SEARCH

SEARCH BY CITATION

Keywords:

  • Functional neuroimaging;
  • Functional magnetic resonance imaging;
  • Epilepsy;
  • Lateralization;
  • Language;
  • Wada

Summary

  1. Top of page
  2. Summary
  3. Methods
  4. Results
  5. Discussion
  6. Conclusions
  7. Acknowledgments
  8. References

Purpose:  This work examines the efficacy of functional magnetic resonance imaging (fMRI) for language lateralization using a comprehensive three-task language-mapping approach. Two localization methods and four different metrics for quantifying activation within hemisphere are compared and validated with Wada testing. Sources of discordance between fMRI and Wada lateralization are discussed with respect to specific patient examples.

Methods:  fMRI language mapping was performed in patients with epilepsy (N = 40) using reading sentence comprehension, auditory sentence comprehension, and a verbal fluency task. This was compared with the Wada procedure using both whole-brain and midline exclusion–based analyses. Different laterality scores were examined as a function of statistical threshold to investigate the sensitivity to threshold effects.

Results:  For the lateralized patients categorized by Wada, fMRI laterality indices (LIs) were concordant with the Wada procedure results in 83.87% patients for the reading task, 83.33% patients for the auditory task, 76.92% patients for the verbal fluency task, and in 91.3% patients for the conjunction analysis. The patients categorized as bilateral via the Wada procedure showed some hemispheric dominance in fMRI, and discrepancies between the Wada test findings and the functional laterality scores arose for a range of reasons.

Discussion:  Discordance was dependent upon whether whole-brain or midline exclusion method–based lateralization was calculated, and in the former case the inclusion of the occipital and other midline regions often negatively influenced the lateralization scores. Overall fMRI was in agreement with the Wada test in 91.3% of patients, suggesting its utility for clinical use with the proper consideration given to the confounds discussed in this work.

In the majority of the adult healthy human population, the left hemisphere is dominant for language function. However, in a small percentage of the population, the right hemisphere is dominant for language, and sometimes there is bilateral dominance (Pujol et al., 1999; Springer et al., 1999). If surgical intervention is planned, both dominant and nondominant activity associated with language processing should be evaluated.

In patients with neurologic disorders such as epilepsy, it is essential to have an understanding of how organization may be changed in patient populations. In presurgical planning it is crucial not only to identify the role of specific brain areas but also to determine hemispheric dominance for language to avoid postoperative deficits. Identifying the network of the different brain regions involved in a particular cognitive task assists the neurosurgeon in surgical decision-making. This is essential, as it is vital to spare these activated brain regions in order to preserve function following surgery.

The Wada procedure (also known as the Intracarotid Amytal Test) (Wada & Rasmussen, 1960) is a standard clinical test to determine hemispheric dominance typical for both language and memory processing. This test determines the lateralization of the cerebral hemispheres by suppressing cortical activity in one of the hemispheres while testing the patient’s ability to perform certain language or memory-related tasks with the other hemisphere, and then repeating the same procedure suppressing activity in the opposite hemisphere. The Wada test findings represent a categorical classification of the left, right or bilateral hemisphere(s) being dominant for a particular cognitive function. This procedure is invasive, entails some risk, and often the patient is not comfortable during the period of the test and so it is usually not repeated. Although currently a gold standard for both memory and language assessments of laterality, the risks and the variable patient response to the procedure has to some extent limited the interpretation of the results (Loring et al., 1990; Snyder et al., 1990; Hart et al., 1991; Jeffery et al., 1991;Meador & Loring, 1999). Furthermore, although the Wada procedure provides significant information about hemispheric dominance, it provides no information on the localization of the brain regions activated in response to a specific cognitive task, which prior to surgery could be of great significance. Many epileptic patients often also undergo electrocortical stimulation mapping (ESM), which is another invasive means of obtaining information on functional localization. Functional magnetic resonance imaging (fMRI) for language mapping has the potential to provide a noninvasive alternative, which may yield more information on spatial localization. Functional imaging is now employed in many centers as a clinical tool to aid in the neurosurgical treatment of focal epilepsy and other conditions (Desmond et al., 1995; Binder et al., 1996; Bookheimer et al., 1997; Springer et al., 1999; Carpentier et al., 2001; Swanson et al., 2007). This work examines the efficacy of the approach through comparison with the Wada test and examines in detail cases where there is disagreement between the Wada results and the results from one or more fMRI paradigms. Sensory motor functions, language processing, and memory testing are some of the functions that form an important part of everyday life routines and, therefore, these cognitive tasks are usually tested in functional imaging.

fMRI detects the BOLD (blood oxygenation level dependent) signal as an indicator of neuronal activity associated with the performance of a specific task (Ogawa et al., 1990). The functional activation map illustrates a network of brain regions that are activated in response to a particular cognitive task. Like the Wada test, fMRI provides information not only about hemispheric lateralization but also about localization of activated brain regions related to a specific task. fMRI is noninvasive and can be easily repeated if necessary.

The aims of this study were (1) to identify cortex activated by language tasks on fMRI in order to determine hemispheric dominance. The different tasks differentially activate different nodes within the language network and thus provide a comprehensive picture of language processing; (2) to validate the language lateralization scores from functional imaging results with the Wada test findings and measure the concordance between the methods; and (3) to investigate the sensitivity of different lateralization measures to threshold effects and specific tasks in determining the hemispheric dominance using fMRI, and to explore the reasons for occasional disagreement between the Wada test and fMRI lateralization of language.

