Functional MRI Predicts Memory Performance after Right Mesiotemporal Epilepsy Surgery

Authors


Address correspondence and reprint requests to Dr. F.G. Woermann at MRI Unit, Mara Hospital, Bethel Epilepsy Center, Maraweg 21, Bielefeld 33617, Germany. E-mail: fgw@mara.de

Abstract

Summary: Purpose: Anterior temporal lobe resection (ATR) is a treatment option in drug-resistant epilepsy. An important risk of ATR is loss of memory because mesiotemporal structures contribute substantially to memory function. We investigated whether memory-activated functional MRI (fMRI) can predict postoperative memory loss after anterior temporal lobectomy in right-sided medial temporal lobe epilepsy (MTLE).

Methods: We included 16 patients (10 women) aged 16–54 years. The mean age at epilepsy onset was 12.5 years (range, 1–26 years). The patients' mean Wechsler IQ score was 95.2 (range, 62–125). The activation condition of fMRI consisted of retrieval from long-term memory induced by self-paced performance of an imaginative walk. All but one patient had left-sided speech dominance according to speech-activated fMRI. Outside the scanner, we evaluated the pre- and postoperative visual memory retention by using Rey Visual Design Learning Test.

Results: We found a correlation between the preoperative asymmetry index of memory-fMRI and the change between pre- and postsurgical measures of memory retention. Reduced activation of the mesiotemporal region ipsilateral to the epileptogenic region correlated with a favorable memory outcome after right-sided ATR.

Conclusions: In light of the postoperative results, the theoretical implication of our study is that fMRI based on a simple introspective retrieval task measures memory functions. The main clinical implication of our study is that memory-fMRI might replace the invasive Wada test in MTLE by using a simple fMRI paradigm. Predictive power, however, will be studied in larger patient samples. Other studies are required for left-sided MTLE and neocortical epilepsies to assess the clinical usefulness of memory-fMRI.

Anterior temporal resection (ATR) is a treatment option in drug-resistant epilepsy and leads to complete cessation of seizures in 60–90% of patients (1). An important risk of ATR is loss of memory because mesiotemporal structures contribute substantially to memory function (2). Although the neuropsychological tests (3) and the presence of structural damage of the mesiotemporal structures (4) might predict postoperative memory loss to some degree, the intracarotid amobarbital (Wada) test is still a basic tool for predicting postoperative memory problems (3).

The predictive value of the Wada test is based on asymmetrically low memory functions on the side of surgery (5) or a preserved memory function on the side contralateral to the planned operation (6), which is associated with good postoperative memory outcome. Thus the lateralization of the memory plays a crucial role in predicting postoperative memory outcome (7).

Conversely, the Wada test is an invasive procedure and does not specifically test memory functions of mesiotemporal structures alone but of the whole investigated hemisphere (8). The mesiotemporal structures, especially the hippocampus, may not receive an anterior circulation supply; therefore the Wada test may not anesthetize the relevant regions of interest: the posterior two thirds of the hippocampus is supplied by the posterior circulation (8). Cross-flow to the opposite anterior cerebral artery and variable perfusion of the posterior cerebral artery by the internal cerebral artery also may influence the clinical effects (9). The procedure does not provide sufficient time for reliable memory testing: the evaluation is performed within a short period (3–5 min) under the behavioral effects of anesthesia (aphasia, attention problems, neglect phenomena, somnolence). Furthermore, the Wada test has a low predictive value for postoperative memory outcome in patients with right-sided mesiotemporal lobe epilepsy (MTLE) (10).

Recently some studies using functional magnetic resonance imaging (fMRI) visualized memory functions of the mesiotemporal structures (11–13). In an earlier study, we reported that the activation of mesiotemporal structures during visuospatial memory retrieval was symmetrical in healthy subjects and asymmetrical in patients with MTLE. The activation of the mesiotemporal structures is usually reduced ipsilateral to the epileptogenic region (Fig. 1) (11). Reduced ipsilateral mesiotemporal activation correlated with both abnormal material-specific neuropsychological and Wada tests results (11).

Figure 1.

The activation of the mesiotemporal structures is typically reduced ipsilateral to the epileptogenic region. Structural magnetic resonance imaging (MRI) with a fluid-attenuated inversion recovery (FLAIR) coronal slice of a 18-year-old female patient with right-sided mesiotemporobasal lesion, which was a ganglioglioma on histopathological examination (A). The memory-functional (fMRI) showed bilateral activity in the mesiotemporal structures, but the larger activation appeared on the contralateral (left-sided) mesiotemporal lobe structures (B).

The present fMRI study aimed to investigate whether memory-fMRI is able to predict postoperative memory loss in right-sided MTLE. Normal symmetrical activation of mesiotemporal structures for visuospatial memory retrieval allowed us to use a simple evaluation of fMRI pictures by analyzing the ratio of activated voxels on both sides.

