Sage Journals: Discover world-class research

Abstract

Repeated functional magnetic resonance imaging (fMRI) during learning and recovery can inform us about functional reorganization in the brain. We examined how, in the absence of reorganization, simple task repetition affects measures of fMRI activation. We studied fMRI activation over 10 consecutive sessions of silent word generation in a healthy subject. Additionally, we performed functional transcranial Doppler sonography (fTCD) to learn about the temporal pattern of corresponding changes in cerebral blood flow velocity (CBFV) and pulsatility. With repetition, word generation-associated increases in heart rate diminished steadily. Task repetition also led to a net increase in CBFV bilaterally and to a bilateral increase in the number of activated voxels on fMRI. As a result, whereas the absolute interhemispheric difference of activated voxels remained constant, there was a decrease in the standard fMRI index for language lateralization [LI_fMRI = 100 (activated voxels in left hemisphere - voxels in right)/(voxels in left + voxels in right)]. Thus task repetition can lead to changes in task-related autonomic drive and an augmentation of bihemispheric blood flow. This can mimic increasing bilaterality of brain activation.

Keywords

Functional magnetic resonance imaging Functional transcranial Doppler sonography Arousal Habituation Language Cerebral lateralization Attention

Functional magnetic resonance imaging (fMRI) and positron emission tomography (PET) can be used to map cerebral reorganization during recovery of function in patients after stroke (Cao et al., 1999; Gold and Kertesz, 2000; Musso et al., 1999; Rosen et al., 2000). Thus increased bilaterality of language related brain activation has been observed in serial studies after stroke to left hemisphere language regions (Silvestrini et al., 1993; Weiller et al., 1993). This increased bilaterality has been speculated to be the result of a transhemispheric reorganization recovery (Thulborn et al., 1999). It could constitute a major basis for language recovery (Weiller et al., 1995). To better understand transhemispheric reorganization, patients need to be scanned at different time points during recovery. This way, we have a chance to observe functional brain repair online. However, care must be taken not to confuse repair-specific changes in brain activation with cerebral blood flow changes associated with repeated scanning.

Activations on fMRI are affected by a number of factors, which may change from one session to another (McGonigle et al., 2000). These factors include technical ones, like magnetic field inhomogenities, and physiologic ones, like autonomic drive, arousal, attention, and strategy. Therefore, the reproducibility of fMRI activation patterns within sessions or between two or more sessions is variable (DeYoe et al., 1994; Gozal et al., 1994; Karni et al., 1995; Loubinoux et al., 2001; Noll et al., 1997; Ramsey et al., 1996; Rombouts et al., 1998; Silva et al., 1999; Tegeler et al., 1999; Yetkin et al., 1996).

To circumvent confounding variations in cerebral blood flow or blood oxygenation level dependent (BOLD) signal, data are frequently normalized. For example, in studies on the interhemispheric reorganization of the language system after stroke, interhemispheric differences in activation are mostly related to the total (bihemispheric) number of activated voxels. The formula used is as follows: (activation in the left hemisphere – activation in the right hemisphere)/(activation in the left hemisphere + activation in the right hemisphere) (Lehericy et al., 2000; Vikingstad et al., 2000). The assumption underlying the use of this formula is that global, that is, bihemispheric signal changes, will cancel out.

In the present study, we challenged this approach by repeating fMRIs during word generation over a total of 10 sessions. To learn about the dynamics of blood flow changes in the supplying arteries during the language task, we additionally measured the cerebral blood flow velocities (CBFV) in the middle cerebral arteries (MCA) by functional transcranial Doppler ultrasonography (fTCD. For a review of the technique see Deppe et al. (1997) and Stroobant and Vingerhoets (2000).

METHODS

Functional MRI and functional TCD were performed in a single subject over 10 consecutive sessions within 2 months with variable intersession intervals. During each session, the subject was examined by fMRI and fTCD using a silent word generation task (see below for details). For each fMRI session, the number of activated voxels and two types of language lateralization indices (LI_fMRI and LI_fMRI-%) were calculated. During each fTCD session, the CBFV changes in the insonated MCAs during task performance relative to rest condition and the fTCD language lateralization index (LI_fTCD) were assessed (see sections below).

Subject

One healthy German-speaking female (24 years of age) participated in this study after giving informed consent. A small hourly expense allowance was paid. The subject was righthanded according to the Edinburgh Handedness Inventory (degree of handedness = +69) (Oldfield, 1971).

Functional magnetic resonance imaging.

Word generation.

For language activation, the subject was required to silently find as many words as possible beginning with a given letter (A, O, L, B, D, and E). Task commands and letter stimuli were presented acoustically using a nonmagnetic headphone. The acoustical letter presentation time was 0.5 seconds. During the resting state, the subject was requested to relax and received no further instructions.

The fMRI session was designed according to a classical “boxcar” paradigm. The first five images of each fMRI session were discarded to account for saturation effects of the MR signal. The fMRI scanning sequence then comprised three activation intervals of silent word generation (SWG-I, -II, -III) separated by two resting periods (R-I, -II). To control for compliance, the last activation interval (SWG-III) was followed by a speaking interval (SP) in which the subject was requested to report the words aloud she found in the previous activation interval (SWG-III). Each activation interval and resting period covered 10 image acquisitions. In the speaking interval (SP), five images were acquired. The scanning sequence was performed three times in succession. Thus, in the whole fMRI session, a series of 3 ϗ (10 + 10 + 10 + 10 + 10 + 5) = 165 images were acquired.

