In the neuro ICU several methods of EEG recording can be applied to obtain EEG data from patients with acute brain injury, among them spot EEG, repeated short EEG recordings, continuous surface EEG, and depth EEG (dEEG). All of these modes of EEG recording can be enhanced by using quantitative EEG (see Chap. 20 for details). Surface EEG (sEEG), as one of the oldest and most intensely studied techniques of monitoring the brain, has its unquestioned diagnostic value in seizure detection, monitoring the depth of sedation, and in prognostication [68]. Open questions in the neuro ICU context include the choice of EEG technique (or combination of techniques) and duration of EEG recording that produces the highest diagnostic yield in a given disease entity and patient [69]. In recent years, technical advancements have made digital transformation of the analog EEG signal, storage, and—at least partly—the automated analysis of large amounts of EEG data possible. Nevertheless, continuous EEG (cEEG) recordings still require a substantial amount of manpower at significant cost, which is why research on effective use is warranted.

Both through modifications of the EEG recording technique and through development of data analysis methods researchers have tried to improve the applicability of sEEG in the neuro ICU [70].

A relatively new development in the neuro ICU is the use of quantitative EEG (qEEG) to aid in the timely extraction of diagnostic information from the large amounts of cEEG data. Various algorithms are available to analyze and visualize qEEG data. After undergoing Fast Fourier Transformation the raw EEG data can either be represented as total power, ratios of power (e.g., alpha delta ratio), or spectrograms. Spectrograms in turn can be graphed as numbers, compressed spectral arrays (CSA), staggered arrays, or histograms [71, 72]. The analogous interpretation of raw EEG data is too time consuming to be feasible by a neurointensive care physician alone whose responsibilities will only grow with the technical progress in the field of neurocritical care , leaving less time for EEG review. The alternative, employment of full time electrophysiologists, is a financial and logistic challenge that most hospitals are not able to afford. In the future, qEEG will hopefully offer efficient screening and alarm mechanisms indicating clinically concerning EEG trends, in order to make big EEG data accessible and useful at the bedside [73].

Another approach to master the task of extracting relevant information from large amounts of raw EEG recordings is the development of seizure detection software. For such software to detect seizures with reasonable sensitivity and specificity, seizures should ideally have a clear on- and offset and should be clearly distinguishable from the EEG background. While this is often the case in a relatively noise-free environment, such as epilepsy monitoring units, it is rarely the case in neurointensive care units that have many ambient sources of noise [74, 75]. Moreover, muscle artifacts and periodic discharges are more common in the NICU. The latter share many electrographic features with seizures and are therefore difficult to distinguish by automated seizure detection software [76, 77]. Indeed there is ongoing debate regarding the clinical relevance of periodic discharges that fall along the ictal-interictal spectrum lacking the discrete on–offset of events that are universally recognized as ictal. Seizure detection algorithms using intracortical EEG data have not been tested so far, but are promising given the overall high signal to noise ratio of dEEG.

Surface EEG has the advantage of providing a global electrophysiological signal covering most parts of the cerebral cortex. However, sEEG electrodes are separated from the source of the EEG signal originating from the brain tissue by several structures: skin, hair, the skull, as well as the meninges and any pathologic material (blood, pus, etc.) that may lie within. Distant from the source of the signal, sEEG is also close to sources of noise. Sources of noise include poor electrode contact and—particularly in the neuro ICU–surrounding electronic devices, which may cause artifacts distorting the EEG signal. Another major source of artifact is the electrical signal generated by muscle and movement. These disadvantages may be overcome by the local insertion of intracortical electrodes, which allows at least local—not global—EEG recordings with minimal artifact and a high signal to noise ratio.

Invasive EEG monitoring as a diagnostic modality arose in the context of seizure focus localization in patients with treatment-refractory focal epilepsy. It is typically achieved either by subdural/epidural grid placement or by implantation of intracortical strip or depth electrodes [78], which share a similar risk profile with other invasive monitoring techniques [79]. Despite comparatively poor spatial resolution, intracortical electrodes may prove to be a valuable diagnostic addition to sEEG monitoring in neurointensive care units that routinely use depth electrodes [79, 80].

