Language tasks used for the presurgical assessment of epileptic patients with MEG

ARTICLE

Auteur(s) : Mona Pirmoradi^1,2, Renée Béland^1,2,3, Dang K Nguyen⁴, Benoit A Bacon⁵, Maryse Lassonde^1,2

¹Centre de Recherche en Neuropsychologie et Cognition, Université de Montréal
²Centre de Recherche, Centre Hospitalier Universitaire Sainte-Justine
³École d'orthophonie et d'audiologie, Université de Montréal
⁴Neurologie, Centre Hospitalier Universitaire de Montréal (Notre-Dame)
⁵Psychology, Bishop's University, Sherbrooke, Quebec

Article reçu le 13 Novembre 2009, accepté le 19 Avril 2010

The most commonly used treatment for epilepsy is pharmacotherapy (Killgore et al., 1999). However, an estimated 35% of patients with epilepsy develop medically intractable epilepsy. In these cases, surgery is widely used to remove the epileptogenic zone (Gates and Dunn, 1999). Resective epilepsy surgery is performed mainly in the temporal and frontal lobes (selective amygdalo-hyppocampectomies, anterior temporal lobectomies, or tailored temporal or frontal corticectomies). However, it must be previously determined that the resection will not have any substantial consequences on cognitive functions, such as language or memory. Determining the language dominant hemisphere and localizing the language function is particularly important in epileptic patients because they present greater variability in language dominance than neurologically healthy individuals (Berl et al., 2005). It is estimated that 94% to 96% of healthy right-handers and 74% of left-handers have left-hemisphere language dominance (Pujol et al., 1999; Springer et al., 1999). In contrast, 63% to 96% of right-handed epileptic patients and 48% to 75% of left-handed or ambidextrous epileptic patients show left-hemisphere language dominance (Helmstaedter et al., 1997; Springer et al., 1999).

The medical standard for determining the language dominant hemisphere prior to surgical resection is the intracarotid amobarbital test (IAT), also known as the Wada test (Wada and Rasmussen, 1960), hereinafter referred to as the IAT. It consists of an injection of sodium amobarbital into the left or right internal carotid arteries. This causes a temporary arrest of function in each hemisphere for approximately six to ten minutes, during which the unanaesthetised hemisphere is functionally assessed. Tasks used to assess language dominance include naming common objects, reading single words aloud, counting, and spelling single words. A major drawback of this test is that it determines lateralization only, and does not allow intrahemispheric localization of language functions. Moreover, because it is relatively invasive, this technique cannot be used with normal volunteers and is difficult to use with children. Finally, the IAT is associated with risks of stroke, infection, and haemorrhage (English and Davis, 2010).

When surgery is believed to put language functions at risk, electrical stimulation mapping (ESM) is used to obtain information on the specific location of the language areas. Using an electrical current, specific brain areas are stimulated while the patient is awake and performing a linguistic task. This method is the most reliable and direct way to localize language areas. However, it has several disadvantages: it is very invasive, there are associated risks such as stimulation-induced seizures, it requires that patients be awake, it is costly, and it cannot be revisited if results are ambiguous (McDermott et al., 2005).

Because of the risks and limitations associated with more invasive techniques of language exploration, it is very important to develop alternate, minimally invasive or noninvasive techniques that offer both lateralization and intrahemispheric localization. Recent advances in imaging technology have produced noninvasive and minimally invasive techniques: functional magnetic resonance imaging (fMRI), positron emission tomography, near infra-red spectroscopy, transcranial magnetic stimulation, and MEG to localize language functions. Functional magnetic resonance imaging is the technique that has received the most attention as a possible replacement for the IAT. However, this method presents certain disadvantages: it is very expensive, requires the patient's cooperation, and is less suitable for young or mentally challenged individuals (Pelletier et al., 2007). Of the remaining techniques, MEG is the only completely noninvasive technique offering excellent temporal and spatial resolution that can be used with children.

