Interhemispheric Preponderance of Noninvasive Source Algorithms: Commentary on Two Recently Published Articles

To the Editor: We read with great interest the articles of Kobayashi et al. (1), reporting on the accuracy and pitfalls of dipole source modeling of epileptic spikes by using simulated spike generators of different cortical extents, and Holmes et al. (2), reporting on localized mesial frontal and frontopolar discharges in patients with absence seizures. During the last two decades, source modeling techniques such as single equivalent dipole estimations or three-dimensional current density distributions have been increasingly used for noninvasive localization of epileptiform and event-related activity. Irrespective of the method used, the scalp-recorded EEG or magnetoencephalogram (MEG) signals do not provide conclusive information about the localization and distribution of putative sources, and a priori assumptions must be implemented into these methods to overcome the nonuniqueness of the inverse solution. With dipole models, the source of the voltage field recorded from the scalp is considered to be a theoretical pointlike dipole, equivalent in location and orientation to the generating cortical area. The intracerebral field of epileptiform activity, however, is often complex and might involve extended areas of active cortex rather than pointlike dipoles, perhaps most evident in the case of primary generalized spike–wave epilepsy. In their study, Kobayashi et al. (1) show perfectly how the locations of estimated dipoles can be very misleading when the source area of a spike is extended over a large area of cortex, despite (or because of) a small residual variance. This problem was particularly conspicuous with very large bilateral sources covering the convexities and mesial surfaces of the frontal lobes. Dipole modeling of this widespread cortical activity resulted in consistent dipoles with minimal residual variance situated in the cingulate gyri, 3 cm away from the cortical generating surface, irrespective of inclusion or exclusion of mesial frontal areas in the source simulation. We too have been bothered by the propensity for noninvasive source models to localize widespread frontally predominant generalized activity into the bilateral anterior cingulate gyri. This interhemispheric preponderance is not restricted to dipole models but occurs also with distributed source models. Holmes et al. (2) present interhemispheric solutions for spike-wave complexes of absence seizures by using both a dipole model (BESA) and a distributed linear source model (low-resolution electromagnetic tomography, LORETA). With both methods, they found similar source solutions for the spike-wave complexes within bilateral frontal interhemispheric cortical areas, including the anterior cingulate gyri and supplementary sensorimotor areas (Figs. 5, 9, and 10 in their study). These findings were interpreted as evidence that absence seizures are not truly “generalized” but rather involve selective cortical networks within mesial frontal and orbital frontal regions. Although these speculations are supported by other lines of evidence, we do not have much confidence in this interpretation of the source modeling data, in that the findings may simply reflect the limitations of the source models. In Fig. 1, we demonstrate the interhemispheric preponderance of noninvasive source models by modeling four different frontally predominant symmetrical EEG patterns: two real, physiologic signals (3-Hz spike-wave, eye blink) and two artificial signals (10-Hz sine wave, random noise). LORETA clearly offers similar interhemispheric solutions for all four symmetrical EEG patterns, irrespective of the underlying signal. In the case of the eye blink, LORETA offers an interhemispheric solution although the two sources are a priori known to be separate and bilateral (the eyeballs act as dipoles, with the negative poles oriented toward the retinas). The interhemispheric solutions for the generalized sine wave activity and random noise exemplify the problem of modeling sources for widespread bilateral signals. Although the noncerebral examples in the figure obviously violate the basic assumptions of LORETA by computing the inverse solutions of extracortical sources, they were chosen here to illustrate a point. Source models, irrespective of their basic assumptions, just offer the putative source(s) that best explain(s) a given scalp voltage field. Given that a symmetrical EEG voltage field can be explained by either (a) a single interhemispheric source, or (b) by two or more symmetric and bilaterally synchronous hemispheric sources, the source models provide us with the simpler solution, which is explanation (a). Low-resolution electromagnetic tomography (LORETA) provides virtually identical interhemispheric solutions for four different symmetrical EEG patterns (calculated during individual time sample points (red vertical lines) or epochs (red horizontal lines); “weighted” denotes adjustment of each electrode in the 10-Hz sine wave and random noise signals to accord with the magnitude of a typical 3-Hz spike-wave discharge, as shown on the left). This preponderance for simple interhemispheric solutions becomes problematic when examining the many EEG phenomena that reflect synergistic subcorticocortical activity, such as generalized epileptiform activity or sleep patterns, which show symmetrical EEG voltage fields that are most likely related to widespread bilateral hemispheric cortical activity and not the result of a distinct, well-defined, midline near generator. In this regard, we are increasingly being confronted with source modeling studies that present bilateral sources in and around the anterior cingulate gyri for a multitude of diseases and conditions. Indeed, it is beginning to seem that activity in the anterior cingulate gyrus may be responsible not only for most psychiatric disturbances such as obsessive–compulsive disorder, depression, and schizophrenia, but also for generalized epilepsy syndromes as well as physiologic sleep patterns. As emphasized by Kobayashi et al. (1), the neurophysiologic principles underlying generation of the recorded EEG signals must always be taken into consideration when interpreting the results provided by these source-modeling methods. It also is important, however, to avoid self-fulfilling prophesies whereby potentially meaningless source-localization data are justified by a priori neurophysiologic constructs. In the case of 3-Hz spike-wave activity, prevailing neurophysiologic concepts implicating a mesial frontal predominance within the thalamocortical circuitry underlying primary generalized epilepsy are not actually further supported (or discredited) by deep midline source solutions for the widespread cortical activity: the findings are simply not helpful one way or the other. We suggest that all interhemispheric solutions of noninvasive source models should be regarded with suspicion whenever widespread and symmetrical EEG voltage fields are analyzed.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.