The concurrent use of EEG with other MMM techniques provides the opportunity to get a more complete picture of ongoing brain physiology, during major pathologic events like seizures, delayed cerebral ischemic events after SAH, or during waves of cortical spreading depression [81–83].

Apart from a few recent studies data are lacking using a combined EEG/MMM methodology [80, 84–86]. One of the most promising areas of research is the characterization of ongoing physiology during seizures, which may potentially help find treatment targets beyond administration of anticonvulsant drugs, with the goal to improve physiology. Seizures in ICU populations are both very common and associated with worse outcomes [85, 87], which is why the characterization of the physiologic signature of seizures through the use of MMM is very promising. With the use of cEEG electrographic seizures have been detected in up to 15% of patients with spontaneous SAH [87, 88], in 15% of patients with post-cardiac arrest syndrome [89], in around one quarter of patients with intracerebral hemorrhage [90, 91], in up to one third of patients with TBI [92, 93] and in one third of patients with CNS infections [94]. By the same token, 16 and 10% of patients in surgical and medical ICU populations are found to have electrographic only seizures [95, 96]. Better understanding of the underlying pathophysiological processes occurring during acute brain injury seizures may allow identification of seizures that deserve more or less aggressive treatment. Lastly, through the combined use of EEG and MMM the so-called ictal-interictal continuum might be characterized further, distinguishing EEG patterns that warrant treatment from those that do not [97].

In animal models researchers have investigated the MMM signature of seizures for decades using varying techniques to measure ICP, CPP, cerebral blood flow , brain oxygen , and cerebral metabolites before, during, and after seizures [99]. Animal models have the advantage of being “clean” models of seizure physiology in contrast to human patients in the neurointensive care unit who tend to have a severe underlying brain pathology and systemic effects such as pneumonia or sepsis to account for [100]. Despite variability among studies regarding the type of animal model, the mode of seizure induction, and the methods employed (including variable sites of MD catheter placement) some comparable results can be extracted, indicating a potential common MMM signature of seizures.

As early as 1939, Penfield and co-workers described an increase of cerebral blood flow during seizures [101] and this finding has been reproduced quite consistently across animal species and seizure models using different types of intraparenchymal probes to measure rCBF [102–105].

Conversely brain tissue oxygen tension eventually decreases with ongoing seizure activity, sometimes following an initial increase of local oxygen tension, indicating a possible initial compensatory effort which eventually ends with decompensation of oxygen delivery [106–109].

Although not consistently reproduced across different disease processes, there are some animal studies showing that prolonged seizures may cause ICP elevations [110, 111]. Consequently CPP may be reduced during seizure activity, unless increased blood pressure during the event compensates the ICP elevation. However CPP is also influenced by various other factors such as presence or absence of functioning autoregulation; therefore, CPP changes during seizures are difficult to investigate and have not been systematically studied in animal models [84].

During experimentally induced status epilepticus initially marked increases in extracellular glucose and lactate levels have been observed, indicating increased energy demand and turnover [112]. This hypermetabolic situation may eventually lead to energy failure as indicated by decreases of extracellular glucose and increases in the LPR after 20–40 minutes of ongoing status epilepticus, as observed in the L-homocysteine induced seizure model in rats [113]. Marked increases of extracellular glutamate and glycerol after prolonged seizure activity have been interpreted as indicative of cell death as a consequence of prolonged status epilepticus [112, 114–118].

MMM studies in humans often face criticism because data is heterogeneous and not always conclusive, perhaps not surprising as most studies include heterogeneous disease processes and disease severities. In clinical practice all efforts are being made to terminate seizures as early as possible. However, a clear picture of MMM changes during seizure activity can often only be seen after prolonged status epilepticus, as evident from the animal studies described above. Figure 6.2 illustrates typical changes of brain oxygen and perfusion preceding, during, and after a prolonged electrographic seizure in a patient with a high-grade spontaneous subarachnoid hemorrhage (Fig. 6.2).

Apart from the dynamics of MMM parameters, secondarily developing lesions are often encountered following acute brain injury such as ischemic and hemorrhagic lesions and will affect the recorded brain physiology. Pathophysiologic changes of primary injury may at times be difficult to differentiate from secondary events. Moreover, variable disease severity at baseline is likely reflected in different baseline MMM values which is hypothetically the reason why pooled data is less likely to produce statistically significant results than comparison of within-subject changes [81, 85]. Despite these “real life” limitations a decent body of scientific evidence has been generated describing MMM changes during seizure activity in human patients.