This paper reviews and examines the efficiency of different language tasks employed in studies that have used MEG to lateralize and localize intrahemispheric language functions in the human brain. The focus is on adaptability to a paediatric population. Following a brief description of the functioning of MEG, a review of studies that have used MEG to lateralize language functions is presented including an overview of the language comprehension and language production tasks used. The second part of the review focuses on studies aimed to determine the intrahemispheric localization of language within the dominant hemisphere. Some of these studies have used language comprehension tasks and others language production tasks. A total of 37 studies from the last decade are reviewed, all of which were conducted either with control subjects or in the context of presurgical assessment of epileptic patients, patients with brain tumours, and other types of patients, including adults and children.

Magnetoencephalography

Hemispheric language lateralization

When patient populations are studied, MEG and IAT findings are often compared. It is important to note that because it is invasive, the IAT cannot be performed on control subjects. Thus, in studies assessing control subjects, handedness is commonly used to determine the accuracy of MEG lateralization findings. However, the discordance between handedness and hemispheric dominance for language in normal populations makes this method problematic (Pujol et al., 1999; Springer et al., 1999).

Language comprehension

Other tasks are more complex and require participants to pay close attention to the stimuli. In a study using a categorization task, controls were instructed to listen to pairs of words belonging to the same or different semantic categories and silently count the number of different semantic pairs. The same procedure was followed with different tones, where participants had to determine whether pairs of tones were the same or different. It was expected that these two tasks would yield opposite lateralization patterns. As expected, greater left hemisphere activation was seen in 87.5% of subjects with the word-matching task, whereas 62.5% of subjects showed asymmetries favouring the right hemisphere with the tone-matching task (Simos et al., 1998).

Breier et al. (1999b) found left hemisphere dominance in 87% of right-handed controls when determining whether a word was repeated, as opposed to 30% when determining whether a low note was repeated. Gootjes et al. (1999) asked controls to determine whether the first and last item in a group of vowels, tones, or piano notes were the same. When looking at activations only for groups in which the first and last item differed, they found that left hemisphere responses to vowels were significantly stronger than for tones or piano notes. Kirveskari et al. (2006) asked Finish-speaking participants to decide whether pairs of tones and Finish vowels were the same or different. When comparing the laterality index for strengths of the auditory-cortex 100 ms responses to vowels vs tones, they found left hemisphere dominance in 80% of right-handed subjects and right hemisphere dominance in 70% of left-handed subjects.

A frequently used word recognition task involves words that are presented either visually or auditorily, with some words being targets and others distractors. Target stimuli are usually presented for study before the test session. Target stimuli are then repeated and mixed with different distractors in each test block. Participants are asked to lift their index finger whenever they detect a repeated word (target). When this task was used with epileptic patients, MEG results showed high concordance with IAT results, varying between 86% and 92% of correct lateralization (Breier et al., 1999a; Breier et al., 2001; Papanicolaou et al., 2004; Maestú et al., 2002; Doss et al., 2009). One group found that, when controlling for IQ and excluding patients with below average scores, the concordance between MEG and the IAT increased from 75% to 90% (Merrifield et al., 2007). It therefore appears that when patients show reduced cognitive capacity, MEG is not 100% specific for language lateralization.

In a semantic judgment task, McDonald et al. (2009), found 75% concordance between MEG and the IAT when examining the laterality of temporoparietal sources, versus 100% with the IAT when examining the laterality of frontal sources. Hirata et al. (2009) used a reading task and found 85% concordance with the IAT in a sample of 60 patients. Finally, when lateralization was determined using both a reading and a picture naming task, it was possible to identify speech-related dominant hemispheric activity in most subjects (Kober et al., 2001).

In summary, although complexity of tasks and stimuli varies greatly, the findings are promising for the use of MEG to lateralize language functions with language comprehension tasks. In the studies that compared frontal and temporal activations to better identify lateralization (Fisher et al., 2008; Kim and Chung, 2008; McDonald et al., 2009), it appears that frontal activations were more accurate. It is important to note that, as indicated in table 1, 10 of the 16 studies summarized in this section compared MEG to IAT findings, with concordance varying between 71% and 94%. Studies comparing handedness with MEG findings showed greater variability in concordance (47% to 100%), and results should be interpreted with caution.
Table 1 MEG studies investigating hemispheric language lateralization.