Methods

  1. Top of page
  2. Summary
  3. Methods
  4. Results
  5. Discussion
  6. Conclusions
  7. Acknowledgments
  8. References

Subjects

Patients (N = 40, 30 right-handed and 10 left-handed, 15 male, Table 1) with intractable epilepsy who were candidates for surgery formed the subject population. All of the patients underwent the Wada procedure, neuropsychological testing, and fMRI for language mapping. Approval for this study was obtained from the Yale Medical Human Investigation Committee, and all the patients signed an informed consent form.

Table 1.   Patient demographics, epilepsy location, and neuropsychological assessment
Patient no.SexAge (years)HandednessEpilepsy locationVerbal IQPerformance IQFull scale IQAge at onset of seizuresDuration of epilepsy (years)
  1. M, male; F, female; RH, right-handed; LH, left-handed.

  2. “Epilepsy location” is the clinical consensus for the patient that includes any visible pathology, video-EEG, interictal EEG, and potentially data from PET and MRS.

  3. EEG, electroencephalography; LH, left-handed; IQ, intelligence quotient; PET, positron emission tomography; MRS, magnetic resonance spectroscopy.

Group A
 1M38RHLeft temporal lobe114991081424
 2F21RHRight anterior medial temporal lobe, Hippocampal sclerosis8368761110
 3F50RHRight anterior medial temporal lobe, Hippocampus9995982525
 4F51RHRight anterior medial temporal lobe, Hippocampal sclerosis104104104744
 5M31RHRight occipital lobe747774724
 6F16LHLeft inferior occipitotemporal gyrus837075115
 7F40RHLeft posterior hippocampus and cingulate area829285Febrile seizures at 3 years, recurrent seizures at 9 years
 8F44RHRight anterior medial temporal lobe, Hippocampal sclerosis807978Febrile seizure age 6 months; Recurrence of seizures age 8 years
 9M43RHRight anterior medial temporal lobe, Hippocampus8192852023
10M46RHLeft temporal lobe7477742719
11F18RHLeft temporal lobe9891951818
12F19RHLeft, right, or both occipital lobes and right inferior posterior temporal cortex109103103415
13F51LHRight anterior medial temporal lobe, Hippocampal sclerosis9797975 months50.7
14F38RHSuperior left anterior medial frontal lobe79871315
15F18RHRight frontal lobe107102105144
16F24LHLeft temporal lobe898587Infancy, hemangioma at birth
17M25RHRight posterior temporal and parietal lobe8470761213
18M27RHLeft temporal lobe114951061017
19M39RHLeft temporal lobe1191191212712
20F47LHLeft anterior medial temporal lobe, Hippocampal sclerosis837981Febrile seizure age 5 months; Recurrence of seizures age 16 years
21F31LHRight anterior medial temporal lobe, Hippocampal sclerosis7891831021
22F21RHLeft occipital lobe130114125129
23M35RHLeft frontal lobe971171071223
24F23RHRight supplementary motor area9010295Infantile spasms at 6 months; partial complex seizures at 11 years
25F43RHLeft anterior medial temporal lobe, Hippocampal sclerosis798580736
26M25RHLeft anterior medial temporal lobe696966817
27M50RHLeft medial temporal lobe829888Childhood seizures with remission until age 20 years
28M33RHLeft temporal lobe9386981617
29M34RHFrontal lobe919090Diagnosed with epilepsy at age 10–11 years; had previous history of “zoning out” since 4 years, possible undiagnosed seizures
30F49RHSeizures971029935 years (after meningitis)14
31F16RHRight temporal lobe79878197
32M15LHLeft frontal lobe796670105
33F29LHLeft temporal lobe7575737 years old; tumor resection at 15 months old22
Group B
34F57RHRight frontal lobe10410710652 5
35M14RHRight temporal lesion that involves superior temporal gyrus101728613 1
36F12LHLeft anterior medial temporal lobe6982736 months  11.6
37M51RHLeft superior temporal gyrus9176831734
38F60RHLeft anterior medial temporal lobe919291Infancy–18 months  58.6
39F45LHLeft anterior medial temporal lobe10410710639 6
40F23LHLeft occipital lobe959595320

Functional imaging paradigms

Language is not a single process, but rather involves a number of specialized systems for reading (visual), listening (auditory), and generation of speech (word production). Therefore, tasks involving understanding of language (reading comprehension and auditory comprehension) (Constable et al., 2004) as well as production of language (verbal fluency task) were designed. In the sentence comprehension tasks, the subjects had to either read or listen to a series of sentences and were instructed to judge if the sentences were semantically and syntactically correct. They responded at the end of each sentence via a button press—“yes” if the sentences were grammatically correct and made sense; otherwise they pushed the “no” button. The nonlinguistic control condition for the reading task consisted of a pair of lines with five oblique lines on each row inline image (with same or different orientation), one immediately above the other, and the subjects had to judge whether the two rows of lines were identical or not. For the auditory task, pairs of high and low pitch tones formed the nonlinguistic control condition, and the subjects made a pitch judgment as to whether the tones in a given pair were the same or different. The verbal fluency task required word generation in response to a visual cue. There were two activation conditions for word generation. In one, subjects were visually presented with a random single letter and they had to generate words beginning with that letter. In the other activation condition, subjects were given a category word (e.g., rivers) and they had to generate as many names of items they could think of in that category. For either cueing condition subjects performed a button press (yes or no) depending on whether they could think of three or more words for the cue. The control condition for this task was the line task presented in a manner identical to that used in the sentence comprehension task. The subjects were asked to perform the verbal fluency task silently in the scanner to prevent head motion and movement artifacts.

