On noninvasive source imaging of the human K-complex
Highlights
• Large, extended superficial cortical sources may pose special problems for noninvasive source localization, in particular, a propensity toward deep midline source solutions. • Various EEG and MEG source imaging techniques were applied to model the human K-complex, and source solutions were compared with the intracranial cortical localization. • No combination of the tested forward and inverse (dipole and distributed source) models could resolve the propensity of the localization algorithms to return invalid deep midline solutions for the extended superficial cortical source.
Introduction
The application of EEG and MEG source imaging to the human K-complex has produced conflicting and occasionally implausible results (Ueno and Iramina, 1990, Lu et al., 1992, Iramina and Ueno, 1996, Numminen et al., 1996, Colrain, 2005).
We have recently described the cortical distribution of the main surface negative peak of the K-complex based on intracranial EEG recording in patients with intractable epilepsy (Wennberg, 2010). In short, compatible with the classic known frontal midline maximum of the scalp EEG K-complex (Davis et al., 1939, Brazier, 1949, Roth et al., 1956), the maximal intracranial field is found over the midline frontal regions, the field reversing medially above the cingulate gyrus and laterally above the inferior temporal gyrus (Wennberg, 2010). Polarity is reversed within the white matter of the frontal lobes, as well as in more distant subcortical structures, confirming cortical generation within the anterior and superior frontal lobe cortices (Fig. 1).
The surface negative peak of the K-complex (sometimes referred to as the N550) is the highest amplitude potential in the normal human EEG (Loomis et al., 1938, Roth et al., 1956, Colrain, 2005). Paradoxically, at the cellular level, it is associated with a widespread cortical down-state characterized by suppression of neuronal activity (Amzica and Steriade, 2002, Cash et al., 2009, Cserca et al., 2010, Dalal et al., 2010, Le Van Quyen et al., 2010). The polarity distribution of the intracranial K-complex electric field is consistent with the classical dipole layer model presumed to underlie generation of most (and certainly all large amplitude) potentials in the EEG (Gloor, 1985, Amzica and Steriade, 1998, Wennberg, 2010). In line with the dipole layer model, the surface negative peak of the K-complex must represent the effects of either: (a) summated excitatory post-synaptic potential (EPSP) inputs to the superficial apical dendrites of frontal lobe cortical pyramidal cells, or (b) summated inhibitory post-synaptic potential (IPSP) inputs at the deeper cell soma level of these same pyramidal cells. The bulk of the available evidence obtained from animal and human microelectrode recordings supports the second option: i.e., that synchronized hyperpolarizing IPSPs synapsing on pyramidal cell bodies in deeper layers of the cortical mantle are primarily responsible for initiation of the principal negative wave of the K-complex (Cash et al., 2009, Cserca et al., 2010, Dalal et al., 2010, Le Van Quyen et al., 2010).
The localization of the presumptive inhibitory inputs is unknown; however, it is reasonable to implicate the thalamus (and possibly other subcortical structures) given the known involvement of reciprocal thalamocortical circuits in slow oscillatory sleep patterns (Amzica and Steriade, 1998, Amzica and Steriade, 2002, Steriade and Amzica, 1998). The paucity of cortical neuronal activity during the K-complex down-state suggests that cortical contributions to the inhibitory inputs may be less likely.
The extent and timing of synchronization during the negative K-complex peak is also incompletely understood. Does cortical synchronization arise from synchrony of the subcortical hyperpolarizing IPSPs synapsing simultaneously on the soma of the relevant frontal lobe cortical pyramidal cells? Or, once initiated in one cortical region, does the K-complex propagate across the cortex, either through synaptic transmission similar to sleep slow wave propagation (Amzica and Steriade, 1995) or perhaps even as a result of endogenous electric field activity (Fröhlich and McCormick, 2010)? If synaptic propagation occurs, would it originate at the inhibitory input neurons (e.g., within the thalamus) or would it arise from corticocortical connectivity? It has been hypothesized, based on high density scalp EEG recordings and noninvasive source localization, that sleep slow waves and K-complexes may represent traveling waves propagating across the cortical surface, usually from front to back, guided along a deep interhemispheric “cingulate highway” (Massimini et al., 2004, Murphy et al., 2009). However, in intracranial EEG recordings, we could find no definite evidence to support the hypothesis of K-complexes as traveling waves, and indeed the cingulate gyrus appeared uninvolved in K-complex generation (Wennberg, 2010). Nevertheless, in some individuals, K-complexes do at times show anterior–posterior time lags of wave onsets or peak maxima in scalp EEG recordings, and this is an unexplained phenomenon.
Recent fMRI studies have correlated K-complexes with positive blood oxygen level-dependent (BOLD) signal changes in subcortical (brainstem and thalamus) and cortical regions, the latter involving mainly paracentral, posterior and inferior parieto-occipital, superior temporal, and midline cingulate and paracingulate structures (Caporro et al., 2012, Jahnke et al., 2012). It is of interest to note that the areas of BOLD changes fairly accurately demarcate the areas of cortex not involved in generation of the negative wave of the EEG K-complex. It is conceivable that the subcortical (and perhaps even the cingulate) BOLD changes might reflect the source(s) of the inhibitory neuronal activity acting upon frontal pyramidal cell bodies to produce the negative wave of the K-complex. However, the other (predominantly posterior and inferior) cortical areas of BOLD changes, situated as they are outside the (anterior and superior) cortical areas of K-complex generation, are most likely representative of increased neuronal activity during the subsequent up-state associated with the later, lower amplitude, positive wave of the K-complex (the temporal resolution of fMRI being insufficient to identify neuronal activity confined to any single wave component of the K-complex; Jahnke et al., 2012). Thus, perhaps unexpectedly, fMRI studies have revealed no K-complex associated BOLD changes in the large frontal cortical areas responsible for generation of the highest amplitude potential in the human EEG.