Language production

Overall, it appears that verb and word generation tasks are more accurate in determining language function lateralization with MEG. Nevertheless, most of these studies were conducted in controls, such that the findings could not be compared with the IAT. However, Bowyer et al. (2005b) compared MEG findings with the IAT and found 82% concordance with the verb generation task.

Intrahemispheric language localization

Language comprehension

Other more complex linguistic tasks have been studied using MEG. Martin et al. (1993) used a listening task in a case study using preoperative MEG to map the speech-receptive cortex in response to auditorily presented phonemes. The consonant-vowel syllables “da” and “ga” were presented. Patients had to covertly count all stimuli. Peak activation was observed anterior to Wernicke's area.

Härle et al. (2002) used a decision-making task in which drawings of objects were presented to German-speaking subjects. In two separate tasks, subjects had to indicate whether the name of the object was masculine or feminine or whether the object was man-made or natural by pressing a button. The grammatical gender decision task was expected to trigger brain activity around 200 ms during the retrieval of morphological information, and the activity was expected to be found predominantly in the left hemisphere. In contrast, the control task, which focused on semantic processes only, was expected to show bilateral activation. Results showed a left-temporal focus of activity 150-275 ms after stimulus onset in the gender decision compared to the semantic classification task, which showed right fronto-central activation as well as more extensive left hemispheric activity in the gender decision task 300-625 ms after stimulus onset.

Three studies (Breier et al., 1999b; Papanicolaou et al., 1999; Sun et al., 2003) used auditory recognition or decision tasks using words, tones, and pictures. Activation was found in the temporal lobe in the dominant hemisphere for all three tasks.

McDonald et al. (2009) used a semantic judgment task to investigate language comprehension. They hypothesized that language-related activity would spread along a posterior to anterior gradient, becoming increasingly left-lateralized in the temporoparietal and frontal lobe regions of interest. Activity was observed in the visual cortex bilaterally from 80-120 ms in response to novel words. Thereafter, activity spread to the fusiform cortex (160-200 ms) and was dominated by left hemisphere activity in response to novel words. From 240-450 ms, novel words produced activity which was left-lateralized in frontal and temporal lobe regions, including the anterior and inferior temporal, temporal pole and pars opercularis, as well as bilaterally in the posterior superior temporal cortex.

The word recognition task, described above in the first section, is probably the most extensively used task with MEG for the intrahemispheric localization of language functions. It has been used with both visual and auditory stimuli and has yielded promising results for localizing activity sources in both the frontal and temporal lobe (Breier et al., 1999a; Simos et al., 1999; Breier et al., 2001; Papanicolaou et al., 2004; Breier et al., 2005). This task has been performed using visual and auditory modalities. Overall, sources of late activity have been observed in the following areas with this task: the posterior part of the superior and middle temporal gyri, the angular and supramarginal gyri, the mesial aspects of the temporal lobe, the inferior frontal areas of the left hemisphere, and the basal temporal areas, although using the visual mode only. Moreover, it is important to note that bilateral activity is often observed in these areas. In three of the studies that used this task, very large samples of control participants (n = 97; Papanicolaou et al., 2006) and large patient populations (n = 100; Papanicolaou et al., 2004; Breier et al., 2005) were studied. Moreover, children were included in some samples. In the large control group study, significant bilateral activity was centred in the superior temporal gyrus (STG) and activity was lateralized to the left middle temporal gyrus (MTG) after 150 ms. These findings were consistent across age, gender, and variation in task characteristics, such as presentation mode or number of stimuli used (Papanicolaou et al., 2006). One group examined the cross-language generalizability of this task with Spanish-speaking patients with epilepsy, and found activation in the left temporoparietal areas and the inferior frontal and insular regions (Maestú et al., 2002). Other groups that attempted to validate this task (Lee et al., 2006; Mohamed et al., 2008) confirmed activation in Wernicke's area.

One group (Levelt et al., 1998) used a reading comprehension task to localize language functions by visually presenting four categories of sentence endings:

• – probable final words;
• – semantically appropriate but unexpected endings;
• – anomalous endings;
• – semantically inappropriate endings that started with the same phonemes as the most probable word.