For each language task, a single run consisted of alternating blocks of baseline and activation conditions. For the reading task, four blocks of sentences (activation) alternated with five blocks of control stimuli, each starting and ending with the control condition. The total task duration was 249.2 s, with 37.3 s of activation block and 20 s of control block. Each sentence was presented for 6.3 s, and the control condition for 2.5 s. The sentences in the auditory task were identical to those used in the reading task. The auditory task duration was 293 s, with 33 s of activation block and 31.5 s of control block. For the verbal fluency language task, four blocks of sentences (activation) alternated with five blocks of control stimuli (control), each starting and ending with the control condition with 208.5 s task duration and with 24 s activation block and 22.5 s control block. For the sentence comprehension task, and the sentences were divided into three parts; each part was presented one at a time at the center of the screen in order to minimize eye movements and prevent the subject from going back to examine the sentence structure. Each task was scanned three times, giving a total of nine runs. All the subjects underwent a practice session prior to scanning.

Magnetic resonance imaging

Imaging was performed on a 1.5T GE Signa LX (GE Medical systems, Waukesha, WI, U.S.A.) and a 1.5T Siemens Sonata system (Siemens, Erlangen, Germany). There was no main effect of magnet in these data, so the patient data from both the MR scanners were combined. Subjects were placed in the supine position in the MR scanner and their heads were placed within a standard head coil. To minimize head movement, foam was placed on either side of the head and tape was placed across each subject’s forehead and attached to the head holder within the head coil. Output from a Macintosh computer running Psyscope program (Cohen et al., 1993) was directed toward a projector that was placed at the feet of the subject so that the subject could view the stimuli by way of back projection through a mirror placed on the head coil. A fiberoptic button box was used by the subject to respond to the stimuli in the scanner. For the auditory comprehension task a set of MR-compatible headphones was used to present the stimuli. To identify the anterior commissure (ac) and posterior commissure (pc), a structural sagittal localizer was scanned (TE = 11 ms, TR = 500 ms, 256 × 192 matrix, field of view (FOV) = 20 cm, slice thickness 6 mm, 0 mm gap, 2 averages). T1-weighted, spin-echo, axial-oblique slices, parallel to and through the ac–pc line (5th slice centered on the ac–pc line) were then acquired (TE = 11 ms, TR = 500 ms, 256 × 192 matrix, FOV = 20 cm, slice thickness 6 mm, 0 mm gap, 16 slices on GE scanner, 18 slices on Siemens scanner) on which the functional images were later overlaid. Functional imaging was performed using gradient echo planar imaging (EPI) sequence (FA = 80, TE = 50, TR = 1800 ms for the reading and auditory task, TR = 1,500 ms for the verbal fluency task, 64 × 64 matrix, 62.5 kHz bandwidth) with the same FOV, slice thickness, and slice location as the T1-weighted anatomic scan. Four warm-up pulses were used to ensure steady state magnetization. A total of 134 images/slice per run for the reading task, 160 images/slice per run for the auditory task, and 138 images/slice per run for the verbal fluency task were obtained. Each of the language tasks was repeated three times (with randomized stimuli presented in every run), giving a total of nine runs in order to maximize the statistical power.

Data processing and analysis

The functional data were corrected for motion using SPM 99 (Friston et al., 1995) and then coregistered to the subject’s two-dimensional (2D) T1-weighted anatomic scans. For each individual language task, using in-house developed software, the runs were averaged and then a Student’s t-test was applied on this averaged run yielding t-maps (Constable et al., 2004). The resultant t-map for each separate condition was overlaid on the subject’s T1-weighted anatomic image to generate brain-activation maps displaying functional activations at a specific threshold. Because the average activation across the entire block was used in the t-test, no correction was made for slice-timing effects. No normalization was performed on the data. A spatial median filter [5 × 5] was applied to the data to remove spurious activations likely due to noise.

A combined task analysis incorporating a logical AND operation, analogous to the conjunction analysis (Bookheimer et al., 1997; Price et al., 1997; Carpentier et al., 2001; Ramsey et al., 2001; Rutten et al., 2002), was also performed on the individual language tasks. For a region to be activated in this conjunction method at a specific threshold, this region must be above that threshold in each task. Therefore, this analysis highlights regions of activation common in all the three language tasks rather than one single task only. In other words, it is independent of a specific task and presentation modality and hence represents a more basic language-processing network.

Lateralization index

To compare the functional activation results with the Wada results, a lateralization index (LI) was calculated to provide a quantitative measure of left (100) versus right (−100) lateralization. This index is defined as LI = {[(L – R)/ (L + R)] × 100}, where L generally indicates the number of voxels activated in the left hemisphere and R in the right hemisphere. Different quantitative measures of activation can be used in this metric. For example, voxel counts, sum of t-values, percentage signal change, product of sum of t-values * voxel count can all be used to determine the LI with the preceding expression. The different approaches to quantifying activation produce different laterality scores, and this was investigated across a range of statistical thresholds. Thresholds have an impact on the score, as a high threshold may result in only a few voxels activated implying that the laterality is driven by only these few voxels.

