely the specificity of the method and the extension of the technique to extracranial EEG signals. The specificity will determine how the method reacts to nonpathological, but relatively abrupt state changes in the temporal lobe, such as with the rapid onset of sleep. Extending the technique to provide seizure detection using extracranial electrodes would provide early seizure detection while minimizing risk to the patient. Presently, the method requires the development of brain stimulation devices implanted in the mesial temporal lobe or other intracranial sites. Hopefully, noninvasive treatment modalities can be developed in the future. —fred a. lado & solomon l. moshe (Dr. Lado is EEG Fellow, Department of Neurology, Montefiore Medical Center—Albert Einstein College of Medicine, Bronx, New York. Dr. Moshe is Professor and Director, Pediatric Neurology and Clinical Neurophysiology, Department of Neurology, Montefiore Medical Center—Albert Einstein College of Medicine, Bronx, New York.)
You have reached your article limit for the month. Subscribe now to access this article plus other member-only content.
- Award-winning Medical Content
- Latest Advances & Development in Medicine
- Unbiased Content