Functional magnetic resonance imaging data acquisition.

FMRI sessions were performed on a 1.5 Tesla Magnetom Vision scanner (Siemens; Erlangen, Germany) using a circular polarized head coil. Data acquisition was conducted using echo planar imaging sequence [EPI; echo time (TE) 64 ms; repetition time (TR) 3 seconds; flip angle 90°; 64 × 128 pixel acquisition matrix, reconstructed to 128 × 128 pixel; 210 mm × 210 mm field of view; slice thickness 3 mm; slice gap 0.75 mm] for functional imaging. Ten axial slices at a distance of 3.75 mm were acquired oriented parallel to the intercommisural plane. Additionally, 10 T₁-weighted images (TE 15 ms; TR 350 ms; flip angle 90°) in the same orientation of the EPI planes were acquired to facilitate the anatomic assignment of the functional images.

High-resolution T₁-weighted reference images in sagittal scanning direction (MPRAGE sequence; 256 mm × 256 mm field of view, slice thickness 1 mm, voxel size 1 mm × 1 mm × 1 mm, TR 11.4 seconds, TE 4.4 seconds, flip angle 15°) were additionally obtained to relate the activation pattern to neuroanatomic structures. Standardized head fixation devices for clinical sessions (head and neck padding) were used to reduce head motion artefacts.

Functional magnetic resonance imaging data analysis.

Image preprocessing and statistical analyses were performed using the SPM99 software package (Wellcome Department of Cognitive Neurology, London, UK). SPM was used for realignment, coregistration, smoothing, and creating statistical maps of significant relative regional BOLD response changes (Friston, 1995a, b ).

The first five images of each time series, during which the MR signal reaches a steady state, were omitted from further analysis. All other functional images were realigned using sinc interpolation to the first remaining image (corresponding to the sixth acquired image) to correct for motion artefacts. A mean functional image was calculated from the realigned images. Subsequently, the functional images were smoothed using a Gaussian kernel of 4 mm × 4 mm × 9 mm (FWHM) corresponding to an anisotropic voxel size of 1.64 mm × 1.64 mm × 3.0 mm. The anatomic image was coregistered to the mean functional image.

The functional data from each session were analyzed using a general linear model approach as implemented in SPM99. A reference waveform corresponding to the alternating activation and control conditions was constructed and convoluted with an estimate of the hemodynamic response function. Specific effects were tested by applying appropriate linear contrasts to the parameter estimates for the experimental condition, resulting in a t-statistic for each voxel. Voxels were identified as significant if they passed a P value of 0.001 uncorrected and belonged to a cluster of activation with an extent of at least 10 voxels. The significance level of P = 0.001 (uncorrected) was selected because it represented the P value where the language lateralization index LI_fMRI (see below) showed the maximal reproducibility and was nearly independent of moderate changes in P. If, for example, P = 0.05 (corrected for multiple comparisons) was applied, the longitudinal reproducibility of LI_fMRI decreased.

Functional magnetic resonance imaging lateralization index.

The calculation of the functional magnetic resonance imaging lateralization index (LI_fMRI) was based upon the activated voxels within a specified region (see below). The LI_fMRI was calculated for each session by the following formula:

L I_{fMRI} = 100 \frac{N_{L} - N_{R}}{N_{L} + N_{R}}

(1)

where N_L and N_R refer to the number of activated voxels within the regions of interest (ROI). Positive values indicate increased numbers of activated voxels in the left hemisphere relative to the right hemisphere and thus reflect left-hemispheric dominance and vice versa.

Two different ROIs were defined. First, all activated voxels in the whole brain (without cerebellum) were accepted to determine global brain activation. Second, language-related ROIs were defined for the experimental task: Based upon experiences from previous fMRI language lateralization studies using word generation tasks, language-related ROIs were restricted to the anterior section of the vascular territory of the MCAs (see Fig. 1). Anatomic landmarks identified in T₁-weighted images acquired before functional scanning guided the determination of the language-related ROIs. ROIs comprised areas surrounding the inferior frontal gyrus, including the classical Broca's area and their contralateral homologues (Binder, 1997; Binder et al., 1997).

Functional magnetic resonance imaging lateralization index based upon percentage signal change.

A second fMRI lateralization index (LI_fMRI-%) was calculated according to Equation 1, but the number of activated voxels was replaced by the mean change of the BOLD signal within two spherical ROIs of 20 mm diameter within the left and right hemispheres. The center of the left ROI was adjusted to the center of mass of the most significant cluster corresponding to Broca's area (BA 44/45). The position of the right ROI (contralateral homologue of Broca's area) was determined by mirroring the left ROI to the right hemisphere.