The purpose of comparing these measures was to determine the best measure to be used to calculate hemispheric lateralization. Figure 1 shows the results for these different metrics and indicates that the functional lateralization score calculated using the product of voxel count * sum of t-values was found to provide the maximum lateralization. This approach was, therefore, used to determine the functional language lateralization hemispheric dominance as LI = {[(L – R)/(L + R)] × 100}, where L indicated the product of voxel count * sum of t-value activated in the left hemisphere and R in the right hemisphere, respectively. The lateralization score is also threshold-dependent, as shown in Fig. 1. Threshold of t = 2 was chosen because it is in the mid-range providing the most stable laterality measure. t-Test was computed for unequal sample sizes and unequal variance as inline image, where inline image, and where inline image are the sample means for the activation and control data, and inline imageand inline image are the sample variances, respectively. Too high a threshold yields only a few activated voxels and unstable laterality calculations, and too low a threshold activates everything, leading to no lateralization. Threshold of t = 2 was used to evaluate the functional MR LIs and then compare these indices with the results of the Wada test. The MR language laterality range varied from +100 (maximal left dominance) to −100 (maximal right dominance). To classify these functional lateralization scores in a manner similar to that in which the Wada findings are usually categorically classified—as left, right, or bilateral dominance—cut-off values were determined from the graphs such that an LI score of greater than +10 was categorized as left hemisphere dominant, less than −10 as right hemisphere dominant, and any LI score in the range +10 to −10 was classified as bilateral. This categorical classification is important when making comparisons with Wada testing.

image

Figure 1.   Dependence of functional laterality on statistical threshold for four different quantitative measures of activation.

Download figure to PowerPoint

Wada procedure

At our institution, the ntracarotid Amytal procedure (“Wada test”) consists of both language and memory assessments. Both hemispheres are studied on the same day, following cerebral angiography. Typically, the side of suspected seizure onset is injected first, followed by the contralateral hemisphere approximately 30 min after the conclusion of the first study. The standard beginning dose is 125 mg, hand injected into the internal carotid artery over approximately 5–8 s, filling the middle cerebral artery distribution. Following the injection, contralateral grip strength is monitored, and language testing and memory testing are conducted. Language assessment consists of evaluating spontaneous speech quality, comprehension of simple commands, object and body-part naming, repetition of simple and complex phrases, and comprehension of complex ideational material. Language lateralization is based on motor speech (e.g., ability to phonate and produce intelligible speech) as well as more complex language assessment. Criteria for establishing hemispheric dominance are based on positive production of speech errors in the period following the injection and prior to complete recovery. We did not encounter instances of totally preserved language functions with anesthesia of both sides.

Bilateral language is determined when the patient demonstrates one or more of the following:

  • • 
    Cessation of speech with preserved consciousness following both right and left hemisphere injections.
  • • 
    Impaired object naming following both right and left hemisphere injections.
  • • 
    Impaired comprehension following both right and left hemisphere injections.

Results

  1. Top of page
  2. Summary
  3. Methods
  4. Results
  5. Discussion
  6. Conclusions
  7. Acknowledgments
  8. References

Functional language LI and the Wada test

From the total epileptic patient population in the current study, 72.5% were classified as left hemispheric dominant, 10% as right hemispheric dominant, and 17.5% as mixed hemispheric dominant (both left and right hemispheres dominant) by the Wada procedure. This sample was divided into two groups: a unilateral group (group A: either left or right) and a bilateral group (group B) as determined by the Wada test. A comparison was then made between the Wada test results and the functional language LIs to examine the concordance between the two methods.

In group A, the fMRI LIs were concordant with the Wada procedure results in 26 of 31 patients (83.87%) for the reading comprehension task, 25 of 30 patients (83.33%) for the auditory comprehension task, 20 of 26 patients (76.92%) for the verbal fluency task, and in 21 of 23 patients (91.3%) for the conjunction analysis (Table 2). The conjunction analysis showed the best overall concordance with the Wada test (91.3%) as compared to the individual language tasks (Table 2). Group B in which the patients were categorized as having bilateral dominance as determined by Wada showed discrepancies between the Wada test findings and the functional laterality scores for various reasons (Tables 2 and 4). For this group too, the conjunction analysis of the individual language tasks illustrated the best concordance with Wada (Table 2).

Table 2.   Summary of the functional lateralization indices and Wada test findings of patients
Patient no.HandednessWadaLI of visual taskLI of auditory taskLI of word generation taskLI of logical AND of three language tasks
  1. LI, lateralization index; LH, left-handed; RH, right-handed; LHD, left hemisphere dominant; RHD, right hemisphere dominant; B, bilateral dominance; L, left; R, right.

  2. –, Functional language task not performed on subject.

