INSIGHTS INTO BRAIN FUNCTION THROUGH MAGNETIC SOURCE IMAGING: RESEARCH AND CLINICAL APPLICATIONS

CONFERENCES
TOPIC: EPILEPSY

---

 
Abstract

The paper presents a brief outline of the rationale behind the use of non-invasive functional imaging and of the features that any imaging technique should display in order to make a substantial contribution to the search of the brain mechanisms responsible for cognitive functions. One such technique, Magnetic Source Imaging (MSI), that meets these specifications, is described in more detail. Advantages of MSI include the capacity to provide direct measures of regional neurophysiological activity, a millisecond-range temporal resolution, and the capacity to provide images of brain activity on an individual basis. We then describe applications of Magnetic Source Imaging to the study of brain mechanisms involved in various language functions such as oral comprehension and reading. Among these applications, the accuracy of MSI protocols in determining hemispheric dominance for language functions and in identifying the precise location and extent language-specific cortex (Wernickeís area) has been verified through comparison with standard invasive techniques (Wada procedure and electrocortical stimulation mapping) in over 35 consecutive cases. In another series of studies we combined data from MSI and direct cortical stimulation to determine the role of temporoparietal areas in phonological analysis of spoken language and in phonological decoding of print. Finally, we have used MSI to gain unique insights into the brain mechanisms that support reading in developmental reading disability. Results from over 21 children diagnosed with this disorder suggest that impaired reading is associated with aberrant functional connections between temporal and temporoparietal areas of the left hemisphere that are normally engaged in reading.
 

Resumen

El presente articulo es una breve aproximación a las posibilidades del uso de las técnicas no invasivas de imagen funcional y de las características que cualquier técnica de neuroimagen debería disponer en orden a realizar una contribución sustancial a la investigación de los mecanismos responsables de las funciones cognitivas. Una de esas técnicas es la Imagen de Fuente Magnética (del ingles Magnetic Source Imaging, MSI) que cumple esas especificaciones y se describe con mas detalle. Entre las ventajas de la MSI se incluye la capacidad de proporcionar medidas directas de la actividad neurofisiologica regional, con una resolución temporal del orden de los milisegundos. Así, nosotros describimos las aplicaciones de la MSI para el estudio de los mecanismos involucrados en varias funciones lingüísticas como la comprensión oral y la lectura. Entre estas aplicaciones destaca la capacidad de los protocolos MSI en la determinación de la dominancia hemisférica para el lenguaje y en identificar la precisa localización y extensión del córtex lingüístico especifico (área de Wernicke), verificando estos hallazgos mediante la comprobación con técnicas invasivas estándar (Wada y estimulación eléctrica cortical), en 35 casos consecutivos. En otra serie de estudios se combinan los datos de la MSI y la estimulación directa cortical para determinar el papel de las áreas temporo-parietales en el análisis fonológico del lenguaje hablado y en la decodificación fonológica de la escritura. Finalmente nosotros hemos usado la MSI para introducirnos en los mecanismos cerebrales que soportan la lectura en el desorden de la lectura en el desarrollo. Los resultados de 21 niños diagnosticados con este desorden sugieren que las dificultades de lectura están asociadas con conexiones funcionalmente aberrantes entre áreas temporales y temporo-parietales del hemisferio izquierdo que normalmente están asociadas a la lectura.
 
 

The advent of modern imaging techniques has stirred many expectations that scientists would finally be able to address definitively the core problem in the field of cognitive neuroscience, namely to describe the brain mechanisms responsible for cognitive functions.  In this article we will summarize results from a series of studies conducted at the Magnetic Source Imaging Laboratory in the University of Texas-Houston, in search of the brain mechanisms involved in various language functions. The rationale behind this research will become apparent following a brief outline of the basic principles and required features of modern functional imaging approaches.

Each functional imaging technique measures one aspect of brain activity in the form of electromagnetic signals. Three-dimentional (spatial) maps of this activity are then reconstructed using special algorithms. These maps are then used to derive information regarding the degree to which neurophysiological activity (i.e., neuronal signaling) increases regionaly above certain "baseline" levels as participants engage in experimental tasks. The latter are used as the context in which the cognitive functions under investigation are evoked.  Such changes in regional activity are considered as evidence that the "activated" brain region actually participates in the performance of the cognitive function(s) that the experimental task used requires.

