9 Functional MRI in Pediatric Epilepsy Surgery
Functional magnetic resonance imaging (fMRI) has rapidly entered the armory of imaging methods available for the management of patients with focal epilepsy who may be surgery candidates. This is particularly true for presurgical investigations of cognitive and motor functions. The factors that have contributed to its success are chiefly its noninvasiveness, reproducibility, and wide availability. fMRI can reveal clinically useful information about the functional status of cortical tissue in relation to eloquent functions and, more recently, about origin and spread of epileptic activity.
In this chapter we will outline some of the basic principles of application of fMRI for complementing the neuropsychological evaluation of children and adolescents undergoing neurosurgical treatment. We will focus on two main issues that have received most attention so far: does fMRI provide a noninvasive means to reliably estimate hemispheric dominance of language and memory functions and localization of eloquent cortex? Finally, we will briefly mention recent developments in the mapping of blood flow changes associated with epileptic activity. The use of fMRI in mapping sensorimotor cortex before surgery has been reviewed previously1,2 and will not be covered in this chapter. Several recent reviews cover different aspects relevant to fMRI application to epilepsy surgery evaluation3–5 and its use in pediatric populations.1,6,7
Instrumentation and Methods Basic Principles of fMRI
fMRI is based on the dependence of T2-weighted signal on the oxygenation status of hemoglobin, and the derived signal is called blood oxygen dependent (BOLD) signal. The fMRI signal is delayed with respect to the onset of neuronal activity by approximately 6 seconds and shows a more temporally protracted time course, lasting approximately 20 seconds before returning to baseline. Neurophysiological studies in the monkey visual cortex have shown that neuronal excitation results in enhanced (positive) BOLD signal, whereas reduction in net neural excitation results in reduced (negative) BOLD signal.8 The spatial resolution is in the order of several mm (typical voxel size: 3 x 3 x 3 mm3), and its effective resolution will depend on the spatial filtering used during signal processing.
Experimental Designs
fMRI activation is detected using one of two experimental paradigms: block designs consisting of repeated activation– baseline state cycles and event-related designs, in which discrete events are analyzed separately, allowing for control of behavioral task performance during data analysis. The latter is used for studies of memory (e.g., comparing recalled versus forgotten/new items), whereas the former is commonly used with language and motor activation studies. Block designs have been used most consistently in children and adolescents, because their higher signal-to-noise ratio reduces the scanning time required to obtain robust fMRI activation.
Stimulation tasks are tailored to the cognitive domains under investigation, using visual or auditory stimulus presentation. Language tasks may include story comprehension, but the most commonly used task to assess hemispheric dominance for expressive language is silent word generation to letters (fluency) or words (verb or synonym generation). Memory tasks commonly use visually presented words, pictures, or human faces.
Response Monitoring
Earlier language fMRI studies have instructed patients to covertly generate verbal responses to avoid head movements caused by overt speech. Although this has generally worked well with most patients, it has the disadvantage of being prone to ambiguous results if activation patterns are atypical or in loss of activation in less cooperative patients. More recent studies have used experimental protocols that require patients to respond using button presses to indicate a choice between different stimulus categories or presence of certain target items.9 An alternative, yet relatively unexplored, method uses gaps in the image acquisition to allow for overt verbal responses.10
Assessment of Laterality of fMRI Activation
The term fMRI lateralization is commonly used to refer to the hemispheric asymmetry of fMRI activation in a given region of interest and may refer to extent of activation, level of activation, or both. Some epilepsy centers successfully rely on a qualitative judgment of laterality, performed by fMRI experienced neuroradiologists,9,11,12 whereas others use a combination of qualitative and statistical measures13,14 as converging evidence.
Applications for Presurgical Evaluation
The majority of published studies have been conducted on adults, and only a small number of pediatric studies have been reported.14–18 We therefore review here the evidence from both adult and childhood epilepsy studies combined while attempting to point out issues that may have specific significance for pediatric practice.