Group A
 1RHLHDL (98)L (92)
 2RHLHDL (97)L (86)
 3RHLHDL (23)L (16)
 4RHLHDL (53)L (50)
 5RHLHDL (26)L (94)L (37)L (97)
 6LHLHDL (88)L (46)L (24)L (98)
 7RHLHDL (94)L (70)L (21)L (99)
 8RHLHDL (61)L (86)L (66)L (95)
 9RHLHDL (99)L (53)L (83)L (100)
10RHLHDL (44)L (72)L (21)L (99)
11RHLHDL (47)L (43)L (67)L (93)
12RHLHDL (96)L (37)L (64)L (99)
13LHLHDL (56)L (20)
14RHLHDL (21)
15RHLHDL (98)L (70)L (92)L (100)
16LHLHDR (−87)L (95)
17RHLHDB (−3)R (−43)B (−6)B (−9)
18RHLHDL (45)R (−13)B (9)L (17)
19RHLHDR (−38)L (15)B (−10)R (−11)
20LHRHDB (−5)
21LHRHDR (−35)R (−16)R (−13)B (−0)
22RHLHDL (62)L (55)L (74)L (100)
23RHLHDL (70)L (60)L (74)L (99)
24RHLHDL (83)L (11)L (35)L (98)
25RHLHDL (100)L (48)L (64)L (100)
26RHLHDL (45)L (19)L (60)L (39)
27RHLHDL (75)R (−40)R (−84)L (98)
28RHLHDL (40)L (22)B (−2)L (26)
29RHLHDR (−67)L (23)R (−14)L (35)
30RHLHDL (72)B (−3)L (50)L (99)
31RHLHDR (−73)L (26)
32LHRHDR (−25)R (−52)
33LHRHDR (−32)R (−38)L (23)R (−93)
Group B
34LHBL (91)L (63)
35RHBR (−12)L (38)
36LHBR (−79)R (−58)R (−77)R (−99)
37RHBR (−73)R (−35)R (−54)R (−95)
38RHBL (94)B (1)L (87)L (99)
39LHBL (19)L (77)B (6)L (95)
40LHBR (−84)L (78)L (92)L (86)
Table 4.   Group B—Summary of patients whose lateralization indices (LIs) and Wada findings disagree
Patient no.HandednessWadaLI scores from fMRILI of visual taskLI of auditory taskLI of word generation taskLI of logical AND of three language tasksPossible explanations for disagreement between functional LI and Wada findings
  1. LI, lateralization index; ✓, functional lateralization indices and Wada findings agree; x, functional lateralization indices and Wada findings disagree, –, functional language task not performed on subject.

34LHBLxxFunctional activation maps indicated bilateral activation with stronger left language regions activated
35RHBLx Functional activation maps indicated bilateral activation with stronger left language regions activated
36LHBRxxxxFunctional activation maps indicated bilateral activation with stronger right language regions activated. Relative dominance of right hemisphere for language function compared with bilateral speech capacity suggested by Wada
37RHBRxxxxFunctional activation maps indicated bilateral activation with stronger right activation
38RHBLxxxFunctional activation maps indicated bilateral activation with stronger left language regions activated
39LHBLxxxFunctional activation maps indicated bilateral primary auditory activation with stronger left language regions activated
40LHBLxxxxFunctional activation maps indicated bilateral activation with stronger right Broca activation for the visual task and bilateral Wernicke activation for the auditory task. Wada results suggested dissociation of language function, with right Broca and bilateral comprehension

Functional LI scores based on whole-brain analysis and midline exclusion method

Initially a whole-brain analysis was used to calculate the functional laterality scores (Table 2). Not using a region of interest (ROI) based approach to laterality calculation allows for shifts in the language network localization, either due to mass effects (in the case of a large lesion) or to accommodate reorganization that may occur associated with the epileptogenic process. Using selected ROIs (based on normative group data, for example) can strongly influence the LI calculation. With use of the whole-brain analysis technique, it was found that the medial regions both in the occipital and frontal lobe often influenced the lateralization scores and that this was the source of some of the discrepancy between the fMRI lateralization and Wada results. To eliminate the contribution of occipital cortex to the laterality score while maximizing the contribution of the classical language areas of Broca and Wernicke, a semiautomated approach was used to eliminate midline cortical areas along the interhemispheric fissure without requiring explicit delineation of inferior frontal gyrus (IFG) and superior temporal gyrus (STG) language areas. This approach removed an approximately 2 cm wide strip of brain tissue centered on the interhemispheric fissure, running from the most anterior to posterior edges of the cortex. After implementing this midline exclusion technique, the functional lateralization results for many patients that were initially not in agreement with Wada demonstrated concordance. This approach produced more concordance with the Wada, and there was no case where this approach introduced disagreement with the Wada test when the two had previously agreed.

Improvement in concordance between functional LI scores and Wada test for group A with midline exclusion

Patient 17 in group A showed left hemispheric dominance (contralateral side of seizures) as determined by the Wada test. The functional language laterality score suggested bilateral dominance (Table 2), as it was driven by the visual cortex (occipital gyri) (Fig. 2A–D). When the visual cortex was excluded from the analysis the laterality score indicated left hemisphere dominance in agreement with the Wada findings. Another case in which the midline exclusion method was crucial was in patient 33, whose Wada indicated right hemispheric dominance (i.e., contralateral side of the resection) and the functional LI also indicated right hemispheric dominance for the reading comprehension, auditory comprehension, and conjunction analysis (Table 2). However, the LI score for the verbal fluency task showed left dominance (disagreement with the Wada result) driven by the left occipital gyri and the left orbital gyri. After the midline exclusion analysis, the LI score indicated right dominance, in agreement with the Wada finding. Wada testing for patient 19 indicated left hemisphere dominance, and again, using the midline exclusion technique to eliminate the occipital and orbital gyri, the functional MR results were in accordance with the Wada test. The midline exclusion technique improved concordance for those patients whose functional lateralization was falsely driven by bilateral or unilateral occipital and/or orbital activation. Therefore, in patient 30, for the auditory task there was no difference in lateralization score when the ROI technique was applied, as the functional maps were driven by bilateral temporal and frontal activation. However, in patient 18, because the orbital and occipital areas were influencing the verbal fluency lateralization results, eliminating them completely using the midline exclusion method improved the lateralization to become left dominant, which was in agreement with the Wada test. For the auditory task, the functional maps showed bilateral temporal and frontal activation and hence disagreement with Wada test results. The functional activation maps for patient 21 showed right hemispheric dominance (ipsilateral side of seizures) for the sentence comprehension, auditory comprehension, and verbal fluency tasks (Fig. 3A–C), which was in agreement with the Wada test results, whereas the intersection analysis showed disagreement (Fig. 3D). This was one case where using the midline exclusion analysis for the conjunction method did not lead to agreement with the Wada test, because the functional activation map demonstrated predominantly bilateral temporal and frontal activation.

image

Figure 2.   Patient 17, Functional language tasks (A) reading comprehension at threshold = 2, (B) auditory comprehension at threshold = 5, (C) verbal fluency at threshold = 4, and (D) conjunction analysis at threshold = 2 showing bilateral activation. The laterality index (LI) is driven by the visual cortex (occipital gyri).

