Brain-Computer Interface (BCI) Applications in Mapping of Epileptic Brain Networks Based on Intracranial-EEG: An Update

Abstract

The main applications of the Brain-Computer Interface (BCI) have been in the domain of rehabilitation, control of prosthetics, and in neuro-feedback. Only a few clinical applications presently exist for the management of drug-resistant epilepsy. Epilepsy surgery can be a life-changing procedure in the subset of millions of patients who are medically intractable. Recording of seizures and localization of the Seizure Onset Zone (SOZ) in the subgroup of "surgical" patients, who require intracranial-EEG (icEEG) evaluations, remain to date the best available surrogate marker of the epileptogenic tissue. icEEG presents certain risks and challenges making it a frontier that will benefit from optimization. Despite the presentation of several novel biomarkers for the localization of epileptic brain regions (HFOs-spikes vs. Spikes for instance), integration of most in practices is not at the prime time as it requires a degree of knowledge about signal and computation. The clinical care remains inspired by the original practices of recording the seizures and expert visual analysis of rhythms at onset. It is becoming increasingly evident, however, that there is more to infer from the large amount of EEG data sampled at rates in the order of less than 1 ms and collected over several days of invasive EEG recordings than commonly done in practice. This opens the door for interesting areas at the intersection of neuroscience, computation, engineering and clinical care. Brain-Computer interface (BCI) has the potential of enabling the processing of a large amount of data in a short period of time and providing insights that are not possible otherwise by human expert readers. Our practices suggest that implementation of BCI and Real-Time processing of EEG data is possible and suitable for most standard clinical applications, in fact, often the performance is comparable to a highly qualified human readers with the advantage of producing the results in real-time reliably and tirelessly. This is of utmost importance in specific environments such as in the operating room (OR) among other applications. In this review, we will present the readers with potential targets for BCI in caring for patients with surgical epilepsy.

Keywords: BCI; coherence analysis; connectivity index; epilepsy surgery; epileptogenicity index; high frequency brain stimulation; high frequency oscillations; single pulse electrical stimulation.

FIGURE 1

Different type of electrodes currently employed in practices. The depth electrodes and stereo-EEG commonly employed in Europe especially in France, and more recently in the United States. Whereas subdural electrodes constituted the mainstay of evaluations in the United States until the last few years.

FIGURE 2

A schematic showing a universal design for BCI systems. Adapted from BCI2000 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 51, NO. 6, JUNE 2004.

FIGURE 3

Function: (Left) hand motor according to commonly employed parametric methods of analysis of task-related gamma activation. (Right) Improving on the results by custom made algorithm. The results of direct electrical cortical stimulation are highlighted in cyan both figures for references. All shown in the electrode space. The size of the dot is proportional with the strength of task-related activation.

FIGURE 4

Identification of hand-motor area and the central sulcus (blue-line in electrode plain) and co-registration of the results with brain MRI in real-time based on free-running ECOG and custom-made software (Alkawadri et al., 2015). (A) Size of circles corresponds to the value M specific to the anterior lip of the central sulcus – whereas the color represents local field normalized power in the electrode space. (B) Co-registration of M values in the MRI space and proper thresholding to demonstrate the localization of the hand area.

FIGURE 5

Representative case: 34-year old woman with left neocortical temporo-parietal epilepsy. The size of the blue dots represents the “spikiness” of the automatic-detected spikes which are located in the left inferior and lateral temporal lobes, as well as in the left inferior parietal lobe, and overlap partly with, and within the vicinity of the seizure onset zone (red circles).

FIGURE 6

Demonstration of spatial distribution of physiologic high frequency oscillations. A reliable classifier to distinguish those from epileptic ones is desired (Alkawadri et al., 2014).

FIGURE 7

To-date no single EEG feature can reliably distinguish epileptic from non-epileptic HFOs.

FIGURE 8

Examples demonstrating a false detection in red (not discerned on raw data and does not occupy a blob in the Morley-Wavelet based spectral window), and true ripple marked by a green line. The spectral analysis is even more important in analysis of fast ripples (lower).

FIGURE 9

Abnormal (left, A,B) and normal evoked responses (right, E,F) to single pulse electrical stimulation. Panels (C,D) represents source localization of late and slow responses after stimulation of epileptic and non-epileptic orbitofrontal brain regions in two patients, respectively (Alkawadri et al., 2015).

FIGURE 10

This figure demonstrates the strong correlation between the duration of epilepsy and the degree of epileptogenicity from non-SOZ tissue as graded by the connectivity index.

FIGURE 11

Quantitative analysis based on cumulative ictal high frequency oscillations in two difficult-to-localize seizures. The size represents the length of electrode’s involvement in ictal HFOs, and the color represents normalized cumulative power up to 20 s after seizure onset. Note the strong spatial overlap and that these seizures were interpreted differently by the clinical team as temporal (above), and fronto-parietal (below).