Ictal hyperperfusion is a well-studied phenomenon, and has been seen in numerous studies using positron emission tomography, CT perfusion, MR perfusion, and Xenon CT in patients with and without acute brain injury, but studies using invasive rCBF measurement are sparse [119, 120]. In line with animal data, one study found that rCBF during seizures increases with continuous seizure activity, after approximately 10 minutes [85]. Also consistent with the animal data, PbtO2 has been found to decrease during seizures prior to the increase of rCBF, suggesting that perfusion may increase as a compensatory reaction to hypoxia [85]. In a group of 10 TBI patients with seizures compared to a matched control group of TBI patients without seizures, the occurrence of seizures correlated with a prolonged sustained increase of ICP [121]. In the study by Claassen et al. electrographic seizures were accompanied by an immediate increase of ICP, CPP, heart rate, MAP, respiratory rate, and minute ventilation [85]. These changes in cardiovascular parameters are common in partial and generalized epilepsies and are consistent with previous studies [122, 123].

Few studies are available evaluating brain MD changes during seizure activity in humans. In 30 subjects with TBI the occurrence of electrographic seizures coincided with a decrease of extracellular glucose concentrations [124] and this finding was reproduced in a recent study in 29 TBI patients in whom extracellular glucose concentrations during seizures were significantly lower than at baseline (mean 0.8 vs. 1.7 mmol/l) [81]. Administration of systemic insulin lowers both systemic and brain extracellular glucose concentrations and correlates with markers of metabolic stress in the brain, namely increased LPR and increased glutamate concentrations [65, 125]. The few available studies investigating lactate and LPR during seizures are in line with these findings . A study by During and colleagues investigated the effect of spontaneously occurring seizures in (otherwise healthy) patients with treatment-refractory complex partial seizures, after bilateral implantation of MD catheters in both hippocampi. Partial seizures with secondary generalization were accompanied by a mean 90% lactate increase in the first 60–90 minutes after seizure onset. This increase was restricted to the side of seizure initiation [126]. Pyruvate concentrations were not determined in this study.

MD changes may last beyond seizure cessation. In a study of non-convulsive seizures in TBI, Vespa et al. found prolonged and significantly higher LPR values in the seizure group compared to the group without seizures. This effect lasted beyond 100 hours after admission [121]. However, this finding was not reproduced in the study by Claassen et al., who did not find significant changes of MD parameters related to the occurrence of electrographic seizures. The discrepancy between both papers is possibly due to the statistical methods employed or due to the different entities studied (TBI versus SAH) [127]. In a recent study in patients with severe TBI undergoing MMM including MD, the comparison of 12 patients with seizures and 8 patients without seizures showed a numerically higher hourly LPR burden in the seizure group (as quantified using the area under the curve). This finding was not statistically significant in a pooled analysis but reached significance using a within-subject-design analyzing LPR changes of single patients on an hour by hour basis [81].

Extracellular glutamate is a key factor in abnormal neuronal discharges and has been implicated in triggering and maintaining seizures, and in kindling epilepsy [128]. MD studies in patients with temporal lobe epilepsy have shown chronically increased extracellular glutamate concentrations on the side of the epileptic focus leading to excitotoxicity, which hypothetically entertains the increased likelihood of further seizures [129]. Ronne-Engström and colleagues investigated extracellular amino acid levels during seizures and during the period immediately following the seizures in seven patients with epilepsy of focal origin by implanting a MD device into the epileptic focus. Extracellular glutamate concentrations rose approximately sevenfold during spontaneously occurring seizures, but normalized immediately after [130]. In patients with TBI, seizures were associated with the occurrence of transient extracellular glutamate increases [84].

Glycerol is an essential ingredient of cell membranes and serves—even more so than the LPR—as a direct marker of cell damage. However, the role of extracellular glycerol in the context of neurointensive care patients with seizures is poorly studied [131]. One of the few studies included just two patients with TBI and electrographic seizures in whom MD detected a higher mean glycerol concentration in comparison with patients who had no seizures. These concentrations, however, were average concentrations during days 1–3 and 7–9 post-injury and not time-locked to the occurrence of seizures [132].