Pixel1.gif (51 bytes)
Pixel1.gif (51 bytes)
Pixel1.gif (51 bytes) Main Page Pixel1.gif (51 bytes)
About DSP Laboratory
Pixel.gif (52 bytes)
Contact Us
Pixel.gif (52 bytes)
Go to FIU's Homepage


 Pixel1.gif (51 bytes)


Curve.gif (104 bytes) Pixel1.gif (51 bytes)

Multi-Resolution Characterization of Interictal Epileptic Spikes based on Wavelet Transformation

Pixel1.gif (51 bytes)

"Multi-Resolution Characterization of Interictal Epileptic Spikes based on Wavelet Transformation", (1994)
Barreto A., Chin N., Andrian J., Riley J.

ABSTRACT: Interictal "spikes" are important features of the Electroencephalogram (EEG) or the Electrocorticogram (ECoG) of epileptic patients, which complement the information that the clinician obtains from the electrographic records during seizures. Unfortunately, attempts to define the fundamental characteristics of these abnormal electrographic transients have achieved only partial success. While the "sharp" character of the transients has played a central role in their detection, both by visual inspection of the EEG chart and by automated methods, it has also been found that other temporal characteristics, such as steepness (first derivative) of the slope preceding the peak, steepness of the slope following the peak, duration and amplitude may also be important factors in discriminating spikes from background EEG activity.

Numerous attempts have been made at the design of automated systems that continuously monitor these characteristics in an EEG channel and implement a decision-making process to indicate the occurrence of spikes, whenever these parameters are found to fall in the intervals expected for an interictal event.

Recently, it has been proposed that the most characteristic feature of interictal events in time may be their temporal sharpness, if considered not as a "point value" (i.e., an instantaneous measurement of second time derivative), but as a property measurable at different observation "spans". While the interictal "spikes" share their pointy apices with other non-epileptiform EEG features, there will be less normal waves that also maintain steep slopes before and after the apex. In other words, true interictal spikes will display sharpness in both the narrow spans of observation and the wider spans of observation, while a number of other waves commonly mistaken for interictal "spikes" such as muscle artifacts, will not be sharp in a wider span of observation.

The previous reasoning suggests that a multi-resolution assessment of sharpness in the EEG may be an efficient way to detect the occurrence of interictal "spikes". The plausibility of this approach has been verified through the implementation of a numerical second time derivative function at different decimation levels (i.e., different temporal resolutions). The encouraging results obtained in that previous study have led us to propose the decomposition of the EEG signal according to wavelets at different dilations to reflect the level of sharpness in the signal under analysis, at different temporal resolutions.

We have selected a wavelet suitable for the proposed decomposition by convolving a number of prospective wavelets with known "spikes". Then, a number of EEG segments have been decomposed according to the chosen wavelet, at different dilations. Plotting of the several component magnitudes through time, arranged in a three-dimensional display allows the identification of characteristic patterns in the decomposition of true interictal events. We propose that this may be an new way to define the temporal evolution of interictal spikes.

We will also discuss the representation of these components as the coordinates in an n-th dimensional space and the visualization of the EEG signal that is obtained for this representation with 3 chosen components, as well as the patterns associated with the occurrence of interictal "spikes".