The burgeoning research interest in sleep neurophysiology is certain to be accompanied by increasing attempts to incorporate noninvasive EEG and MEG source localization techniques to study various aspects of the K-complex, and it will be important to resolve whether or not these techniques are able to accurately depict the brain areas involved in K-complex generation. Previous reports describing EEG or MEG source localization of the K-complex using dipole mapping or, in one case, distributed source modeling, presented various source solutions in deep centrotemporal (Ueno and Iramina, 1990, Iramina and Ueno, 1996), inferior parietal (Lu et al., 1992) or frontal and parietal areas (Numminen et al., 1996), all of which are at odds with the true cortical localization of the intracranially recorded K-complex electric field (Wennberg, 2010).
One particular problem for noninvasive source localization of the K-complex is the inherent tendency of inverse models to provide deep interhemispheric solutions for large, bilaterally synchronous cortical sources (Nunez, 1990, Romani and Pizzella, 1990, Hämäläinen et al., 1993, Numminen et al., 1996, Colrain, 2005, Kobayashi et al., 2005, Zumsteg and Wennberg, 2005). This may be especially relevant to source localization using single equivalent current dipole models, although it is unclear whether cortically-constrained distributed source models will necessarily perform better (Zumsteg and Wennberg, 2005).
In this study we apply a variety of EEG source localization techniques to the same K-complexes recorded in the same 6 patients with epilepsy that were presented in our study defining the intracranial cortical localization of the human K-complex (Wennberg, 2010). In this way, the true neurophysiologic “forward model” of the EEG waveforms is known. Furthermore, we apply different source localization techniques to K-complexes recorded with MEG in 3 of the patients and to K-complexes recorded with MEG and high density EEG in an additional subject without epilepsy. Specifically, we compare single moving, fixed coherent and multiple dipole models with standardized low-resolution electromagnetic tomography (sLORETA) and low-resolution electromagnetic tomography (LORETA) distributed source modeling, techniques theoretically better suited to modeling extended cortical sources. The goal of the study is to determine whether EEG source imaging (ESI) or magnetic source imaging (MSI) can model human K-complexes in a valid and reliable fashion.
Section snippets
EEG: subjects and recordings
EEG source localization was performed on the data acquired from six patients (patients 1–6) with medically-refractory epilepsy during their routine pre-surgical stereoelectroencephalographic recordings. This is the same data set of simultaneously acquired intracranial and scalp EEG used to determine the intracranial localization of the K-complex (Wennberg, 2010). Details of the patients’ epilepsies, neuroimaging and anti-epileptic medications have been presented previously (Wennberg, 2010). All
EEG: dipole mapping
Fig. 2 shows the results of FCD mapping in patients 1–6. Localization results were essentially identical whether the generic FEMi volume conductor or the patients’ own BEM volume conductors were used as the forward model. In all cases deep interhemispheric dipole solutions were provided to explain the large mid frontal maximum electric field. Comparison with the known intracranial superficial frontal cortical source (Fig. 1; Wennberg, 2010) reveals that these dipole source localizations are not
Discussion
This study sought to determine whether source localization techniques that are currently used to describe the underlying generators of sensory, motor or cognitive brain responses in humans could also be used to model human K-complexes in a valid and reliable fashion. The K-complex was selected for study for two reasons: (a) burgeoning research interest in sleep neurophysiology would benefit from being able to noninvasively model the K-complex with ESI or MSI techniques, and (b) as the largest
Acknowledgments
Taufik Valiante, MD, PhD, FRCSC, and Andres Lozano, MD, PhD, FRCSC, performed the surgical implantations of the intracranial electrodes. Nat Shampur, RET, and his team of EEG technologists provided expert technical assistance.
References (51)
- et al.
The functional significance of K-complexes
Sleep Med Rev
(2002) The electrical fields at the surface of the head during sleep
Electroencephalogr Clin Neurophysiol
(1949)- et al.
Functional MRI of sleep spindles and K-complexes
Clin Neurophysiol
(2012) - et al.
Supplementary motor area activation preceding voluntary movement is detectable with a whole-scalp magnetoencephalography system
Neuroimage
(2000) - et al.
Endogenous electric fields may guide neocortical network activity
Neuron
(2010) - et al.
Boundary element method volume conductor models for EEG source reconstruction
Clin Neurophysiol
(2001) - et al.
Confidence limits of dipole source reconstruction results
Clin Neurophysiol
(2004) - et al.
To wake or not to wake? The two-sided nature of the human K-complex
Neuroimage
(2012) - et al.
Epileptic source localization with high density EEG: how many electrodes are needed?
Clin Neurophysiol
(2003) - et al.
Generator sites of spontaneous MEG activity during sleep
Electroencephalogr Clin Neurophysiol
(1992)