Language laterality determined by MEG mapping with MR-FOCUSS

Abstract

Magnetoencephalography recordings were made on 27 patients with localization related epilepsy during two different language tasks involving semantic and phonological processing (verb generation and picture naming). These patients underwent the semi-invasive intracarotid amobarbital procedure (IAP), also referred to as the Wada test, to determine the language-dominant hemisphere. Magnetoencephalography (MEG) data were analyzed by MR-FOCUSS, a current density imaging technique. A laterality index (LI) was calculated from this solution to determine which hemisphere had more neural activation during these language tasks. The LIs for three separate latencies, within each language task, were calculated to determine the latency that correlated best with each patient’s IAP result. The LI for all language processing was calculated for the interval 150–550 ms, the second LI was calculated for the interval 230–290 ms (Wernicke’s activation), and the third LI was calculated for the interval 396–460 ms (Broca’s activation). In 23 of 24 epilepsy patients with a successful IAP, the LIs for Broca’s activation, during the picture naming task, were in agreement with the results of the IAP (96% agreement). One of three patients who had an undetermined or bilateral IAP had an LI calculated for Broca’s activation (396–460 ms) that agreed with intracranial mapping and clinical testing. These results indicate an 89% agreement rate (24 of 27) for magnetoencephalographic LI determination of the hemisphere of language dominance.

Introduction

Determining the hemisphere that is language-dominant or contains the major language areas (Broca’s and Wernicke’s) is imperative prior to many neurosurgical procedures, including the most common epilepsy surgery procedure, the temporal lobectomy. The medical standard for determining the language-dominant hemisphere is the intracarotid amobarbital procedure (IAP), also known as the Wada test [1], [2]. The IAP usually identifies the hemisphere containing eloquent cortex for language and reduces the risk of postsurgical language morbidity, such as aphasia. Newer, noninvasive diagnostic techniques for determination of the hemispheric dominance for language and memory are under development because of the risks associated with the IAP. Magnetoencephalography (MEG) [3] studies have been performed to find a noninvasive alternative to the IAP. Using a MEG language-evoked response protocol, Simos et al. [4], Breier et al. [5], [6], and Zouridakis et al. [7] calculated a laterality index using MEG by counting the number of single equivalent-current dipole (ECD) localizations found in each hemisphere. Simos et al. [4] found 14 of 16 or 87.5% agreement for dipoles localized to the left hemisphere during an auditory word matching language task. Breier et al. [5] found that MEG language laterality correlated highly (R > 0.87) with IAP results in 19 children. In another study Brier et al. [6] reported that 13 of 15 (87% agreement) right-handed subjects had more dipoles in the left compared with the right hemisphere during an auditory word recognition task. Zouridakis et al. [7] found that all 11 of 11 right-handed normal subjects had more ECD localizations in the left hemisphere. Recently Papanicolaou et al. [8] compared MEG ECD localizations with IAP results in 85 patients and obtained a high degree of concordance (87%). Kober et al. [9] used the MEG current source strength in each hemisphere to determine the dominant hemisphere. They found all 15 of their right-handed normal subjects were left hemisphere dominant for language. The results of these studies indicate MEG may be a promising method for determining the dominant hemisphere for motor speech.

Functional MRI (fMRI) studies have also been performed to find a noninvasive alternative to the IAP [10], [11], [12], [13]. The asymmetry index for fMRI is calculated by counting voxels of activation in each hemisphere; the hemisphere with more activation is considered the language-dominant hemisphere. Woermann et al. [10] used a Picture Naming task during fMRI and obtained 91% concordance with IAP results in 100 patients with epilepsy. A Verb Generation task has also been used during fMRI to calculate an asymmetry index to determine the language-dominant hemisphere [11]. In this study, IAP and fMRI results agreed in 27 of 30 patients (90% accuracy). Binder et al. [12] used a single-word semantic decision task and obtained a correlation of 0.96 between fMRI lateralization and the IAP results in 22 patients with epilepsy. Rutten et al. [13] used four different language tasks and reported an overall 83% correlation accuracy between fMRI laterality and the IAP results in 18 patients. Abou-Khalil and Schlagger [14] recently asserted that fMRI was not yet ready to replace the IAP and ECoG mapping until a battery of fMRI paradigms can be validated for identifying only critical language cortex. When the IAP is inconclusive, direct cortical stimulation using implanted electrodes is used to determine the cortical language areas.

As part of our investigation of evoked responses to language stimuli in normal subjects, we found that MEG signals from Wernicke’s area located in the superior temporal gyrus (STG), angular gyrus (AG), and/or supramarginal gyrus (SMG) were active approximately 239 ms after stimulus onset [15]. We also found that Broca’s area located in the inferior frontal gyrus (IFG) and/or the middle frontal gyrus (MFG) was active approximately 436 ms after stimulus onset [15]. These latencies are similar to those reported in MEG studes by Simos et al. [4], Levelt et al. [16], and Helenius et al. [17]. A third language area, located in fusiform gyrus, has been identified with fMRI and PET [18], [19]. This language area, known as the basal temporal language area (BTLA), can also be imaged by MEG. Our studies indicate activation of the BTLA occurs between 149 and 185 ms after stimulus onset (see the accompanying paper [20]).

MEG language mapping studies in our laboratory have been performed using a multiple source analysis technique, MR-FOCUSS [15], [21]. We developed this technique for use when the conditions for ECD analysis technique are unfavorable. The ECD technique is a robust method when a single nondistributed stationary source is present, such as the early cortical response to somatosensory stimulation or an isolated interictal epileptic spike. Multisource analysis is useful for other types of cortical activation such as studies of language, memory, or analysis of epileptic seizures. In the accompanying article, we provided details of the MR-FOCUSS technique that are useful in identifying the critical latencies and locations of cognitive activity in our evoked response studies of language [20]. This article introduces a new development in MEG language analysis, the laterality index (LI), which we have applied to our evoked response studies of language. Our laterality index is similar to the asymmetry index used in fMRI analysis of language. We have also explored whether the LI should be applied to the full data set (after activation of the visual cortex at 100–150 ms) or at the times when certain language areas seem most active.