Using the Spatio-Temporal Variogram for the Classification of Electroencephalographic (EEG) Assessment

Electroencephalography (EEG) is an important tool in neuroscience used to study the electrical brain responses following traumatic brain injuries (TBI), with the goal of clinical diagnosis and assessment. It is widely recognized that EEG produces inherently spatio-temporal data; however, the use of analytical methods that explicitly account for auto-correlation in both space and time have been limited. The lack of appropriate statistical methods, coupled with increased prevalence of TBI in both athletic and military settings, necessitates the development of sophisticated techniques for analysis of EEG data. We propose a novel method of EEG classification based on the spatio-temporal variogram. Using data from subjects with and without a history of TBI symptoms, we first computed spatio-temporal variograms for each EEG assessment. Second, we produced group-median variograms for both the healthy and the TBI groups. Third, we developed a two-parameter measure of dissimilarity between variogram surfaces and applied this measure to subject-specific and groupmedian variograms. Results indicate that our proposed measure out-performs several established measures of dissimilarity for the classification of EEG assessments.