METHODS

We included consecutive right-sided MTLE patients who underwent our adult presurgical evaluation program including continuous video-EEG monitoring. All patients met the following inclusion criteria:

  • 1Epileptogenic lesion in the right hippocampal formation proved by high-resolution MRI
  • 2Right-sided ATR between January 2001 and September 2001.

MRI acquisition

MRI scanning was performed on a 1.5-T scanner (Siemens Magnetom Symphony, Erlangen, Germany) equipped with a standard head coil. Scout and sagittal T1-weighted images were obtained in every subject to position the coronal T*2-weighted images perpendicular to the long axis of the hippocampus. For fMRI, 16 contiguous coronal T*2-weighted images covering the temporal lobes with a slice thickness of 5 mm were obtained by using a standard echo-planar imaging sequence (TR, 1,600 ms; TE, 50 ms; FOV, 192 mm; matrix, 64 × 64). To reduce head motion a vacuum cushion was used.

fMRI task design

The paradigm consisted of 10 activation blocks and 10 baseline blocks. Each block was introduced by spoken commands by using built-in communication devices. The duration of each block was 30 s. During each block, 10 sets of 16 coronal T*2-weighted MR slices were obtained. During the activation block, retrieval from long-term memory without language was induced by self-paced performance of Roland's Hometown Walking task (14). The task was explained to the subjects in detail before scanning. For each subject, an individual hometown walk encompassing 10 destinations was prepared. If a subject had recently moved, then the most familiar hometown was chosen. The walk started either at home or at a well-known central point (e.g., main station). Then the subjects were asked to select a familiar landmark as destination. This landmark served as the starting point for the walk to the next destination. Subjects were instructed to seek destinations within a 300- to 400-m range. After preparation of 10 pairs of starting points and destinations, the complete route was presented to the subject to ensure consistency and comprehension of the words denoting the route. Subjects were asked to navigate mentally through the 10 different routes and to imagine as many details as possible while navigating. Subjects were instructed to look around the destination if they reached the destination before the beginning of the baseline condition. After 30 s, each route was interrupted by the baseline task. The baseline condition consisted of covertly counting odd numbers starting with 21.

Image analysis

On-line image processing was performed by using software provided with the commercially available scanner, which recently has been compared with standard off-line postprocessing software (SPM 99; Wellcome Department of Cognitive Neurology, London, U.K.) (15). The T*2-weighted images were corrected for subject movement by using an algorithm for realignment in k space. Images were smoothed by using a gaussian filter (width 2.0) to prepare statistical comparisons on a voxel-by-voxel basis. Voxel-by-voxel z tests were performed for each subject, identifying average signal-intensity increases as measured during the activation phases compared with the average signal intensity acquired during baseline conditions. The statistical threshold chosen was z > 4; p < 0.00003, for a single activated voxel. Statistical maps showing activated voxels were projected onto echo-planar images of the same patient, thus using images for display purposes with geometric distortions similar to the fMRI data. To perform group data analysis, two investigators blinded to clinical data, except for the right-sided lesion, counted the numbers of voxels in a predefined region of interest over both mesiotemporal regions including the parahippocampal gyrus and the hippocampus. Counting used the crus of the fornix as the posterior starting point and continued anteriorly until no activated voxels were found. The size of this region of interest was 600 to 800 voxels per person. A threshold of z > 4 ensures that the error probability for a single activated voxel within a region of interest of <800 voxels is <0.05. Activated voxels were defined as those voxels that had neighboring activated voxels, that is, at least one adjacent activated voxel in plane and one neighboring activated voxel on at least one adjacent MR slice. Clustering removed small and scattered activated regions that were unlikely to represent genuine brain activation and further reduced the real type I error probability of a z > 4 threshold.

To establish the speech dominance, all patients also had speech-activated fMRI. The methodologic details are already published (16).

For each subject, we calculated an asymmetry index for the mesiotemporal activation: AI = (activated voxels on the left – activated voxels on the right)/all activated voxels.

Thus positive values indicated that the activation was more pronounced on the left than on the right side.

Outside the scanner, we evaluated visual memory retention by using Rey Visual Design Learning Test (RVDLT). The patient's task was to learn 15 geometric patterns during five learning trials. After each trial, the patient had to draw as many patterns as possible. One hour after the learning period, the patient was asked to draw as many patterns as possible (17). The retention of the memorized information was characterized by the ratio consisting of the number of patterns correctly reproduced by the delayed recall divided by the correctly reproduced items during the fifth learning trial. The test was performed before surgery and 6 months postoperatively. Postoperative memory changes were expressed as the difference between postoperative and preoperative retention performance.