Functional transcranial Doppler sonography

Word generation

The hemispheric language lateralization was assessed by a standardized word generation task used in a number of previous studies and validated by direct comparison with the Wada procedure (Deppe et al., 2000; Knake et al., 2003; Knecht et al., 1998; Wada and Rasmussen, 1960). In short, during the fTCD session, the subject was expecting a target letter on a computer screen 5 seconds after a cueing tone. The subject was instructed to then silently generate as many words as possible with the given letter within 15 seconds (silent word generation phase). For control, the subject was requested to report the words aloud after a second cueing tone. A third tone initiated a resting phase of 30 seconds. Each session consisted of 20 trials. Each trial began with a different target letter. The order of the target letters was randomized. The letters “X” and “Y” were excluded from presentation.

Data recording.

During the word generation task, the event-related cerebral blood flow velocities (CBFV) in both MCAs were recorded using a 2 MHz-pulsed Doppler device (DWL Multi Dop T; Siplingen, Germany). The transducer probes were attached to a headband and fixed at the temporal bone windows bilaterally. The spectral envelope curves of the Doppler signal of both MCAs were recorded at a sample rate of 28 points per second. The data were stored for further data analysis. Further technical details and information about practical techniques of the sonographical assessment have been published elsewhere (Ringelstein et al., 1990).

Data analysis.

Offline data analysis was performed by the computer program AVERAGE, a software package for functional CBFV data analysis introduced by one of the authors (Deppe et al., 1997). Data analysis involved several processing steps, that is, artefact reduction, filtering, cardiac integration, and normalization of the data. The continuously acquired data set of 20 trials was segmented into 20 epochs related to the onset of the cueing tone. The epochs range from 15 seconds before to 35 seconds after the first cueing tone. The mean velocity in the 15-second precueing interval (V_pre.mean) was chosen as the CBFV baseline value in the resting phase. Relative changes of the CBFV (dV) in the word generation task were calculated by the formula:

d V = 100 \frac{V (t) - V_{pre . mean}}{V_{pre . mean}}

(2)

where V(t) is the CBFV over time.

Functional transcranial Doppler sonography lateralization index.

The Functional transcranial Doppler sonography lateralization index (LI_fTCD) was computed by the following formula:

L I_{fTCD} = \frac{1}{t_{int}} \int_{t_{{max}^{- 0.5 t_{int}}}}^{t_{{max}^{+ 0.5 t_{int}}}} Δ V (t) d t

(3)

where

Δ V (t) = d V {(t)}_{left} - d V {(t)}_{right}

(1)

was the difference between the relative CBFV changes in the left and right MCA. The time point t_max represents the latency of the absolute maximum of V (t) within the word generation interval (5 to 22 seconds). As the integration interval, a time period of t_int = 2 seconds was chosen. The LI_fTCD defines the maximum of relative CBFV changes in a preset period within the silent word generation phase in comparison with the CBFV baseline. A positive value of the LI_fTCD represents a relative increase of the CBFV in the left MCA during the word generation phase and vice versa. Thus a positive value of LI_fTCD indicates left hemispheric language processing.

Accuracy.

The coefficients of variability (CV) for all three lateralization indices (CV_fMRI, CV_fMRI-%, and CV_fTCD) were defined by the following formula:

C V = \frac{s_{L I}}{L I}

(5)

where

\overline{L I} = \frac{1}{N} \sum_{i = 1}^{N} L I^{i}; N = 10

(6)

is the mean of all 10 consecutively measured LIs and

s_{L I} = \sqrt{\frac{1}{N} \sum_{i = 1}^{N} (L I^{i} - \overline{L I})}

(7)

is the corresponding standard deviation. A lower CV represents lower relative variation of the LIs assessed by one method and thus better accuracy.

Cardiovascular parameter.

During the fTCD sessions, the heart rate was monitored to assess changes in cardiovascular drive. This was accomplished by a program option offered by the software AVERAGE (Deppe et al., 1997).

Statistical analysis

Linear regression functions (Spearman's R, two-tailed) were computed to examine the time dependency (independent variable) of the following parameters (dependent variables):

Number of activated voxels in the defined ROIs (see Figs. 3A–C)

Degree of two fMRI lateralization indices (LI_fMRI and LI_fMRI-%) (see Figs. 3D–E)

Task-related CBFV increases (see Fig. 4B and Table 2)

Changes in the heart rate during fTCD task performance relative to rest (see Fig. 5 and Table 2).

Fig. 1.

Language task-related regions of interest: the vascular territories of the three basal cerebral arteries are indicated in green (middle cerebral artery), red (anterior cerebral artery), and blue (posterior cerebral artery). The anterior sections of the vascular territory of the middle cerebral arteries (areas marked in linked grey bars) were defined according to the program VOXELMAN (Höhe, 1997).

Fig. 2.

Distribution of activated voxels during word generation in every second (1, 3, 5, 7, and 9) session, illustrated by one identical functional axial slice: there is a clear left-sided asymmetry of activated voxels and a clustering within Broca's area, superimposed by an increasing bilateral activation.

Fig. 3.

