Effects of Coherent Noise on Ictal Component Selection for EEG Source Imaging
This paper examines the effects of coherent noise on the scalp voltage topography, activity power spectra and dipole residual variances of the independent components (ICs) of ictal EEG signals. Eleven different sets of ictal EEG signals are generated by a
- PDF / 2,408,574 Bytes
- 4 Pages / 595 x 791 pts Page_size
- 59 Downloads / 168 Views
Abstract— This paper examines the effects of coherent noise on the scalp voltage topography, activity power spectra and dipole residual variances of the independent components (ICs) of ictal EEG signals. Eleven different sets of ictal EEG signals are generated by adding various amounts of coherent noises. All of these simulated EEGs are decomposed into their corresponding ICs. Single dipole source that helps to distinguish ictal components from noise components, two-dimensional (2D) topographic map, activity power spectrum, and dipole residual variance were estimated for each of these ICs. Topographic maps show that the number of ictal components decreases with the increase of noise level. Activity power spectrum analysis supports the result of topographic map analysis. The average residual variances not only increase with the increment of noise level but sometimes decrease as well. Simultaneous consideration of these three features is helpful for better selection of ictal components. Keywords— Independent Component Analysis, Ictal EEG, Ictal Component, EEG Source Imaging, Noise.
I. INTRODUCTION Electroencephalography (EEG) is used, almost inevitably, for the pre-surgical evaluation of medically intractable focal epilepsy. Possible cortical sources of scalp EEG can be estimated and depicted with the help of a computational technique, namely EEG source imaging (ESI). Either ictal or interictal EEG events are analyzed for ESI based source estimation. Although ictal EEG is comparatively difficult to analyze with ESI because of the low Signal to Noise Ratio (SNR) [1], ictal EEG is believed to be more reliable than interictal EEG in localizing the epileptogenic focus [2]. Ictal EEG measures cortical seizure discharges superposed with various artifacts, external noises and other background brain oscillations. The unwanted parts of ictal EEG can be minimized by utilizing digital filters, various artifact removal/rejection algorithms including independent component analysis (ICA) and sometimes by averaging the selected ictal discharges. The ICA-based ictal ESI studies [35] decomposed each set of ictal EEG into a series of spatially fixed and temporally independent components and then analyzed only the selected ictal components for ESI.
The major challenge of ICA-based ictal ESI algorithms is to select the correct ictal ICs that correspond to the actual seizure discharge. There is no concrete rule to identify those ictal ICs. Selection techniques are mostly visual inspection dependent. Jung et al. [3] excluded the ICs of muscle artifacts, eye movements, and 60-Hz noise by visually inspecting the scalp voltage topography and activity power spectra. They localized a single dipole source for each IC using DIPFIT and excluded the ICs of dipole sources located outside the head model or with a residual variance of more than 20%. Yang et al. [4] and Lu et al. [5] used dynamic seizure imaging (DSI) approach for ictal ESI. They also removed the unwanted ICs (e.g. ICs related to eye movement and muscle artifacts) through the visual
Data Loading...