Words were presented one at a time, and participants were instructed to concentrate on the meaning of the sentences. The cortical structures most consistently involved with comprehension were located near the left auditory cortex. The inappropriate final words evoked longer activation (250-600 ms). This activation could be related to the analysis of the meaning of the word and its role in the sentence. Kober et al. (2001) conducted a silent reading task with words presented visually to German-speaking participants. Wernicke's area was localized in the posterior part of the superior temporal gyrus and Broca's area was localized in the left frontal gyrus in all subjects. Hirata et al. (2009) also used a silent reading task with healthy subjects and patients to examine local oscillatory changes in the brain. Activation profiles differed between the two groups. In healthy volunteers, the left frontal and parietotemporal areas showed oscillatory changes. In the patient group, left frontal language areas were detected in 95.9% of cases, although activity in the posterior language areas was not as lateralized.

Finally, Kamada et al. (2006 and 2007) used a word categorization task to localize language functions intrahemispherically. Activation was found in the superior temporal, middle temporal, and supramarginal gyri of the dominant hemisphere. Moreover, Kamada et al. (2007), in a study of 177 patients, found that combined MEG and fMRI data yielded a 100% match with IAT results, including data on two patients who showed dissociation of expressive and receptive language areas. Grummich et al. (2006) used different language tasks with patients who had tumours to compare MEG and fMRI findings. Congruence was found between fMRI and MEG in 77% of patients for intrahemispheric language localization, results differed in 4% of cases, and in 19% of cases one modality showed activation but not the other. They concluded that more information about language centres is obtained by combining measurements and using multiple paradigms.

In summary, the different language comprehension tasks used to localize intrahemispheric sources of activity showed activation in the left temporal lobe in most cases, in both control and patient populations.
Table 2 MEG studies investigating intrahemispheric localization of language.

Language production

Source localization methods

Inverse solutions or source localization methods can be divided into two big groups: the equivalent current dipoles (ECD) and the distributed solutions. In most of these studies, the neuromagnetic fields elicited by the stimuli were recorded and the sources modelled as single ECD fitted at different successive time intervals (e.g. 1 ms, 4 ms). A current dipole consists of a point source, with a given position, orientation and dipolar moment (strength). The ECD is the best-fitting current dipole, in terms of maximum field variance. In some cases the estimated activity sources associated with the late components of the ERFs (200 ms after stimulus onset) were examined (Simos et al., 1998). Others limited ECD computation to latency periods during which a single pair of magnetic flux extremes dominated the left and/or right half of the head surface (e.g. Maestú et al., 2002). According to the article by Simos et al. (1998), the single ECD model was part of the standard analysis protocol in essentially all clinical MEG applications. A single ECD has been found sufficient to account for 90–95% of the variance in ERF data. Levelt et al. (1998) integrated the ECDs in a multidipole source model, derived by fitting dipoles to the entire spatiotemporal field pattern. They obtained source models which explained 80-90% of the data variance. However, such findings should be interpreted with caution due to the ill-posed nature of the inverse problem, given that the possible sources are far more than the number of sensors used to measure the source activities. Boundary effects, multiple dipolar activity, and cancellation effects can influence the brain's neuromagnetic fields and the resultant ECD modelling.

Other studies used distributed solutions such as the minimum norm estimate (MNE) and multi resolution FOCUSS (MR-FOCUSS). For example, Härle et al. (2002) used the MNE, an inverse method for reconstructing the primary current underlying extra-cranially recorded responses. Unlike ECD modelling, MNE requires no a priori knowledge of the possible source configuration or restriction of the MEG channels included in the model (Breier and Papanicolaou, 2008). McDonald et al. (2009) applied a spatiotemporal analysis to estimate the time courses of cortical activity using a distributed source solution.