Download figure to PowerPoint

image

Figure 3.   Patient 21, Functional activation maps (A) reading comprehension at threshold = 3, (B) auditory comprehension at threshold = 4, (C) verbal fluency at threshold = 5.5 showing bilateral activation with right hemispheric dominance in agreement with the Wada test results. (D) conjunction analysis functional map at threshold = 2 indicating bilateral temporal and left frontal activation.

Download figure to PowerPoint

Dependence of functional LI on threshold

A second issue to consider is that functional laterality scores are also threshold dependent. Binder et al. (1996) in their report correctly pointed out that language lateralization is a continuous rather than dichotomous variable. For the current study, from the graphs plotted as a function of laterality score versus threshold, a threshold value of t = 2 was found to be the most appropriate threshold to perform the LI calculation to validate the results with the Wada findings. An example of the dependence of the functional laterality score on the threshold is shown with patient 30. In this patient, for the auditory task when the threshold was changed from a value of 2 to a value of 3, the language lateralization shifted from bilateral dominance to left dominance, which was consistent with the Wada findings. Another example is patient 18, whose Wada findings indicated left hemispheric dominance. At a threshold of 2, the laterality of the verbal fluency task indicated bilateral dominance. But as the threshold was changed to 3 and higher, the laterality score demonstrated left dominance, which was concordant with the Wada test (Fig. 4A–B). At a threshold of 2, the functional activity map of the auditory task for patient 18 showed bilateral activation in the primary auditory cortex and bilateral frontal Broca activation with stronger right dominance, as indicated by the LI. As the threshold was increased to 3 and higher, the LI continued to demonstrate right dominance (Fig. 4C–D).

image

Figure 4.   Patient 18, Verbal fluency task at a threshold of (A) t = 2 showing bilateral dominance, (B) t = 3 showing left hemispheric dominance in agreement with the Wada finding. Auditory language task at a threshold of (C) t = 2, (D) t = 4 illustrating bilateral activation with right dominance.

Download figure to PowerPoint

Comparison of functional LI and Wada test for group B

The present study examined not only patients who had unilateral dominance but also those patients demonstrating bilateral language networks as determined by the Wada test. Several previous studies (Desmond et al., 1995; Benson et al., 1999; Lehéricy et al., 2000; Carpentier et al., 2001) have focused on patients with either left or right dominance as described by the Wada test. In many cases the fMRI results indicated some level of lateralization, noting of course that the LI score for fMRI ranges from −100 to +100, whereas the Wada score in the current study is essentially −1, 0, and 1, for right, bilateral, and left lateralization, respectively. The Wada test findings for patient 34 indicated bilateral speech, whereas the language LIs showed left hemispheric dominance (Table 2). The functional activation maps demonstrated bilateral activation with stronger left (Broca and Wernicke) than right hemisphere activation, and so the laterality scores were left dominant (Table 4). Patient 39 showed bilateral dominance as suggested by the Wada test, and the laterality score indicated left dominance (Table 2). However, the functional LI for the verbal fluency task demonstrated bilateral dominance (Fig. 5A), which is consistent with the Wada findings. The functional maps for the remaining language tasks (Fig. 5B–D) showed bilateral primary auditory cortex activation and left Broca activation (Table 4). Another example of such a case is patient 38, whose Wada test results demonstrated bilateral dominance. The MR laterality index for all the language tasks showed left hemispheric dominance, except for the auditory task, which showed bilateral activation in agreement with the Wada test (Table 2). The Wada results for patient 40 suggested dissociation of language function, with right Broca and bilateral Wernicke (Fig. 6A–B). For group B, the functional activation maps often had activations present on both sides (in agreement with the Wada test if these areas were indeed critical for language), but this activation was rarely completely even in extent, and thus the fMRI findings were often lateralized even when the Wada testing was not.

image

Figure 5.   Patient 39, (A) Verbal fluency task at a threshold of t = 2 showing bilateral dominance in agreement with the Wada finding. Functional language activation maps for (B) reading comprehension, (C) auditory comprehension, (D) conjunction analysis at a threshold of t = 2, illustrating bilateral primary auditory cortex activation and left Broca activation.

