A Case Study on EEG Signal Correlation Towards Potential Epileptic Foci Triangulation

Abstract

The precise localization of epileptic foci with the help of EEG or iEEG signals is still a clinical challenge with current methodology, especially if the foci are not close to individual electrodes. On the research side, dipole reconstruction for focus localization is a topic of recent and current developments. Relatively low numbers of recording electrodes cause ill-posed and ill-conditioned problems in the inversion of lead-field matrices to calculate the focus location. Estimations instead of tissue conductivity measurements further deteriorate the precision of location tasks. In addition, time-resolved phase shifts are used to describe connectivity. We hypothesize that correlations over runtime approaches might be feasible to predict seizure foci with adequate precision. In a case study on EEG correlation in a healthy subject, we found repetitive periods of alternating high correlation in the short (20 ms) and long (300 ms) range. During these periods, a numerical determination of proportions of predominant latency and, newly established here, directionality is possible, which supports the identification of loops that, according to current opinion, manifest themselves in epileptic seizures. In the future, this latency and directionality analysis could support focus localization via dipole reconstruction using new triangulation calculations.

Keywords: ECoG; EEG; clinical electric source imaging; signal propagation; time delay correlation.

Figure 1

Surface (x,y-) triangulation scheme of an ictal source in the depth (z) of about 50 mm with electrodes of 10 mm pitch (spacing center to center). Neural propagation leads to different transit times of 0.7 ms, 0.8 ms, and 1.0 ms, respectively.

Figure 2

Correlation coefficient r of channels Oz and PO3 within a 500 msec correlation interval. The x-axis denotes Oz time over 500 ms, against which PO3 was delayed by 10 ms up to 500 ms. For the shorter delays, a positive correlation is always found, which tends to zero around 100 ms and turns into partial anticorrelation for delays <250 ms.

Figure 3

Correlation difference of the pairing FC4-FT8 versus FT8-FC4.

Figure 4

Directionalities of the EEG correlations depicted for the visual and auditory cortices. The percentage values denote the maximum differences [%] together with the latencies [ms] of those maxima. The informational flux is yielded in a correct way. So do the latencies, which are short for the more primary areas and become prolonged for the more associative spots. The auditory system shows prolonged maximum directionality latencies when compared to the visual system.

Figure 5

Correlation analysis over several spots and longer times reveals phases of strong short correlations, which alternate with periods where the longer delays gain strength.