(A) Increase in activated voxels over the course of 10 sessions in the regions corresponding to the anterior vascular territories of the middle cerebral arteries (MCAs; loosely termed Broca and Broca Homologue). (B) Difference of activated voxels (lower curve; V_L - V_R; squares) in the anterior vascular territories of the MCAs, indicating a slight but not significant increase, whereas there is a substantial bilateral increase within the (left) Broca region and its (right) homologue (upper graph; N_L + N_R; circles). (C) The same trend is found within the whole vascular territory of the MCAs, indicating a global increase of activated voxels. (D) As a consequence from the changes depicted in B, lateralization indices LI_fMRI decrease with repeated examinations. (E) In contrast, the lateralization indices based upon percentage change (LI_fMRI-%) were variable without a statistically significant trend. Trends, thick solid and dotted lines; regions of confidence, 95%, thin solid and dotted curves.

Fig. 4.

(A) Averaged changes in CBFV in the left and right MCA assessed by fTCD during task performance relative to resting condition. Note that the CBFV in the left MCA increases more in comparison with the right MCA, reflecting language processing in the left hemisphere. (B) Averaged event-related CBFV in the left and right MCA with standard deviation (bars) and trends (solid and dotted lines). The bilateral decrease in CBFV (Fig. 4A) diminishes over the course of the 10 sessions. This results in a global increase of cerebral perfusion in both hemispheres. The three phases of CBFV changes are marked with numbers for later discussion. CBFV, cerebral blood flow velocity; MCA, middle cerebral artery; fTCD, functional transcranial Doppler sonography.

Fig. 5.

Heart rate changes during the word generation phase relative to the resting condition averaged over 20 trials in each session. Trends, solid lines; regions of confidence, 95%, dotted curves.

RESULTS

Functional magnetic resonance imaging

Initially, word generation led to activation in the left inferior frontal lobe. Over the course of repeated examinations, this left-sided activation was superimposed by a bilateral increase in activated voxels (Fig. 2). Numerical analysis indicates a significant increase in activated voxels in the regions corresponding to the anterior territories of the left and right middle cerebral arteries (see Figs. 3A and 3B, upper panel; for trend statistics, see Table 1, nos. I, II, and IV) and the entire vascular territory of the left and right MCA (see Fig. 3C; for trend statistics, see Table 1, no. VI).

Conversely, the difference between the activated voxels in the left versus right anterior vascular territory of the MCA did not change significantly (see Fig. 3B, lower panel; for trend statistics, see Table 1, no. III).

In each of the 10 sessions, positive LI_fMRI and LI_fMRI-% were found, indicating left hemispheric language processing. Over repeated sessions, the LI_fMRI decreased significantly (see Fig. 3D; for trend statistics, see Table 1, no.V) in contrast to the LI_fMRI-% (Fig 3E; Table 1, no. VII).

Functional transcranial Doppler sonography

In all 10 fTCD sessions, positive LI_fTCD were obtained, indicating left hemispheric language dominance. Unlike in the fMRI assessments, no substantial trend in the language lateralization index (LI_fTCD) was observed over time (for trend statistics, see Table 2, no. I).

TABLE 1.

Repeated fMRI sessions: trend statistics

No.	Area/index	Regression function	Correlation (R)	P level
I	Broca (left)	N = 366 + 90 S	0.69	0.025
II	Broca homologue (right)	N = −23 + 64 S	0.74	0.013
III	Difference between (I) and (II)	N = 389 + 26 S	0.29	0.405
IV	Sum of (I) and (II)	N = 343 + 154 S	0.69	0.025
V	LI_fMRI (=III/IV)	LI = 0.769 – 0.05 S	-0.75	0.011
VI	Vascular territory MCA (left + right)	N = 545 + 159 S	0.66	0.038
VII	LI_fMRI-%	LI = 23.9 + 1.8 S	0.25	0.480

N = activated voxels; S = session no.; significance based on Spearman's R, two-tailed; MCA, middle cerebral artery.

TABLE 2.

Repeated fTCD sessions: trend statistics

No.	Index/parameter	Regression function	Correlation	P level
I	LI_fTCD	LI = 5.0 – 0.008 S	-0.16	0.581
II	Averaged event-related CBFV (left)	CBFV = −0.3 + 0.16 S	0.34	0.223
III	Averaged event-related CBFV (right)	CBFV = −4.8 + 0.25 S	0.29	0.184
IV	∆ Heart rate relative to baseline	∆ HR = 0.02 – 0.005 S	-0.48	0.153

Significance based upon Spearman's R, two-tailed.

HR, task-related heart rate changes relative to baseline; S, session number; fTCD, functional trans- cranial Doppler sonography; CBFV, cerebral blood flow velocity; LI, lateralization index.