A functional imaging technique can be useful in both clinical practice and as a research tool in cognitive neuroscience if:
(1) it has the capacity to provide images of the working brain of individual subjects;
(2) is capable of capturing the relevant aspects of brain activation, in the form of activity that reflects the actual engagement / participation of a particular area in the function under investigation; and,
(3) is capable of capturing both the spatial as well as the temporal features of activation.
In order to meet the first requirement, functional images must have sufficiently high fidelity and also display stability over repeated measurements. The second requirement can be met if activation profiles reflect, in a direct fashion, critical aspects of neuronal activity (i.e. neuronal signaling, as opposed to secondary delayed effects of neuronal activity such as blood flow/metabolism). Finally, the latter requirement can be met if the technique can relay changes in neuronal activity in real time. In other words it is crucial that regional profiles of activation are reconstructed during the temporal evolution of the function under investigation.  Most cognitive functions, such as word recognition, are performed within a fraction of a second after the presentation of a task-relevant stimulus.

MSI is unique among other functional imaging techniques for its ability to provide spatiotemporal brain activation profiles that reflect not only where activity occurs in the brain but also when this activity occurs in relation to the presentation of an external stimulus. In this way information can be drawn more directly regarding which areas participate in a particular cognitive/linguistic and also how these areas might interact with each other, in real time, to enable such complex functions. The latter type of information is very critical in order to determine functional connections between brain areas that show increased activation in the context of reading tasks in the effort to describe the brain mechanism serving word recognition.

In the studies summarized below, all MSI recordings were made with a multichannel neuromagnetometer (BTi, Magnes 2500) consisting of 148 magnetometers arranged so that they can cover the entire head.  The instrument is housed in a magnetically shielded chamber designed for reducing environmental magnetic noise that interferes with the recordings of physiological signals.  The entire recording session, during which the subjects or patients must remain immobile while lying on a bed with their head inside the helmet-like container of the magnetometers, does not exceed 10 minutes.  Thus, repeated measurements for the purpose of establishing the reliability of the results are feasible.

The principles underlying the capacity of MSI to identify brain areas that show event-related increases in local neuronal activity can be described briefly as follows: External stimuli are known to evoke neurophysiological activity as soon as they impinge upon sensory receptors.  One basic aspect of such activity is the intra-and extracellular flow of ions which is associated with electrical currents and magnetic flux.  The latter can be recorded from the head surface in the form of event-related potentials and fields (ERFs).  Application of a train of similar stimuli results in the repeated evocation of such activity which is recorded and averaged to improve signal quality.  The resulting average ERFs, in all cases, consist of early (30-150 ms. poststimulus) and late (150-700 ms poststimulus) components.  To identify the intracranial origin of the latter, the magnetic field distribution that had been recorded simultaneously over the entire head surface at successive points (4 ms apart) during the evolution of each component and which represented the amplitude of the component at each successive time point was analyzed.  The analysis consisted of the application of a mathematical model which considers the intracranial activity sources (sets of active cells) as equivalent to physical current dipoles (Papanicolaou, 1998; Sarvas, 1987) and is intended to provide estimates of the location and strength of these sources, the activity of which produced the recorded magnetic fields at that point in time.

The location estimates of each "dipolar" source are specified with reference to a Cartesian coordinate system, anchored on three fiducial points on the head (the nasion and the external meatus of each ear).  The same fiducial points were marked with vitamin pills, thus enabling superimposition of the precise location of each dipolar source on the subjectís (or patientís) MRI.
Thus, the dipolar sources that account for a particular ERF component projected onto the MRI identify the brain areas activated during that time interval in response to the stimulus.  The degree of activation of a particular area (or the total duration of its activation) following a stimulus, is estimated by the total number of successive dipoles that account for the ERF components.  The validity of this estimate was not based on any theoretical considerations, but it was empirically derived.  Namely, among all other possible indices of the degree of activation of an area (e.g. mean or median amperage of all dipolar sources), the number of sources was the one which resulted in the most consistent mapping results (Breier et al., 1999a; Papanicolaou et al., 1999; Simos et al., 1998a).
 

We have described in two studies the spatiotemporal activation profile associated with phonological decoding in neurologically intact adult volunteers without reading problems (Breier et al., 1998; in press-a). This profile features, initially, bilateral activation of occipital regions, followed by left basal temporal activation, which is in turn followed by activation in left posterior temporal and inferior parietal (i.e. temporoparietal) regions. A clear predominance of left over right hemisphere activation of basal temporal and temporoparietal areas was noted consistently across participants. Given the similarity in both degree and timing of activity in basal temporal areas between words and pseudowords (Breier et al., 1998, in press-a) this activity probably reflects the engagement of neurophysiological processes involved in pre-lexical analysis of print.

