Localizing acute stroke-related EEG changes: assessing the effects of spatial undersampling

Because of its sensitivity to metabolic and ionic disturbances related to ischemia, the EEG can be a potentially useful tool for acute stroke detection and for monitoring affected tissue. However, the clinical use of the EEG in detecting stroke is determined in part by how accurately the spatial information is characterized. The purpose of the current study was to determine the effects of spatial undersampling on the distribution and interpretation of the stroke-related topographic EEG. Using a 128-channel sensor montage, EEG was recorded from six stroke patients acutely (between 8 and 36 hours) after symptom onset. The EEG was submitted to a spectral analysis and was compared with patient symptoms and MRI and computed tomographic findings. To determine loss of spatial and clinical information resulting from spatial undersampling, the average-referenced data from the original 128-channel recording montage were subsampled into 64-, 32-, and 19-channel arrays. Furthermore, the analytical findings were compared with a board-certified electroencephalographer's review of the raw EEG using a conventional clinical montage. As predicted, the results showed that accurate description of stroke-related topographic EEG changes is dependent on adequate spatial sampling density. Accurate description of the spatial distribution of the stroke-related EEG was achieved only with the 64- and 128-channel EEG. As the recording density decreases to 32 channels, the distribution of the scalp EEG spectra is distorted, potentially resulting in mislocalization of the affected region. Results of the clinical review by an expert electroencephalographer corroborated the quantitative analyses, and the results also demonstrated the shortcomings of the conventional 10-20 recording density for capturing focal EEG abnormalities in several cases. The EEG provides useful information about the localization of acute cerebral ischemia, but recording densities of 64 channels or higher are required for accurate spatial characterization of focal stroke-related EEG changes.