The hemodynamic response in both MCAs during task performance is shown in Fig. 4. All following data are based upon averaged event-related cerebral blood flow data processed by the software AVERAGE (Deppe et al., 1997). They represent the mean CBFV data of 20 trials during a specific session. Figure 4A shows the typical hemodynamic response pattern in the left and right MCAs during rest and task performance. The graph plots the percentage of CBFV changes during task performance relative to the resting condition. During task performance (time period, 0 to 23 seconds), the CBFV modulation in both MCAs is characterized by a biphasic increase in response to the cueing (first peak, Fig. 4A, mark 1) and the letter presentation (second peak, Fig. 4A, mark 2), followed by a marked bilateral decrease (Fig. 4A, mark 3). Figure 4B demonstrates that the bilateral decrease in CBFV at the end of word generation diminished over the course of the 10 sessions. This resulted in a relative bilateral increase of the CBFV over the course of the 10 sessions (for statistics, see Table 2, nos. II and III) relative to the resting condition. The pattern of a relative bilateral CBFV increase was confirmed in a second subject (female, 31 years of age; regression functions: averaged event-related CBFV_{left MCA} = 2.01 + 0.44 S; R = 0.58, P = 0.06; averaged event-related CBFV_{right MCA} = 0.35 + 0.33 S; R = 0.61, P = 0.07M) (unpublished data)

The heart rate during the word generation phase relative to the resting condition was approximately 103% (first session) and showed a negative trend with repetition of the task (Fig. 5 and Table 2, no. V for trend statistics). No such trend was observed during the resting phase.

Accuracy

The coefficients of variability were CV_fMRI-% = 0.63, CV_fMRI = 0.35, and CV_fTCD = 0.17, respectively.

DISCUSSION

The present study demonstrates that repeated fMRI examinations during word generation lead to a systematic decline in the fMRI lateralization index, calculated by the formula LI_fMRI = 100 (V_L - V_R)/(V_L + V_R). This decrease is caused by an increase in bilaterally activated voxels, whereas the difference between activated voxels in left and right hemisphere remains constant. Functional TCD demonstrates that the bihemispheric increase in activated voxels is caused by a bilateral increase in the event-related cerebral blood flow velocity.

Methodologic considerations

We did not conduct a group study. Pseudoincrease in bilateral blood flow was demonstrated in two subjects examined by fTCD and confirmed in one of them by fMRI. The effect was a consistent and incremental phenomenon over the 10 repeated examinations by fTCD as well as fMRI. We cannot exclude that there are individuals who will not show such response properties, but we believe that our data prove the principle.

Two different approaches have been taken in the past to calculate the fMRI hemispheric lateralization index for brain functions. The first is to measure brain activity based upon the extent of the activated brain region, that is, the number of activated voxels in a predefined ROI at a chosen statistical threshold (e.g., Binder et al., 1996; Deppe et al., 2000; Desmond et al., 1995). The second is to measure the magnitude of the fMRI signal change in a ROI (e.g., Adcock et al., 2003; Cohen and DuBois, 1999).

For clinical studies, the first approach is still the most popular (Binder et al., 1996; Pujol et al., 1999; Springer et al., 1999; Thulborn et al., 1999). Therefore, we focused the data analysis on this method. However, a number of recent studies showed that the stability of this measure is relatively poor. It is biased by several factors, most of all by the statistical threshold used to identify activated voxels (Nagata et al., 2001). Furthermore, its reproducibility is low (Adcock et al., 2003; Cohen and DuBois, 1999; Nagata et al., 2001; Rutten et al., 2002). Therefore, we recalculated the lateralization index using the signal change between the task and control condition as measure of activity (LI_fMRI-%). No choice of threshold is involved in calculating these indices.

The lateralization index based on the magnitude of signal change does not show a significant correlation with time (r = 0.25, n.s.). However, this is caused by a lower accuracy (CV_fMRI-% = 0.63) compared with the lateralization index based upon voxel statistics (CV_fMRI = 0.35) and fTCD (CV_fTCD = 0.17). The two latter approaches demonstrated congruently bilateral increases in perfusion. We acknowledge that our findings may not be a general phenomenon but specific to our case. Nevertheless, our results show that calculating a lateralization index based upon the magnitude of signal change does not necessarily produce more stable results.

Potential explanations

There are several different mechanisms that could explain our findings. The effect could be related to variations of hardware characteristics of the MR scanner, increasing practice and proficiency on the task, or autonomic habituation.

Changes in brain activation across fMRI sessions can be influenced by a number of technical factors. In our scanning series, all scanning parameters were kept as constant as possible. However, it cannot be excluded that there were still variations in some hardware characteristics, such as magnetic field inhomogenities (Howseman et al., 1998). These variations should lead to random fluctuations in the number of activated voxels. In contrast, our fMRI data indicate a systematic increase of bilateral activated voxels across sessions. That makes it unlikely that technical factors are responsible for the systematic increase of activated voxels. Furthermore, our fMRI results are corroborated by the concomitant fTCD data, which are based upon completely different physical phenomena, that is, the CBFV-dependent Doppler shift of ultrasound waves. Event-related CBFV responses increased across sessions, corresponding to the fMRI findings. FTCD and fMRI both measure blood flow increases. Different studies report a close relationship between regional cerebral blood flow (rCBF) as measured by fMRI and CBFV changes as measured by fTCD. The results by the one technique can be translated in a linear fashion into that by the other (Deppe et al., 2000; Sitzer et al., 1994).