RESULTS

Sixteen patients (10 women) aged 16 to 54 years met the inclusion criteria and comprised 14 right-handed and two left-handed patients. The mean age at epilepsy onset was 12.5 years (range, 1–26 years). The patients' mean Wechsler IQ was 95.2 (range, 62–125). Six months after the ATR, 14 (87.5%) patients were seizure free. Histopathologic examination showed hippocampal sclerosis in 12 and low-grade tumors in four cases. All but one patient had left-sided speech dominance according to speech-activated fMRI. Other details are presented in Table 1. Wada tests were not performed because Wada tests for right-sided MTLE have been infrequently used at our center, as in right-sided lesional MTLE, they are rarely informative (10).

Table 1. General data of patients
CaseSexAge at onset (yr)Age (yr)HandednessLesionSeizure typesScalp EEG dataSpeech-fMRI
  1. HS, hippocampal sclerosis; DNT, dysembryoplastic neuroepithelial tumor; CPS, complex partial seizure; GMS, secondarily generalized tonic–clonic seizure; IED, interictal epileptiform discharge; SP, seizure pattern.

 1.F1439RightDNTCPS, GMSRight temporal SP, right temporal IEDLeft-sided speech dominance
 2.F2654RightHSCPS, GMSRight temporal SP, bitemporal independent IED, right and left temporal slow wavesLeft-sided speech dominance
 3.M2444RightHSCPS, GMSRight temporal SP, bitemporal independent IED, right and left temporal slow wavesLeft-sided speech dominance
 4.F637RightHSCPSRight temporal SP, right temporal slow wavesBilateral speech dominance
 5.M1720LeftGangliogliomaCPSRight temporal SP, right temporal IED, right temporal slow wavesLeft-sided speech dominance
 6.F1745RightHSCPSRight temporal SP, right temporal IED, right temporal and generalized slow wavesLeft-sided speech dominance
 7.F1252RightHSCPS, GMSRight temporal SP, right and left temporal slow wavesLeft-sided speech dominance
 8.M1433RightHSCPS, GMSRight temporal SP, bitemporal independent IED, right temporal slow wavesLeft-sided speech dominance
 9.F118RightGangliogliomaCPS, GMSRight temporal SP, bitemporal independent IED, right temporal slow wavesLeft-sided speech dominance
10.M1122RightHSCPS, GMSRight temporal SP, right temporal IED, right temporal slowingLeft-sided speech dominance
11.F1216RightAstrocytoma WHO ICPSRight temporal SP, bitemporal independent IED, right temporal and generalized slow wavesLeft-sided speech dominance
12.M637RightHSCPS, GMSRight temporal SP, right temporal IED, right temporal slowingLeft-sided speech dominance
13.F651LeftHSCPSRight temporal SP, right temporal IED, right temporal slowingLeft-sided speech dominance
14.F1239RightHSCPSRight temporal SP, bitemporal independent IED, right temporal slow wavesLeft-sided speech dominance
15.F523RightHSCPS, GMSRight temporal SP, right temporal IED, right temporal slowingLeft-sided speech dominance
16.M1343RightHSCPSRight temporal SP, bitemporal independent IED, right temporal slow wavesLeft-sided speech dominance

Considering the whole patient group, neither the nonverbal immediate recall (correctly reproduced items during the fifth learning trial) nor the retention rate measured by RVDLT showed significant postoperative changes. Regarding the immediate recall, the patients could remember 9.31 ± 3.8 items on average (median, 9) preoperatively, whereas they could remember 9.0 ± 3.8 items on average (median, 9) after the operation (p = 0.47; Wilcoxon test). The preoperative retention rate was 94.7 ± 15% on average (range, 66.7–125%), whereas after the surgery, it was 100.7 ± 21% on average (range, 66.7–150%). This difference was not significant according to the Wilcoxon test (p = 0.35). On the individual level, however, we found a correlation between preoperative AI on memory-fMRI and the change between pre- and postsurgical measures of memory retention (Fig. 2) by using the Spearman rank correlation test (r = 0.71; p = 0.002). Reduced activation of the mesiotemporal region ipsilateral to the epileptogenic region correlated with a favorable memory outcome after right-sided anterior temporal lobectomy. Fig. 3 demonstrates a patient with larger mesiotemporal activation ipsilateral to the epileptogenic region.

Figure 2.

Relation between preoperative asymmetry indices on memory-functional magnetic resonance imaging [fMRI; activated voxels on the left; (activated voxels on the left−activated voxels on the right)/(all activated voxels)] and the postoperative memory outcome. Reduced activation of the mesiotemporal region ipsilateral to the epileptogenic region predicts favorable memory outcome after right-sided anterior temporal lobectomy. With the Spearman rank correlation test, this association is highly significant (r = 0.71; p = 0.002).

Figure 3.