In order to determine the role of temporoparietal engagement in reading we combined data obtained non-invasively with MSI and results from electrical stimulation mapping performed either extra- or intraoperatively in patients undergoing left temporal lobectomies for epilepsy treatment. Three preliminary studies have been conducted thus far. In the first study with four patients, we examined the effects of direct electrical interference on distinct patches of the superior temporal gyrus on ability to read pseudowords (such as dofe) and orthographically irregular words (such as yacht and have). The rationale for using these types of stimuli was that pseudoword reading necessarily involves phonological decoding (in other words involves processes of assembled phonology) while correct pronunciation of irregular words requires (or addressed phonology). In all four patients we were able to identify a small region along the posterior portion of the superior temporal gyrus (typically in the order of 1-2 cm) stimulation of which led to complete (in three case) or partial (in one case) inability to read pseudowords. Correct reading of exception words was spared in all cases. Further testing with the same patients revealed that the cortical regions that were indispensable for deriving the assembled phonology of pseudowords were key components of the mechanism responsible for the phonological analysis of spoken words and sentences.  In a second experiment spatiotemporal brain activation profiles were obtained during reading of real words and pseudowords using Magnetic Source Imaging from ten neurologically intact volunteers.  Based on these profiles we obtained information regarding the manner in which cortex in the posterior part of the superior temporal gyrus interacts with other temporal and inferior parietal areas during reading. In the third experiment of this series, spatiotemporal activation profiles were obtained from patients prior to undergoing a left temporal lobectomy in the context of a printed and an auditory word recognition task. The results suggested that portions of the superior temporal gyrus involved in the phonological analysis of speech, as indicated by severe deficits (elicited by direct electrical stimulation) in the ability to process complex speech utterances, are routinely engaged (i.e. show activation) during silent word reading. However, engagement of these regions does not appear to be necessary for reading real words aloud, as suggested by the results of the first experiment. These results are consistent with the prediction of dual route models for reading aloud (e.g. Coltheart et al., 1993).
 

Non-invasive mapping of language-specific cortex.
Among the issues that MEG could eventually address, identification of the brain regions mediating language has always been the most urgently sought after, due to its obvious practical implications:  Advance knowledge of the language-specific zones can facilitate surgical planning and reduce morbidity associated with resection of eloquent cortex, especially in cases of epilepsy surgery.  Such knowledge is currently sought through invasive means such as the Wada procedure (Strauss & Wada, 1983; Loring et al., 1990) and direct cortical stimulation either intraoperatively or extraoperatively via implanted electrodes or subdural electrode grids (e.g., Ojemann et al., 1989; Lesser et al., 1986).  Successful mapping of language-specific cortex or, for that matter, cortex specialized for any cognitive function, can be defined as the verification of expectations regarding the functional specialization of the brain by the MEG-derived maps.  Specifically, in the case of receptive language, two aspects of functional specialization are beyond dispute:  First, that the overwhelming majority of people are left hemisphere dominant for language and second, that the receptive language mechanism involves the posterior temporal and parietal cortex (Wernickeís area). Consequently, MEG mapping could only be considered successful if it showed clearly greater activation of the left hemisphere in virtually all subjects, concentrated in Wernickeís area, during the linguistic but not during the non-linguistic tasks. In addition, it is essential that activation maps are stable over time in a given individual (test-retest reliability). The tasks we developed for language mapping (two versions of a continuous recognition memory task, one for spoken and the second for printed words) can be completed rapidly (each session lasts a maximum of 40 min) and performed easily by individuals with basic communication skills in a relatively rapid fashion.

We initially verified that the MSI protocols we had developed met these expectations in a series of four studies using neurologically intact volunteers (Breier et al., in press-a,b; Papanicolaou et al., 1999; Simos et al., 1998; Zouridakis et al., 1998). Based on the outcomes of these studies, the measure that proved to be the most valid and reliable index of language-specific activation is the number of activity sources computed over the late portion of the evoked magnetic response located in perisylvian areas traditionally associated with language function, i.e. the posterior portion of the superior and middle temporal gyri, the supramarginal and angular gyri, and the inferior frontal gyrus.