Bowyer et al. (2004, 2005a, 2005b) used multi resolution FOCUSS (MR-FOCUSS), a current density imaging technique that detects focal concentrations of cortical activity. MR-FOCUSS enables a time sequence of whole brain images of focal and extended source structures to be constructed. They also used ECD source localization in their analysis and compared the two methods. Results showed that MR-FOCUSS analysis can provide the anatomical location of the multiple cortical areas involved in the language process. Moreover, because MR-FOCUSS produced reasonable localizations in a large number of patients, with similar temporal and spatial evolution in the several patients with whom it was not possible to fit dipoles even when using less rigid criteria, it would appear that MR-FOCUSS is more sensitive and useful than ECD. The authors argue that ECD works well for stationary, non-distributed sources such as early cortical latencies in evoked response data. However, for spontaneous transients such as language comprehension, the model would not be robust, in part, because multiple cortical sites originating from non-stationary distributed sources are active for only a short period. Because language processing involves numerous cortical areas that may be simultaneously active, current density imaging techniques such as MR-FOCUSS are well suited for mapping MEG data onto corresponding cortical structures. This approach provides a temporal display of all the concurrent activity involved during language processing.

Supplementary analyses

Conclusion

MEG directly measures neurophysiological processes with a high temporal resolution and therefore has the potential to localize neurophysiological processes within the whole brain. It has been useful in determining hemispheric language dominance in presurgical patients and mapping language function areas. MEG has been used to identify both frontal and temporal areas of activation and to identify language dominance in agreement with other methods, such as fMRI and the IAT. The reliability and validity of this technique have also been confirmed.

When drawing from the literature to develop a language protocol, certain factors need to be taken into account, especially if the protocol must be adapted for children. For example, tests should be relatively short because MEG requires immobility. The presentation mode can also influence results. Some argue that the auditory mode elicits asymmetric cerebral activation in favour of the left hemisphere, while others prefer visual presentation because visual stimuli activate areas located further from Broca's and Wernicke's areas. Of the studies reviewed here, many more used auditory than visual mode. Moreover, it is easier to use auditory stimuli with children who cannot read or who have reading disabilities. It is imperative that the tasks are accomplished by a paediatric population. In addition, the complexity of stimuli may influence results. For example, vowels are acoustically and linguistically simpler than words. Therefore, a word-related task would more likely evoke a greater portion of the linguistic neural pathways involved in lexical and semantic processing. It is also very important to note that because MEG has high temporal resolution, when long stimuli are used and analyzed (sentences), more variability will be found between participants due to inter-subject differences in processing. Consequently, averaged signals will be blurred and imprecise. Ideally, the analysis should be limited to a portion of the signal equal to or smaller than the word length. Moreover, in any language protocol, it is important to assess language comprehension and language production, especially if the findings are to be compared with IAT results. Based on the studies reviewed here, covertly produced responses allow investigating language production and yield activation in the areas of interest (Wernicke's and Broca's areas).

For the reviewed studies, different methods of analysis were used to determine the location of cortical sources involved in language processing and these locations were subsequently mapped using brain MRIs. These methods need to be taken into account when addressing the limitations of MEG, as they restrict the potential for interpretation. They may also contribute to differences in findings. The inverse solution is often used to estimate the source of language activation. However, this method presents drawbacks, and it allows only indirect estimates of the activity source based on MEG findings. Most of the studies reviewed here used ECD to model the data. Other analysis methods (MNE, MR-FOCUSS, SAM, etc.) were also used, and in all cases, activation in regions of interest was obtained. However, when different methods are compared for similar tasks, the findings are inconsistent. Across studies, the timing of lateralization and localization also varied. In most cases, late fields were analyzed (after 150 ms), but in some cases early fields (before 150 ms) yielded interesting findings. The paradigm used can influence these findings (Shtyrov and Pulvermüller, 2007; Gootjes et al., 1999). Overall, there is a clear need for standardized protocols and methods of analysis to enable comparisons of findings from different research centres.

A significant advantage of MEG is that it allows examining both hemispheres simultaneously, which is especially useful in epileptic populations, in which language lateralization is more variable. Similarly, in neurologically intact individuals, language often involves bilateral cortical networks. This was observed in the studies reviewed here, which showed bilateral activity in many cases, although left hemisphere activations were generally stronger. Furthermore, there is rarely a single source of activation during language comprehension, but rather multiple areas of activation. From the results of these studies, one might conclude that no task is purely linguistic: they all involve to a greater or lesser degree other cognitive operations such as attention or memory. Thus, the results of language studies also showed cortical activity that depended on other cognitive functions.