Only a few studies have investigated the effect of practice on changes the pattern of brain activation (Petersen et al., 1998; Raichle, 1994). Changes in brain activation associated with practice effects during a lexical verbal fluency task have not been examined so far, to our knowledge. One could argue that task repetition leads to more bilateral brain activation because of an increasingly easier word retrieval and more effective lexical search recruiting distributed (bilateral) cerebral networks. Because of a limited number of target letters in the fMRI experiment, the same trials have to be presented more than once. However, results from our own laboratory do not support this idea. From a previous fTCD study, we know that differences in performance in a word generation task do not influence the degree of hemispheric activation among subjects (Knecht et al., 2000). Dräger and Knecht found no increased bilaterality in CBFV during a word generation task within subjects when task difficulty is increased (Dräger and Knecht, 2002).

Thus the present findings seem to be best explained by a change in the interplay of autonomic factors. Such changes take place when cognitive tasks are performed repeatedly. They are most marked for task-related arousal and attention (Näätänen, 1992). They involve rate and volume of respiration, as well as cardiac output and heart rate (Faucheux et al., 1989; Frey and Siervogel, 1983; Rombouts, 1982). It has been known for a long time that with task repetition cerebral blood flow changes even during the resting phases (Blauenstein et al., 1977; McHenry et al., 1978; Warach et al., 1988). Our CBFV data indicate that autonomic habituation has also occurred in our study. The confrontation with the experimental setting can be conceived as a novel stimulus, which arouses and attracts attention that then diminishes with repeated exposure (Warach et al., 1992). Respiration changes during the orienting response when a subject focuses attention to a stimulus or reacts to it (Näätänen, 1992).

Respiration affects CBFV (Tiecks et al., 1995). After cueing and during word generation, subjects usually breathe in deeply. Tiecks and colleagues demonstrated a breathing-related bilateral decrease in CBFV, which was similar to that observed in our study. Inspiration reduces venous return and can lower cardiac output which is then compensated by an increased heart rate via the Bainbridge reflex (Bainbridge, 1915). This could explain why fTCD showed a CBFV pattern consistent with a task-related CBFV increase superimposed upon a global CBFV decrease during word generation (Fig. 4A, phase 3). Additionally, this scenario could explain an increase in heart rate during the task. Increasing familiarity with a task should attenuate changes in respiration. This could explain why global CBFV during word generation increased and heart rate decreased with task repetition.

Our data show that the usual normalization procedure used in measuring task related lateralization of brain activation by fMRI cannot compensate for the confounding effect of task repetition. This is critical in studies on brain activation during learning or recovery of function. To control for the influence of a changing global signal, alternative statistical approaches have been proposed for fMRI activation. In a single-subject fMRI scanning series, McGonigle and co-workers demonstrated that the inclusion of session-by-condition interaction (i.e., influence of session context on the activation effect) explained a significant amount of error variance (McGonigle et al., 2000). The use of a statistical random-effects-analysis was recommended to account for both within- and between-session variance. However, this requires an appropriate sample size (i.e., number of sessions), which is not feasible in most clinical settings. As an alternative, a covariate could be introduced to the analysis of covariance, which models the shifting global signal (Desjardins et al., 2001; Friston et al., 1990). Defining such a covariate could involve the use of a control task. Our findings underscore the importance of accounting for repetition-related changes to improve the validity of fMRI as a tool for understanding reorganization. Distinguishing changes in brain activation patterns related to task repetition will help us to more precisely identify the changes related to the dynamic process under study, such as brain repair. This in turn will give us a better basis for understanding the underlying mechanisms of functional recovery.

References

Adcock

Wise

Oxbury

Matthews

(2003) Quantitative fMRI assessment of the differences in lateralization of language-related brain activation in patients with temporal lobe epilepsy. Neuroimage 18:423–438

Bainbridge

(1915) The influence of venous filling upon the rate of the heart. J Physiol 50:65–78

Binder

(1997) Neuroanatomy of language processing studied with functional MRI. Clin Neurosci 4:87–94

Binder

Frost

Hammeke

Cox

Rao

Prieto

(1997) Human brain language areas identified by functional magnetic resonance imaging. J Neurosci 17:353–362

Binder

Swanson

Hammeke

Morris

Mueller

Fischer

Benbadis

Frost

Rao

Haughton

(1996) Determination of language dominance using functional MRI: a comparison with the Wada test. Neurology 46:978–984

Blauenstein

Halsey

Wilson

Wills

Risberg

(1977) ¹³³Xe inhalation method. Analysis of reproducibility and some of its physiological implications. Stroke 8:92–95

Cao

Vikingstad

George

Johnson

Welch

(1999) Cortical language activation in stroke patients recovering from aphasia with functional MRI. Stroke 30:2331–2340

Cohen

DuBois

(1999) Stability, repeatability, and the expression of signal magnitude in functional magnetic resonance imaging. J Magn Reson Imaging 10:33–40

Deppe

Knecht

Henningsen

Ringelstein

(1997) AVERAGE: a Windows program for automated analysis of event related cerebral blood flow. J Neurosci Methods 75:147–154

10.

Deppe

Knecht

Papke

Lohmann

Fleischer

Heindel

Ringelstein

Henningsen

(2000) Assessment of hemispheric language lateralization: a comparison between fMRI and fTCD. J Cereb Blood Flow Metab 20:263–268

11.

Desjardins

Kiehl

Liddle

(2001) Removal of confounding effects of global signal in functional MRI analyses. Neuroimage 13:751–758

12.