Download figure to PowerPoint

image

Figure 6.   The Wada results for patient 40 suggested dissociation of language function, with right Broca and bilateral comprehension. The functional activation maps indicated bilateral activation with (A) stronger right Broca activation for the reading comprehension task at threshold = 2 and (B) bilateral Wernicke activation for the auditory comprehension task at threshold = 2.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Summary
  3. Methods
  4. Results
  5. Discussion
  6. Conclusions
  7. Acknowledgments
  8. References

Previous studies have used various language paradigms in functional MRI to identify brain regions involved in different aspects of language processing (Desmond et al., 1995; Binder et al., 1996; Yetkin et al., 1998; Benson et al., 1999; Pujol et al., 1999; Springer et al., 1999; Lehéricy et al., 2000; Ramsey et al., 2001). In a report by Kamada et al. (2007), the authors investigated 87 patients with brain lesions to localize language dominance by applying fMRI and magnetoencephalography (MEG) techniques, and compared the results with those of the Wada test. Several studies have examined whole-sentence processing with emphasis on different components of semantic and syntactic language processing (Müller et al., 1997;Kang et al., 1999; Caplan et al., 2000, 2002;Ni et al., 2000; Caplan, 2001; Carpentier et al., 2001; Michael et al., 2001). A study by Kamada et al. (2006) reported dissociation of expressive and receptive language functions using combined fMRI and MEG methods. Binder et al. (2008) compared fMRI protocols for mapping speech comprehension systems and found that they differed in patterns of activation and lateralization. The functional paradigms used in the current study have been scanned previously in a population of healthy normal controls (Constable et al., 2004) and were designed to incorporate several features of language processing simultaneously, involving both interpretation of language (semantic and syntactic decision making) as well as verbal fluency, thereby activating distributed language circuits, including the temporal and frontal cortices (Constable et al., 2004). The goal of this study was to validate language LIs using functional imaging through comparison with the Wada procedure on a group of patients with intractable epilepsy who were candidates for neurosurgical intervention and explore the reasons for occasional disagreements between fMRI and Wada testing. For functional imaging to have a clinical application, especially for presurgical patients, it is crucial to predict hemispheric lateralization and localization and to employ paradigms involving both interpretation and production of language. In addition, both print and sound were used for stimulus presentation, allowing modality-specific regions to be assessed. Previous studies (Desmond et al., 1995; Benson et al., 1999; Lehéricy et al., 2000; Carpentier et al., 2001) evaluated patient groups with either left or right language dominance as determined by the Wada test and did not consider bilateral patients. In the present study, patients who had lateralized or bilateral hemispheric activation as determined by Wada testing were investigated to test the applicability of these language tasks for fMRI lateralization. From a practical point of view it is convenient to have both auditory and visual presentation of stimuli such that those subjects with very poor vision or who cannot read can still be mapped using the auditory task and those with hearing aids (that cannot be used in the magnet) can be mapped using the reading comprehension task.

Lateralization index

The language LI was calculated for each individual task alone, and a conjunction analysis was also performed. The conjunction analysis proved superior in establishing hemispheric dominance as determined by concordance with Wada findings (91.3%) compared to the individual language tasks alone (Table 2). For the present study, LI score of greater than +10 was categorized as left hemisphere dominant, less than −10 as right hemisphere dominant, and any LI score in the range +10 to −10 was classified as bilateral. This categorical classification is important when making comparisons with the Wada testing. If the range of laterality scores that determine bilateral language is broadened beyond the ±10 range that was used, the concordance between the Wada and fMRI improves as one would expect (e.g., ±40 has a concordance of 0% for the conjunction analysis and ±95 yields a concordance of 60%); however, this leads to a decrease in concordance for the unilateral patients (e.g., for the conjunction analysis with the ±40 cutoff, the concordance for the unilateral patients drops from 91.3% to 69.57%, and to 56.52% for the ±95 threshold). Because the primary goal for fMRI in this application is to lateralize patients, we used the ±10 conservative threshold to maximize the number of patients that were lateralized. A number of activation measures such as voxel count, sum of t-values, percentage signal change, product of voxel count * sum of t-values can be employed to calculate the language LI. The results demonstrate that such measures are sensitive to threshold effects and that the product of voxel count * sum of t-values provided the highest lateralization coefficients (Fig. 1).

Midline exclusion method

Many studies (Gaillard et al., 2002; Woermann et al., 2003; Benke et al., 2006) have employed a visual assessment method to determine hemispheric lateralization, whereby activated regions known not to be of interest for language mapping can be simply ignored. The midline exclusion approach used in this study accomplishes this task automatically by eliminating from the analysis many noncritical language regions along the interhemispheric fissure. Eliminating some of these midline activations improved the concordance between the fMRI and Wada findings. Initially a whole-brain analysis to calculate the functional LI should be applied. In patients demonstrating significant bilateral or unilateral occipital and/or orbital activation, which often reduce the lateralization scores, we suggest the midline exclusion method be applied to provide a more accurate functional LI. Also for patients in whom the functional activation maps are noisy at lower thresholds, using a higher threshold will generate cleaner activation maps and may improve the lateralization.

Role of the nondominant hemisphere

A study by Just et al., 1996, suggested that increased recruitment of the nondominant hemisphere in language tasks might reflect a compensatory response to impoverished left hemisphere language function. Springer et al. (1999) found a greater variability of language dominance in epilepsy patients when compared to a control group, and Carpentier et al. (2001) also found increased bilateral activation in patients. The nondominant hemisphere also appears to be recruited with increasing task difficulty. In the present study, patient 17 had difficulty understanding the language tasks and the functional laterality showed atypical dominance (Tables 2 and 3). Previous findings have suggested that epileptic patients with left handedness are the most likely to illustrate atypical language organization, establishing a relationship between left handedness and atypical dominance (Rey et al., 1988;Woods et al., 1988; Loring et al., 1990; Strauss et al.,1990; Satz et al., 1994; Helmstaedter et al., 1997; Risse et al., 1997). In the present study, in group A, seven patients were left handed. The Wada for four of them showed atypical dominance (right), and left dominance for the remaining three patients (Table 2). Three of these left-handed patients had conjunction analysis functional laterality scores, of which two showed atypical dominance and the remaining one showed left dominance (Table 2).