To conclude, MEG offers many important advantages: it is completely noninvasive, can be used with children, has excellent temporal resolution, and allows intrahemispheric localization of sources of activity. In short, it is an excellent presurgical assessment tool for localizing language functions. Nonetheless, MEG has some limitations. For example, it cannot be used with patients who have metal implants, very young children or non-cooperative patients, and it is relatively expensive. It has also been argued that it is difficult to assess language production with MEG. In the studies reviewed here it was possible to examine language production using MEG. However, it should be noted that the language production tasks did not systematically activate frontal regions, which should have been the case. This may be due to the fact that most tasks used covert production of answers. MEG is less sensitive than other techniques to detect deep and very small sources. Ultimately, it appears that using more than one technique could yield a more complete picture of activation profiles. It is important to include more than one task when assessing language functions in patient populations prior to surgery, and to include both language comprehension and production tasks, which have been shown to yield activation in the regions of interest. Thus, the best task to assess language comprehension in both adults and children appears to be a word recognition task. A verbal fluency task could be used to assess language production in children and a verb generation task in adults.

Acknowledgments

Financial support.

This study was supported by the Canada Research Chair Program (Maryse Lassonde), the Fonds de Recherche en Santé du Québec (FRSQ) (Maryse Lassonde, Renée Béland, Dang K. Nguyen and Mona Pirmoradi), and a scholarship awarded by the Canadian Institutes of Health Research (Mona Pirmoradi).

Disclosure.

None of the authors has any conflict of interest to disclose.

References

[Bowyer and Flemming, 2005a] Bowyer SM, Flemming T, Greenwald ML et al. Magnetoencephalographic localization of the basal temporal language area. Epilepsy Behav 2005a; 6: 229-34.

[Bowyer et al., 2005b] Bowyer SM, Moran JE, Weiland BJ. Language laterality determined by MEG mapping with MR-FOCUSS. Epilepsy Behav 2005b; 6: 235-41.

[Bowyer et al., 2004] Bowyer SM, Moran JE, Mason KM, et al. MEG localization of language-specific cortex utilizing MR-FOCUSS. Neurology 2004; 62: 2247-55.

[Breier and Papanicolaou, 2008] Breier JI, Papanicolaou AC. Spatiotemporal patterns of brain activation during an action naming task using magnetoencephalography. J Clin Neurophysiol 2008; 25: 7-12.

[Breier et al., 2005] Breier JI, Castillo EM, Simos PG, et al. Atypical language representation in patients with chronic seizure disorder and achievement deficits with magnetoencephalography. Epilepsia 2005; 46: 540-8.

[Breier et al., 2001] Breier JI, Simos PG, Wheless JW, et al. Language dominance in children as determined by magnetic source imaging and the intracarotid amobarbital procedure: A comparison. J Child Neurol 2001; 16: 124-30.

[Breier et al., 1999a] Breier JI, Simos PG, Zouridakis G, et al. Language dominance determined by magnetic source imaging: a comparison with the Wada procedure. Neurology 1999a; 53: 938-45.

[Breier et al., 1999b] Breier JI, Simos PG, Zouridakis G, Papanicolaou AC. Lateralization of cerebral activation in auditory verbal and non-verbal memory tasks using magnetoencephalography. Brain Topogr 1999b; 12: 89-97.

[Cornelissen et al., 2009] Cornelissen PS, Kringelbach ML, Ellis AW, Whitney C, Holliday IE, Hansen PC. Activation of the left inferior frontal gyrus in the first 200 ms of reading: evidence from magnetoencephalography (MEG). PLos One 2009; 4: e5359.

[Doss et al., 2009] Doss RC, Zhang W, Risse GL, Dickens DL. Lateralizing language with magnetic source imaging: Validation based on the Wada test. Epilepsia 2009: 1-7.

[English and Davis, 2010] English J, Davis B. Case report: Death associated with stroke following intracarotid amobarbital testing. Epilepsy Behav 2010; 17: 283-4.

[Fisher et al., 2008] Fisher AE, Furlong PL, Seri S, et al. Interhemispheric differences of spectral power in expressive language: a MEG study with clinical applications. Int J Psychophysiol 2008; 68: 111-22.