Having established appropriate testing MSI protocols we conducted two large-scale studies to examine the degree of concordance between the MSI and invasive procedures for determining hemispheric dominance and identifying language-specific cortex within the dominant hemisphere, namely the Wada procedure and electrocortical stimulation mapping, which are considered as the "gold-standards" in the field. In the first of these studies we obtained excellent concordance between MSI-based estimates of hemispheric dominance and Wada results in 26 consecutive patients (Breier et al., 1999). Using MSI and Wada indices three distinct groupings of language laterality scores were identified. As expected, the majority of patients belonged to a group that exhibited a strong left hemisphere-favoring asymmetry by displaying 20% or more activity sources in left over right hemisphere perisylvian regions. A second group exhibited a strongly right hemisphere favoring-asymmetry, while a third exhibited relatively bilaterally symmetrical data. The two methods agreed completely regarding the members of each of these groups.

Non-invasive lateralization of language function using a methodology such as MSI has a number of advantages, including elimination of health risk, potential for test-retest reliability studies, and ability to use a number of different tasks of extended duration. In addition, the problems inherent in the Wada procedure, including potential over- or under-anesthetization and anomalous distribution of anesthetic due to crossflow or atypical vascularization, are eliminated. Furthermore, while the Wada procedure provides only data regarding lateralization of language function, MSI is capable of providing data regarding precise intrahemispheric localization of areas involved in language function as well, and this data has been shown to be valid by comparison with intraoperative electrocortical stimulation as described below. Currently we are in the process of establishing the utility of a variety of tasks to provide reliable information regarding medial temporal lobe function (or dysfunction), an important step if the procedure is to be considered as a replacement for the Wada procedure in epilepsy surgery.

FIGURE 1. MSI - MRI co-registered scan from a representative patient who was judged to left-hemisphere dominant for language based on the Wada procedure. Note the clear preponderance in the number of MSI-derived cerebral activity sources in perisylvian areas of the left over the right hemisphere.

 
 

In a second study, we demonstrated the concordance between MSI and direct cortical stimulation for mapping receptive language cortex in a consecutive series of 13 patients (Papanicolaou et al., 1999; Simos et al., 1999a,b). Areas specialized for receptive language function were identified preoperatively as the region(s) that consistently displayed activation during both versions of our word recognition task (i.e., one involving spoken and the other printed words) or two repetitions of the auditory word recognition task. Importantly, the MSI technique was successful in identifying areas involved in receptive language function located outside of the traditional anatomical borders of  Wernickeís area.

FIGURE 2. Preoperative MRI scan from a patient with a left posterior inferior temporal cyst.  Clusters of MSI-derived activity sources obtained in the context of two repetitions of a word recognition task are shown as red or yellow circles. Crossed lines indicate sites of effective intraoperative electrical stimulation.

Such areas were found, as indicated by both MSI and intraoperative stimulation mapping, in the anterior portion of the superior temporal gyrus, the posterior-inferior portion of the middle temporal gyrus, basal temporal cortex, and the lateral temporo-occipital cortex. In all of these atypical cases, MSI-derived information was found to be extremely useful in surgical planning by: (a) helping to determine the optimal extent of the craniotomy, (b) helping to assess surgical risk and, (c) helping to tailor the location and extent of the cortical resection.  The localization accuracy of the procedure was apparently unaffected by the type and extent of brain pathology, or the presence of preoperative language and cognitive deficits.

Given the success in mapping temporal lobe language areas (Wernickeís region) with MSI, we are now assessing the validity and reliability of various tasks that could be used for the identification of frontal lobe areas critically involved in aspects of expressive language function.

Developmental Reading Disability
Developmental reading disability (dyslexia) affects a significant proportion of otherwise normal children. It is believed that the core deficit in dyslexia is in the operation of a decoding process that performs grapheme to phoneme conversions at a sublexical level, known as phonological assembly (Coltheart et al., 1993). Although phonological analysis is by definition required for reading aloud and for performing phonological similarity judgments on pseudowords (such as DOFE and HOAF), it is also believed to be critically involved in silent reading of real words by skilled readers (Seidenberg et al., 1994; Van Orden, 1987).