Desmond

Sum

Wagner

Demb

Shear

Glover

Gabrieli

Morrell

(1995) Functional MRI measurement of language lateralization in Wada- tested patients. Brain 118:1411–1419

13.

DeYoe

Bandettini

Neitz

Miller

Winans

(1994) Functional magnetic resonance imaging (FMRI) of the human brain. J Neurosci Methods 54:171–187

14.

Dräger

Knecht

(2002) When finding words becomes difficult: is there activation of the subdominant hemisphere? Neuroimage 16:794–800

15.

Faucheux

Lille

Baulon

Landau

Dupuis

Bourliere

(1989) Heart rate and blood pressure reactivity during active coping with a mental task in healthy 18- to 73-year-old subjects. Gerontology 35:19–30

16.

Frey

Siervogel

(1983) Cardiovascular response to a mentally stressful stimulus. Jpn Heart J 24:315–323

17.

Friston

(1995a) Spatial registration and normalization of images. Hum Brain Map 165–189

18.

Friston

(1995b) Statistical parametric maps in functional imaging: a general linear approach. Hum Brain Map 189–210

19.

Friston

Frith

Liddle

Dolan

Lammertsma

Frackowiak

(1990) The relationship between global and local changes in PET scans. J Cereb Blood Flow Metab 10:458–466

20.

Gold

Kertesz

(2000) Right hemisphere semantic processing of visual words in an aphasic patient: an fMRI study. Brain Lang 73:456–465

21.

Gozal

Hathout

Kirlew

Tang

Woo

Zhang

Lufkin

Harper

(1994) Localization of putative neural respiratory regions in the human by functional magnetic resonance imaging. J Appl Physiol 76:2076–2083

22.

Höhne

(1997) Voxel-Man: Part I Brain and Skull Edition 2. New York: Springer Verlag

23.

Howseman

McGonigle

Grootoonk

Ramdeen

Athwal

Turner

(1998) Assessment of the variability in fMRI data sets due to subject positioning and calibration of the MRI scanner. Neuroimage 7:S599

24.

Karni

Meyer

Jezzard

Adams

Turner

Ungerleider

(1995) Functional MRI evidence for adult motor cortex plasticity during motor skill learning. Nature 377:155–158

25.

Knake

Haag

Hamer

Dittmer

Bien

Oertel

Rosenow

(2003) Language lateralization in patients with temporal lobe epilepsy: a comparison of functional transcranial Doppler sonography and the Wada test. Neuroimage 19:1228–1232

26.

Knecht

Deppe

Ebner

Henningsen

Huber

Jokeit

Ringelstein

E-B

(1998) Non-invasive determination of hemispheric language dominance using functional transcranial Doppler sonography: a comparison with the Wada test. Stroke 29:82–86

27.

Knecht

Deppe

Dräger

Bobe

Lohmann

Ringelstein

Henningsen

(2000) Language lateralization in healthy right handers. Brain 123:74–81

28.

Lehericy

Cohen

Bazin

Samson

Giacomini

Rougetet

Hertz-Pannier

Le Bihan

Marsault

Baulac

(2000) Functional MR evaluation of temporal and frontal language dominance compared with the Wada test. Neurology 54:1625–1633

29.

Loubinoux

Carel

Alary

Boulanouar

Viallard

Manelfe

Rascol

Celsis

Chollet

(2001) Within-session and between-session reproducibility of cerebral sensorimotor activation: a test–retest effect evidenced with functional magnetic resonance imaging. J Cereb Blood Flow Metab 21:592–607

30.

McGonigle

Howseman

Athwal

Friston

Frackowiak

Holmes

(2000) Variability in fMRI: an examination of intersession differences. Neuroimage 11:708–734

31.

McHenry

jr Merory

Bass

Stump

Williams

Witcofski

Howard

Toole

(1978) Xenon-133 inhalation method for regional cerebral blood flow measurements: normal values and test-retest results. Stroke 9:396–399

32.

Musso

Weiller

Kiebel

Muller

Bulau

Rijntjes

(1999) Training-induced brain plasticity in aphasia. Brain 122:1781–1790

33.

Näätänen

(1992) Attention and brain. New York, LEA

34.

Nagata

Uchimura

Hirakawa

Kuratsu

(2001) Method for quantitatively evaluating the lateralization of linguistic function using functional MR imaging. AJNR Am J Neuroradiol 22:985–991

35.

Noll

Genovese

Nystrom

Vazquez

Forman

Eddy

Cohen

(1997) Estimating test-retest reliability in functional MR imaging. II: Application to motor and cognitive activation studies. Magn Reson Med 38:508–517

36.

Oldfield

(1971) The assessment and analysis of handedness: the Edinburgh inventory. Neuropsychologia 9:97–113

37.

Petersen

van Mier

Fiez

Raichle

(1998) The effects of practice on the functional anatomy of task performance. Proc Natl Acad Sci U S A 95:853–860

38.

Pujol

Deus

Losilla

Capdevila

(1999) Cerebral lateralization of language in normal left-handed people studied by functional MRI. Neurology 52:1038–1043

39.

Raichle

(1994) Visualizing the mind. Sci Am 270:58–64

40.