[Gates and Dunn, 1999] Gates JF, Dunn ME. Presurgical assessment and surgical treatment for epilepsy. Acta Neurol Belg 1999; 99: 281-94.

[Gootjes et al., 1999] Gootjes L, Raji T, Salmelin R, Hari R. Left-hemisphere dominance for processing of vowels: a whole-scalp neuromagnetic study. Neuroreport 1999; 10: 2987-91.

[Grummich et al., 2006] Grummich P, Nimsky C, Pauli E, Buchfelder M, Ganslandt O. Combining fMRI and MEG increases the reliability of presurgical language localization: A clinical study on the difference between and congruence of both modalities. Neuroimage 2006; 32: 1793-803.

[Härle et al., 2002] Härle M, Dobe C, Cohen R, Rockstroh B. Brain activity during syntactic and semantic processing—a magnetoencephalographic study. Brain Topogr 2002; 15: 3-11.

[Helmstaedter et al., 1997] Helmstaedter C, Kurthen M, Linke DB, Elger CE. Patterns of language dominance in focal left and right hemisphere epilepsies: relation to MRI findings, EEG, sex, and age at onset of epilepsy. Brain Cogn 1997; 33: 135-50.

[Hirata et al., 2009] Hirata M, Goto T, Barnes G. Language dominance and mapping based on neuromagnetic oscillatory changes: comparison with invasive procedures. J Neurosurg 2009: 1-11.

[Kamada et al., 2007] Kamada K, Sawamura Y, Takeuchi F. Expressive and receptive language areas determined by a non-invasive reliable method using functional magnetic resonance imaging and magnetoencephalography. Neurosurgery 2007; 60: 296-305.

[Kamada et al., 2006] Kamada K, Takeuchi F, Kuriki S, Todo T, Morita A, Sawamura Y. Dissociated expressive and receptive language functions on magnetoencephalography, functional magnetic resonance imaging, and amobarbital studies. J Neurosurg 2006; 104: 598-607.

[Killgore et al., 1999] Killgore WDS, Glosser G, Casasanto DJ, French JA, Alsop DC, Detre JA. Functional MRI and the Wada test provide complementary information for predicting post-operative seizure control. Seizure 1999; 8: 450-5.

[Kim and Chung, 2008] Kim JS, Chung CK. Language lateralization using MEG beta frequency desynchronization during auditory oddball stimulation with one-syllable words. Neuroimage 2008; 42: 1499-507.

[Kirveskari et al., 2006] Kirveskari E, Salmelin R, Hari R. Neuromagnetic responses to vowels vs. tones reveal hemispheric lateralization. Clin Neurophysiol 2006; 117: 643-8.

[Kober et al., 2001] Kober H, Möller M, Nimsky C, Vieth J, Fahlbusch R, Ganslandt O. New approach to localize speech relevant brain areas and hemispheric dominance using spatially filtered magnetoencephalography. Hum Brain Mapp 2001; 14: 236-50.

[Lee et al., 2006] Lee D, Sawrie SM, Simos PG, Killen J, Knowlton RC. Reliability of language mapping with magnetic source imaging in epilepsy surgery candidates. Epilepsy Behav 2006; 8: 742-9.

[Levelt et al., 1998] Levelt WJ, Praamstra P, Meyer AS, Helenius P, Salmelin R. An MEG study of picture naming. J Cogn Neurosci 1998; 10: 553-67.

[Maestú et al., 2002] Maestú F, Ortiz T, Fernandez A, et al. Spanish language mappint using MEG: a validation study. Neuroimage 2002; 13: 1579-86.

[Martin et al., 1993] Martin NA, Beatty J, Johnson RA, et al. Magnetoencephalographic localization of a language processing cortical area adjacent to cerebral arteriovenous malformation. Case report. J Neurosurg 1993; 79: 584-8.

[McDermott et al., 2005] McDermott KB, Watson JM, Ojemann JG. Presurgical language mapping. Curr Dir Psychol Sci 2005; 14: 291-5.