Preoperative Assessment of Language Lateralization
Language fMRI tasks are used in candidates for epilepsy surgery with a view to improve the prognosis for postsurgical speech and language deficits. The main questions are whether surgery is planned in the language-dominant hemisphere and whether language cortex is located near the planned resection. The increased frequency of atypical language lateralization (i.e., right-sided or bilateral) in patients with focal epilepsy or lesions of the left hemisphere has been known for more than 30 years.19 These early studies, using the intracarotid amobarbital test (IAT, or Wada test), also suggested that mainly lesions within classic language cortex (Broca’s and Wernicke’s regions) are responsible for inducing a shift of language to the right hemisphere. More recent investigations, with the benefit of modern neuroimaging, have, nevertheless, shown that a considerable proportion of patients with early acquired or developmental left-sided perisylvian lesions showed evidence of intrahemispheric reorganization of language (i.e., retain typical left-sided lateralization, see example in Fig. 9.1A ), often near the lesion.17,18,20 Furthermore, patients with epilepsy arising from pathology in regions remote from classic language cortex, especially in the mesial temporal cortex, often show atypical, often bilateral, language representation17,21,22 (see example in Fig. 9.1B ). A recent study in a large cohort of patients with a left hemisphere epileptogenic focus (including children) identified the following factors associated with atypical language lateralization12,23: left-handedness, onset of epilepsy before age 6, and MRI type. Regarding the latter pathology factor, it is notable that patients with stroke showed a very high rate of reorganization and that approximately 35% of patients with a normal MRI had atypical language. This finding points to the possibility that epileptic activity drives functional reorganization.24 In summary, localization-related epilepsies are associated with widespread changes in the structural and functional organization of the brain.
Assessment of Language Lateralization: fMRI and IAT
Studies on Adults
Although the IAT remains the gold standard for determining hemispheric dominance for speech and language, many epilepsy centers now routinely use fMRI.5,25 Comparative studies using both methods11,12,26 in adults have found agreement in approximately 80 to 90% of cases. Notwithstanding the obvious difference between an observation-based method (fMRI) and an inhibition procedure (IAT),2 multiple other factors are likely to cause disparity,9,11,27 including differences between fMRI and IAT tasks12,28 and variability in the regions of interest chosen for fMRI lateralization analysis.27,29 Finally, in contrast to the IAT, fMRI will also detect activation in regions that are not essential for the performance of a task (redundant activation). If these regions are localized in the nondominant hemisphere, this is likely to render fMRI findings more bilateral (see Woermann et al12 for examples). No method is currently available that could distinguish essential from redundant activation foci on fMRI. The contralesional activation detected on fMRI may also indicate the potential for postsurgical reorganization of function. Indeed, two studies that have used both fMRI and IAT have found fMRI to be a better predictor of postoperative cognitive outcome. The study by Sabsevitz et al30 in 24 patients who underwent left anterior lobe lobectomy (L-ATL) showed the fMRI lateralization index to be 100% sensitive and 73% predictive of significant visual naming decline. A recent study by Binder et al31 corroborates this finding for verbal memory outcome after L-ATL and will be discussed in more detail later.
Although the direct comparison of fMRI with IAT was an obvious first step in validating this new method, the IAT itself is not free of potential problems,32 such as lack of standardization and reproducibility, agitation and obtundation in some patients, potential cross-flow between the hemispheres, and many others (see Chapter 12). Indeed, there are reports of erroneous lateralization using the IAT, as evidenced by electrocortical stimulation33 or postsurgical dysphasia34 in which functional neuroimaging had indicated the correct lateralization. Therefore, only further postoperative outcome studies of the kind reported by Sabsevitz et al30 and Binder et al31 can demonstrate the true predictive value of fMRI-derived language lateralization.
In summary, a review of all available studies35 concluded that fMRI increases importantly the probability of correctly predicting language dominance in multiple subgroups of surgery patients with and without epilepsy.
Pediatric Studies
Only a few studies have specifically investigated the role of fMRI in pediatric epilepsy surgery candidates,14–16,18 commonly comparing fMRI with a mixture of invasive investigations (IAT, electrocortical stimulation [ECS]) or clinical observations. These studies have confirmed the feasibility and accuracy of fMRI in estimating language dominance in children, with the proviso that in some bilateral fMRI cases, only unilateral corroborating evidence was available.14,18 A study involving a small series of children who also underwent ECS and IAT reported bilateral fMRI activation more often than suggested by IAT36; however, details of the procedure were not given. Longitudinal fMRI studies are particularly useful for investigating children with epilepsy caused by extensive left hemispheric injury or progressive neurodegenerative conditions, such as Rasmussen encephalitis ( Fig. 9.1C ), in revealing the gradual process of language reorganisation,37 which can be used to optimize the timing of surgery and perhaps also for predicting the level of language proficiency after surgery.10