Ramsey

Tallent

van Gelderen

Frank

Moonen

CTW

Weinberger

(1996) Reproducibility of Human 3D fMRI Brain Maps Acquired During a Motor Task. Hum Brain Mapp 4:113–121

41.

Ringelstein

E-B

Kahlscheuer

Niggemeyer

Otis

(1990) Transcranial Doppler sonography: anatomical landmarks and normal velocity values. Ultrasound Med Biol 16:745–761

42.

Rombouts

(1982) The reproducibility of cardiovascular reactions during cognitive tasks. Act Nerv Super Praha Suppl 3:285–294

43.

Rombouts

Barkhof

Hoogenraad

Sprenger

Scheltens

(1998) Within-subject reproducibility of visual activation patterns with functional magnetic resonance imaging using multislice echo planar imaging. Magn Reson Imaging 16:105–113

44.

Rosen

Petersen

Linenweber

Snyder

White

Chapman

Dromerick

Fiez

Corbetta

(2000) Neural correlates of recovery from aphasia after damage to left inferior frontal cortex. Neurology 55:1883–1894

45.

Rutten

Ramsey

van Rijen

van Veelen

(2002) Reproducibility of fMRI-Determined Language Lateralization in Individual Subjects. Brain Lang 80:421–437

46.

Silva

Lee

Yang

Iadecola

Kim

(1999) Simultaneous blood oxygenation level-dependent and cerebral blood flow functional magnetic resonance imaging during forepaw stimulation in the rat. J Cereb Blood Flow Metab 19:871–879

47.

Silvestrini

Caltagirone

Cupini

Matteis

Troisi

Bernardi

(1993) Activation of healthy hemisphere in poststroke recovery. A transcranial Doppler study. Stroke 24:1673–1677

48.

Sitzer

Knorr

Seitz

(1994) Cerebral hemodynamics during sensorimotor activation in humans. J Appl Physiol 77:2804–2811

49.

Springer

Binder

Hammeke

Swanson

Frost

Bellgowan

Brewer

Perry

Morris

Mueller

(1999) Language dominance in neurologically normal and epilepsy subjects: a functional MRI study. Brain 122:2033–2046

50.

Stroobant

Vingerhoets

(2000) Transcranial Doppler ultrasonography monitoring of cerebral hemodynamics during performance of cognitive tasks: a review. Neuropsychol Rev 10:213–231

51.

Tegeler

Strother

Anderson

Kim

(1999) Reproducibility of BOLD-based functional MRI obtained at 4 T. Hum Brain Mapp 7:267–283

52.

Thulborn

Carpenter

Just

(1999) Plasticity of language-related brain function during recovery from stroke. Stroke 30:749–754

53.

Tiecks

Lam

Matta

Strebel

Douville

Newell

(1995) Effects of the valsalva maneuver on cerebral circulation in healthy adults. A transcranial Doppler Study. Stroke 26:1386–1392

54.

Vikingstad

George

Johnson

Cao

(2000) Cortical language lateralization in right handed normal subjects using functional magnetic resonance imaging. J Neurol Sci 175:17–27

55.

Wada

Rasmussen

(1960) Intracarotid injection of sodium amytal for the lateralization of cerebral speech dominance. J Neurosurg 17:266–282

56.

Warach

Gur

Skolnick

Obrist

Reivich

(1992) Decreases in frontal and parietal lobe regional cerebral blood flow related to habituation. J Cereb Blood Flow Metab 12:546–553

57.

Warach

Gur

Skolnick

Obrist

Reivich

(1988) The reproducibility of the ¹³³Xe inhalation technique in resting studies: task order and sex related effects in healthy young adults. J Cereb Blood Flow Metab 7:702–708

58.

Weiller

Isensee

Rijntjes

Huber

Muller

Bier

Dutschka

Woods

Noth

Diener

(1995) Recovery from Wernicke's aphasia: a positron emission tomographic study. Ann Neurol 37:723–732

59.

Weiller

Rijntjes

Isensee

Muller

Kapperler

Huber

Diener

(1993) Right hemispheric contribution to recovery from aphasia: PET in Wernicke-type aphasia. Can J Neurol Sci 20:87

60.

Yetkin

McAuliffe

Cox

Haughton

(1996) Test-retest precision of functional MR in sensory and motor task activation. AJNR Am J Neuroradiol 17:95–98

Task Repetition Can Affect Functional Magnetic Resonance Imaging-Based Measures of Language Lateralization and Lead to Pseudoincreases in Bilaterality

Abstract

Keywords

METHODS

Subject

Functional magnetic resonance imaging.

Word generation.

Functional magnetic resonance imaging data acquisition.

Functional magnetic resonance imaging data analysis.

Functional magnetic resonance imaging lateralization index.

Functional magnetic resonance imaging lateralization index based upon percentage signal change.

Functional transcranial Doppler sonography

Word generation

Data recording.

Data analysis.

Functional transcranial Doppler sonography lateralization index.

Accuracy.

Cardiovascular parameter.

Statistical analysis

RESULTS

Functional magnetic resonance imaging

Functional transcranial Doppler sonography

Accuracy

DISCUSSION

Methodologic considerations

Potential explanations

References