[McDonald et al., 2009] McDonald CR, Thesen T, Hagler Jr DJ, et al. Distributed source modeling of language with magnetoencephalography: Application to patients with intractable epilepsy. Epilepsia 2009; 50: 2256-66.

[Merrifield et al., 2007] Merrifield WS, Simos PG, Papanicolaou AC, Philpott LM, Sutherling WW. Hemispheric language dominance in magnetoencephalography: sensitivity, specificity, and data reduction techniques. Epilepsy Behav 2007; 10: 120-8.

[Mohamed et al., 2008] Mohamed IS, Cheyne D, Gaetz WC, et al. Spatiotemporal patterns of oscillatory brain activity during auditory word recognition in children: A synthetic aperture magnetometry study. Int J Psychophysiol 2008; 68: 141-8.

[Papanicolaou et al., 2006] Papanicolaou AC, Pazo-Alvarez P, Castillo EM, et al. Functional neuroimaging with MEG: normative language profiles. Neuroimage 2006; 33: 326-42.

[Papanicolaou et al., 2004] Papanicolaou AC, Simos PG, Castillo EM, et al. Magnetoencephalography: a noninvasive alternative to the Wada procedure. J Neurosurg 2004; 100: 867-76.

[Papanicolaou et al., 1999] Papanicolaou AC, Simos PG, Breier JI, et al. Magnetoencephalographic mapping of the language-specific cortex. J Neurosurg 1999; 90: 85-93.

[Pelletier et al., 2007] Pelletier I, Sauerwein HC, Lepore F, Saint-Amour D, Lassonde M. Non-invasive alternatives to the Wada test in the presurgical evaluation of language and memory functions in epilepsy patients. Epileptic Disord 2007; 9: 111-26.

[Pujol et al., 1999] Pujol J, Deus J, Losilla JM, Capdevilla A. Cerebral lateralization of language in normal left-handed people studied by functional MRI. Neurology 1999; 52: 1038-43.

[Salmelin et al., 1994] Salmelin R, Hari R, Lounasmaa OV, Sams M. Dynamics of brain activation during picture naming. Nature 1994; 368: 463-5.

[Shtyrov and Pulvermüller, 2007] Shtyrov Y, Pulvermüller F. Early MEG activation dynamics in the left temporal and inferior frontal cortex reflect semantic context integration. J Cogn Neurosci 2007; 19: 1633-42.

[Simos et al., 1999] Simos PG, Papanicolaou AC, Breier JI, et al. Localization of language-specific cortex by using magnetic source imaging and electrical stimulation mapping. J Neurosurg 1999; 91: 787-96.

[Simos et al., 1998] Simos PG, Breier JI, Zouridakis G, Papanicolaou AC. Assessment of functional cerebral laterality of language using magnetoencephalography. J Clin Neurophysiol 1998; 15: 364-72.

[Springer et al., 1999] Springer JA, Binder JR, Hammeke TA, et al. Language dominance in neurologically normal and epilepsy subjects : A functional MRI study. Brain 1999; 122: 2033-45.

[Sun et al., 2003] Sun J, Wu J, Li S, Wu Y, Liu L. Localization of the human language cortex by magnetic source imaging. Hum Brain Mapp 2003; 116: 1039-42.

[Szymanski et al., 2001] Szymanski MD, Perry DW, Gage NM, et al. Magnetic source imaging of late evoked field responses to vowels: toward an assessment of hemispheric dominance for language. J Neurosurg 2001; 94: 445-53.

[Szymanski et al., 1999] Szymanski MD, Rowley HA, Roberts TP. A hemispherically asymmetrical MEG response to vowels. Neuroreport 1999; 10: 2481-6.

[Wada and Rasmussen, 1960] Wada J, Rasmussen T. Intracarotid injection of sodium amytal for the lateralization of cerebral speech dominance. J Neurosurg 1960; 17: 266-82.

[Yamamoto et al., 2006] Yamamoto M, Ukai S, Shinosaki K. Spatially filtered magnetoencephalographic analysis of cortical oscillatory changes in basic brain rhythms during the Japanese ‘Shirotiri’ word generation task. Neuropsychobiology 2006; 53: 215-22.