Preoperative (A) and postoperative (C) structural magnetic resonance imaging (MRI) as well as functional MRI pictures (B) of a 44-year-old male patient with right-sided hippocampal sclerosis (A). The memory-fMRI showed a large activation in the ipsilateral (right-sided) mesiotemporal lobe structures (B). After the mesiotemporal resection (C), the patient showed loss in visual memory retention compared with the preoperative performance.

DISCUSSION

Using Roland's Hometown Walking task, we found that reduced fMRI activation of the mesiotemporal region ipsilateral to the epileptogenic region correlated with favorable nonverbal memory outcome after right-sided ATR. Neither the nonverbal immediate recall nor the retention rate measured by RVDLT showed postoperative changes considering the whole patient group, but some of our patients showed individual changes including postoperative improvement of visual memory function, as reported elsewhere (18,19). All patients in whom the preoperative fMRI showed a larger or equal activation on the side later operated on compared with the contralateral side had a postoperative memory decline (see Fig. 2).

Our study showed that memory-fMRI can predict postoperative memory decline in MTLE. A preliminary study using fMRI with another nonverbal paradigm also suggested that postsurgical memory loss correlated with fMRI activation ipsilateral to temporal lobe resection (20). A recent H215O-positron emission tomography (PET) study using nonverbal and verbal learning tasks for activation found that memory lateralization revealed by PET could predict postsurgical memory outcome after amygdalohippocampectomy (21). Similar to our results, this study found that patients significantly activating to-be-resected ipsilateral mesiotemporal structures during the ipsilateral learning task experienced a postoperative memory decline.

In our study, memory-fMRI findings correlated with individual memory decline measured by a nonverbal memory test after right-sided ATR conducted outside the scanner (10). In light of postoperative results, the theoretical implication of our study is that fMRI based on a simple introspective retrieval task measures memory functions. From a neuropsychological perspective, introspection in general and Roland's home town walk in particular are viewed as common performances in everyday life, but poorly controlled and unspecific memory tests. Visual detail, levels of familiarity, use of verbal strategies, or use of spatial navigation strategies (e.g., egocentric or allocentric) may differ between patients and may go undetected in our experiment. From a clinical perspective, however, a simple fMRI paradigm with the experimental control based on fMRI contrast changes in brain areas implicated in memory functions rather than based on patient interaction is best suited for patients with epilepsy and cognitive impairment (11). The correlation between mesiotemporal fMRI activation with postsurgical memory outcome demonstrated here further characterizes the memory component of Roland's home town walk in a clinically meaningful way.

In the present study, we chose exclusively right-sided MTLE because our fMRI paradigm challenges the nonverbal memory retrieval. In our previous study, using the same paradigm, we found that the right-sided MTLE could be differentiated better from normal control subjects than could left-sided MTLE (11). Right-sided MTLE is a challenging subgroup in predicting memory outcome because the Wada test has a limited role in predicting memory functions after right-sided ATR (10). Moreover, although left-sided MTLE patients are considered to have a high risk for postoperative memory loss, a recent study reviewing previously reported patients with unexpected amnesia after unilateral temporal lobectomy found that three of seven reported patients had had right-sided TLE (2). In the future, however, other studies are required investigating left-sided MTLE and neocortical epilepsies to assess the clinical usefulness of memory-fMRI in predicting the postsurgical memory outcome in general.

One of the limitations of this study is that we used only a single memory test (RVDLT) for measuring postoperative nonverbal memory changes. We chose this test to challenge and compare nonverbal visual memory functions within and outside the fMRI experiment. Cognitive activity required to perform these tasks might be confined to the same mesiotemporal brain structures.

A second limitation of this study may be that we restricted our patient population to the lesional cases; thus we do not know whether our results can be applied to nonlesional right-sided MTLE. Normal MRI findings had been reported as a strong negative predictor for long-term surgical outcome after ATR (22). Thus our center has tended to operate on the lesional MTLE cases (after 1995, <1% of ATRs were performed in patients with normal MRI findings). Conversely, looking at the favorable outcome in lesional MTLE, the preoperative localization of the epileptogenic region was defined with a higher probability in this patient group compared with nonlesional cases.

Binder et al. (23) concluded that it will be difficult to demonstrate that fMRI is equivalent to the Wada test because both of them are nonspecific general procedures rather than standardized tests of the mesiotemporal regions. The variability within the two procedures will lengthen and increase the number of validation procedures. Ultimately, the validation of fMRI should depend on its relation to postoperative memory outcome (23). Thus the main clinical implication of our study is that memory-fMRI might replace the invasive Wada test in MTLE by using a simple fMRI paradigm. Predictive power, however, will be studied in larger patient samples.

Acknowledgments

Acknowledgment:  This work was supported by a grant from the Deutsche Forschungsgemeinschaft (DFG-Eb 111/2-2), from the Society for Epilepsy Research Bielefeld, Germany (Dr. Woermann), and the Humboldt Stiftung (Dr. Janszky).

Ancillary