Table 3.   Group A—Summary of patients whose lateralization indices (LIs) and Wada findings disagree
Patient no.HandednessWadaLI of visual taskLI of auditory taskLI of word generation taskLI of logical AND of three language tasksPossible explanations for disagreement between functional laterality index and Wada findings
  1. LI, lateralization index; ✓, functional lateralization indices and Wada findings agree; x, functional lateralization indices and Wada findings disagree; –, functional language task not performed on subject.

16LHLHDxSusceptibility artifacts—presence of metal clips from prior brain surgery. Also subject was slow in responding to visual task. Only left hemisphere injection was performed, resulting in significant aphasia; considered typical dominant hemisphere injection
17RHLHDxxxxFunctional activation maps indicated bilateral activation as LI is driven by visual cortex (occipital gyri). Also subject had difficulty with understanding the functional language tasks
18RHLHDxxFunctional activation map indicated bilateral primary auditory cortex activation with bilateral frontal Broca activation
19RHLHDxxxFunctional activation maps indicated bilateral activation with stronger left language regions activated. LI is driven by occipital gyri and rbital gyri
20LHRHDxFunctional activation map indicated bilateral language regions activated. The Wada test suggested relative dominance of the left hemisphere for writing, drawing and right hemisphere for speech
21LHRHDxFunctional activation maps indicated bilateral language regions activated with stronger right hemispheric activation. Occipital gyri and orbital gyri activation influence lateralization scores
27RHLHDxxFunctional activation maps indicated bilateral language regions activated with stronger left activation. LI driven by occipital gyri and orbital gyri activation
28RHLHDxFunctional activation map indicated bilateral language regions activated with stronger left activation. LI driven by occipital gyri and orbital gyri activation
29RHLHDxxFunctional activation maps indicated bilateral language regions activated
30RHLHDxFunctional activation map indicated bilateral language regions activated with stronger left activation
31RHLHDxFunctional activation map indicated bilateral language regions activated with stronger right activation
33LHRHDxFunctional activation maps indicated bilateral language regions activated with stronger right activation. LI driven by occipital gyri and orbital gyri activation

Patient 36 was a remarkable case as the Wada test findings suggested relative dominance of the right hemisphere for language function compared with bilateral speech capacity, and the laterality findings were right dominant (Tables 2 and 4). Rare patients have been reported with interhemispheric dissociation of component linguistic processes (Kurthen et al., 1994; Risse et al., 1997). Springer et al. (1999) have hypothesized that in such cases the fMRI data should correlate more closely with the language comprehension aspects of Wada than with the vocal speech components as is seen with patient 36 in the present study. A relative dominance of the left hemisphere for writing and drawing and the right hemisphere for speech was suggested by the Wada test for patient 20, whereas the functional activation map indicated bilateral dominance (Tables 2 and 3).

Discrepancies exist between various functional studies when evaluating the Wada results and the functional language lateralization scores. Reasons for these discrepancies may include differences in the paradigm used, variations in subject population (pathology, epileptic location), and a lack of uniformity across different hospitals in the application of the Wada procedure. Heterogeneity of the patient populations themselves may also play a role. Patients with epilepsy typically cover a spectrum in terms of handedness, location of epilepsy, age of onset of seizures, neuropsychological profile, and anticonvulsant therapies, all of which can influence the results measured. For the group of patients determined to have bilateral language via Wada testing and lateralized fMRI findings in this study, there are a number of differences between these techniques that could contribute to these errors. As discussed by Swanson et al. (2007), some differences arise simply by the fact that fMRI LI is a continuous variable while we are comparing this to a trichotomous variable in the Wada measure. Other factors include vascular effects (Hietala et al., 1990), arousal effects (Malmgren et al., 1992), and effects related to the time-response of the language and/or motor regions to the drug administered.

Conclusions

  1. Top of page
  2. Summary
  3. Methods
  4. Results
  5. Discussion
  6. Conclusions
  7. Acknowledgments
  8. References

A number of studies have demonstrated good concordance between the functional MR laterality scores and the Wada test findings. The current study found excellent agreement between the two methods but also explored factors that influence the laterality score measured with fMRI. It is crucial to understand the neurophysiology of language processing, both in control subjects and in epilepsy patients, in order to better interpret the results from fMRI. Functional MRI provides a noninvasive alternative to Wada testing for lateralizing language with the added benefit of providing within-hemisphere regional localization of function. At many clinical sites, Wada testing is being eliminated in cases in which the fMRI results are highly lateralized. If the results from the fMRI lateralization are equivocal, Wada testing is then used to resolve the issue. This work suggests that care must be taken in using fMRI for language lateralization and that there are many sources of error that must be considered.

Acknowledgments

  1. Top of page
  2. Summary
  3. Methods
  4. Results
  5. Discussion
  6. Conclusions
  7. Acknowledgments
  8. References

Support from NIH NINDS NS38467-07. We thank Hedy Sarofin, Karen Martin, and Terry Hickey for help with running the MR experiments. We confirm that we have read the Journal’s position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.

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

References

  1. Top of page
  2. Summary
  3. Methods
  4. Results
  5. Discussion
  6. Conclusions
  7. Acknowledgments
  8. References