Functional imaging methods are uniquely suited to provide information regarding the brain mechanisms that support the cognitive processes, such as word recognition and phonological decoding which, in turn, make reading possible. Brain areas that have been shown to display reduced engagement in a variety of reading tasks in dyslexics are those located in the vicinity of the temporoparietal junction and typically include the posterior part of the superior and middle temporal gyri, the angular, and supramarginal gyri predominantly in the left hemisphere (Eden and Zeffiro, 1998; Rumsey et al., 1992; Shaywitz et al., 1998). Unfortunately, functional brain imaging methods that rely on measures of blood flow/metabolism are largely capable of providing profiles of relative activation that are meaningful primarily on a group basis. On the other hand, MSI can provide detailed information regarding the temporal course as well as the spatial extent of task-related regional activation. Moreover, as described in more detail above, the sensitivity of this method for identifying areas involved in language-specific processes has been established in a series of studies.

It should also be noted that most existing functional imaging studies on dyslexia have examined adults with developmental reading disability. However, there is evidence that dyslexics may develop alternative reading strategies in order to compensate for their poor phonological decoding skill (e.g., Olson et al., 1985). Little is known at present of how this compensation may affect brain activation profiles of adult dyslexics, creating the need for studies of younger individuals with reading problems.

We have recently completed two studies in which 21 children diagnosed with dyslexia and 18 age-matched controls were tested on tasks involving word recognition and phonological decoding (Simos et al., 1999c).

FIGURE 3. Three dimensional views of the surface of the brain of a dyslexic (top set of images) and a non-dyslexic child (lower set of images) during reading.  Note the nearly complete absence of activity in parts of the left half of the brain (blue square) in the dyslexic reader coupled with a wealth of activity in mirror areas of the opposite half of the brain. The opposite is true in the case of the normal, non-dyslexic child.

 

All dyslexic children displayed a severe impairment in phonological decoding skills as indicated by performance below the 30th percentile on the Word Attack subtest of the WJ-R Battery (Woodcock and Johnson, 1989). Activation maps were obtained using MSI during engagement in a visual and an auditory version of a word recognition task (Exp. 1), and during a pseudoword rhyme-matching task (Exp. 2). In both experiments all dyslexic children displayed spatiotemporal activation profiles that were markedly and consistently different from those obtained from age-matched non-dyslexic children. Brain activation profiles in children with dyslexia during word reading were characterized by: (1) markedly reduced left temporoparietal activation, (2) markedly increased activation in homologous areas of the right hemisphere starting at approximately the same latency (after the onset of the printed word stimuli) as activity in left temporoparietal regions in non-dyslexic children, (3) activity in ventral visual association cortices (basal temporal cortex) that was very similar in both degree and timing with the activity profile observed in the non-impaired readers. By virtue of the fact that MSI-based profiles of activation reveal both spatial and temporal aspects of brain function, the data presented here afford a closer look at the details of the mechanism that supports reading in children with developmental dyslexia. Whereas in non-impaired readers, this mechanism involves early engagement of left basal temporal regions followed by engagement of left posterior temporal and temporoparietal areas, in impaired readers this mechanism consists mainly of early left basal temporal activation followed by engagement of right posterior temporal and temporoparietal areas. On the other hand, activation of left temporoparietal areas during auditory word processing in impaired readers was virtually identical with activation observed in non-impaired readers. Taken together the data are consistent with the hypothesis that reading difficulties in developmental dyslexia are associated with an aberrant pattern of functional connections between brain areas normally involved in this process rather than a general dysfunction of a particular area.

Future research will examine systematically the relative contribution of different brain areas to reading and attempt to describe in more detail the pattern of regional interactions in relation to the emergence and refinement of reading skills. The study of individual differences in both brain function and behavioral skill is an essential part of this research.  Finally, after obtaining a clear picture of the course of normal development of the  brain mechanisms responsible for reading (and also the range of "normal" variability in the manifestation of these mechanisms) research can begin to address developmental problems in reading skills. This research is the first, yet necessary step, before we begin to understand how education (in general) as well as specific educational intervention strategies work on the brain to alter behavior.
 
 

Breier, J.I., Simos, P.G., Papanicolaou, A.C., Zouridakis, G., Wilmore, L.J., Wheeles, J.W., Constantinou, J.C., & Maggio, W.W. (1999a). Language dominance determined by magnetic source imaging: A comparison with the Wada Procedure. Neurology, 53.

Breier, J.I., Simos, P.G., Zouridakis, G., & Papanicolaou, A.C. (1998). Relative timing of cortical activation during a word recognition task. Journal of Clinical and Experimental Neuropsychology, 20, 782-790.

Breier, J.I., Simos, P.G., Zouridakis, G., & Papanicolaou, A.C. Temporal course of regional activation associated with phonological decoding: A MEG study. Journal of Clinical and Experimental Neuropsychology, 21, in press-a.

Breier, J.I., Simos, P.G., Zouridakis, G., & Papanicolaou, A.C. Lateralization of cerebral activation in verbal and non-verbal recognition memory tasks using magnetoencephalography. Brain Topography, in press-b.

Breier, J.I., Simos, P.G., Zouridakis, G., & Papanicolaou, A.C. Replicabibility of MEG-derived cortical activation maps in a visual word recognition task. Journal of Clinical Neurophysiology, in press-c.

Coltheart, M., Curtis, B., Arkins, P., & Haller, M. (1993). Models of reading aloud: Dual route and parallel distributed-processing approaches. Psychological Review, 100, 589-608.

Eden, G.F. & Zeffiro, T.A. (1998). Neural systems affected in developmental dyslexia revealed by functional neuroimaging.  Neuron, 21, 279-282.

Olson, R.K., Kliegl, R., Davidson , B.J., & Foltz, G. (1985). Individual and developmental differences in reading disability. In G.E. MacKinnon & T.G. Waller (Eds.), Reading research: advances in theory and practice (Vol. 4, pp. 1-64). Orlando, FL: Academic.

Papanicolaou, A.C. (1998). Fundamentals of Functional Brain Imaging. Lisse, The Netherlands: Swets & Zeitlinger.

Papanicolaou, A.C., Simos, P.G., Breier, J.I., Zouridakis, G., Wilmore, L.J., Wheeles, J.W., Constantinou, J.C., Gormley, W., & Maggio, W.W. (1999). Magnetoencephalographic mapping of the language-specific cortex. Journal of Neurosurgery, 90, 85-93.

Rumsey, J.M. et al. (1992). Failure to activate the left temporoparietal cortex in dyslexia: an oxygen 15 positron emission tomographic study Archives of Neurology, 49, 527-534.

Sarvas, J. Basic mathematical and electromagnetic concepts of the biomagnetic problem. Phys Med Biol 1987;32:11-22.

Seidenberg, M.S. & McClelland, J.L. (1989). A distributed, developmental model of word recognition and naming. Psychological Review, 96, 523-568.

Shaywitz, S. et al. (1998). Functional disruption in the organization of the brain for reading in dyslexia.  Proceedings of the National Academy of Sciences (USA), 95, 2636-2641.

Simos, P.G., Breier, J.I., Fletcher, J.M., & Papanicolaou, A.C. (1999c). Failure to engage brain areas involved in phonological decoding in dyslexic children: A magnetic source imaging study. Society for Neuroscience Abstracts, 25, S2170.

Simos, P.G., Breier, J.I., Maggio, W.W., Gormley, W., Zouridakis, G., Wilmore, L.J., Wheeles, J.W., & Papanicolaou, A.C. (1999a). Atypical temporal lobe language representation revealed by MEG and intraoperative stimulation mapping. Neuroreport, 10, 139-142.

Simos, P.G., Breier, J.I., Zouridakis, G., & Papanicolaou, A.C. (1998a). Assessment of cerebral dominance for language using magnetoencephalography. Journal of Clinical Neurophysiology, 15, 364-372.

Simos, P.G., Breier, J.I., Zouridakis, G., & Papanicolaou, A.C. (1998b). Identification of language-related brain activity using magnetoencephalography. Journal of Clinical and Experimental Neuropsychology, 20, 706-722.

Simos, P.G., Papanicolaou, A.C., Breier, J.I., Wilmore, L.J., Wheeles, J.W., Constantinou, J.C., Gormley, W., & Maggio, W.W. (1999b). Localization of language-specific cortex using MEG and intraoperative stimulation mapping. Journal of Neurosurgery, 91, 787-796.

Van Orden, G.C. (1987). A ROWS is a ROSE: spelling, sound and reading. Memory and Cognition, 15, 181-198.

Woodcock, R.W. & Johnson, M.B. (1989-1990). Woodcock-Johnson Psycho-Educational Battery - Revised. Allen, Texas: DLM Teaching Resources.

Zouridakis, G., Simos, P.G., Breier, J.I., & Papanicolaou, A.C. (1998). Functional hemispheric asymmetry assessment in a visual language task using MEG. Brain Topography, 11, 57-65.

---