SEARCH

SEARCH BY CITATION

Keywords:

  • Epilepsy;
  • Magnetic resonance imaging;
  • Postprocessing;
  • Hippocampal sclerosis

Summary

  1. Top of page
  2. Summary
  3. Methods
  4. Results
  5. Discussion
  6. Conclusion
  7. Acknowledgments
  8. Disclosure of Conflict of Interest
  9. References

Purpose

Limbic encephalitis is an autoimmune-mediated disease leading to temporal lobe epilepsy, mnestic deficits, and affective disturbances. Magnetic resonance imaging (MRI) usually shows signal and volume changes of the temporomesial structures. However, these abnormalities may be subtle, thereby hampering the diagnosis by conventional visual assessment. In the present study we evaluated the diagnostic value of a fully automated MRI postprocessing technique in limbic encephalitis and hippocampal sclerosis.

Methods

The MRI postprocessing was based largely on a recently described method allowing for an observer-independent quantification of the fluid-attenuated inversion recovery (FLAIR) signal intensities of amygdala and hippocampus. A 95% confidence region was calculated from the FLAIR intensities of 100 healthy controls. We applied this analysis to the MRI data of 39 patients with antibody-associated limbic encephalitis and 63 patients with hippocampal sclerosis. Moreover, the results were compared to those of visual assessment by an experienced neuroradiologist.

Key Findings

The method detected limbic encephalitis and hippocampal sclerosis with a high sensitivity of 85% and 95%, respectively. The detection rate of the automated approach in limbic encephalitis was significantly superior to visual analysis (85% vs. 51%; p = 0.001), whereas no statistically significant difference for the detection rate in hippocampal sclerosis was found. Patients with limbic encephalitis had significantly higher absolute intensity values of the amygdala and a significantly higher percentage fell outside of the amygdalar confidence region compared to those with hippocampal sclerosis (79% vs. 27%; p < 0.001), whereas we found opposite results in the hippocampal analysis (38% vs. 95%; p < 0.001).

Significance

The FLAIR analysis applied in this study is a powerful tool to quantify signal changes of the amygdala and hippocampus in limbic encephalitis and hippocampal sclerosis. It significantly increases the diagnostic sensitivity in limbic encephalitis in comparison to conventional visual analysis. Furthermore, the method provides an interesting insight into the distinct properties of these two disease entities on MRI, indicating a predominant affection of the amygdala in limbic encephalitis, whereas the affection of the hippocampus is far less pronounced when compared to hippocampal sclerosis.

Limbic encephalitis (LE) was initially described in the 1960s as a clinicopathologic entity with mediotemporal symptoms caused by inflammation in limbic structures in adults (Brierley et al., 1960). Symptoms consist of temporal lobe epilepsy, mnestic deficits, and affective disturbances. In recent years a growing number of autoantibodies has been found to be associated with this disorder. Here, “well characterized” onconeural antibodies (Hu, Yo, Ri, Ma, amphiphysin, collapsin response mediator protein 5), voltage-gated potassium channel (VGKC) antibodies, and glutamic acid decarboxylase (GAD) antibodies are the most frequently detected ones (Vincent et al., 2004; Dalmau & Rosenfeld, 2008; Malter et al., 2010; Irani et al., 2011; Vincent et al., 2011). Moreover, antibodies to the N-methyl-d-aspartate receptor (NMDAR) have also been described to be associated with LE (Dalmau et al., 2007; Novillo-Lopez et al., 2008; Johnson et al., 2010; McCoy et al., 2011), although NMDAR may also lead to a diffuse and multifocal encephalitis with predominant extralimbic manifestations (Dalmau et al., 2008; Niehusmann et al., 2009).

Magnetic resonance imaging (MRI) in LE usually shows hyperintensity and volume changes of amygdala and/or hippocampus. In most cases volume is increased in the acute disease stage (swelling), and it resolves thereafter over time sometimes ending up in atrophy (Urbach et al., 2006). However, MRI changes can be subtle and may be challenging for the neuroradiologist, especially in bilateral LE as the side comparison is hampered in these cases. Furthermore, normal MRI findings after visual assessment have been described, particularly in VGKC-associated LE (VGKC-LE) (Vincent et al., 2004; Irani et al., 2011; van Vliet et al., 2012).

Huppertz et al., (2011) recently introduced an MRI postprocessing analysis tool allowing for a quantification of the fluid-attenuated inversion recovery (FLAIR) signal intensities of the hippocampus. This analysis was capable of detecting hippocampal sclerosis (HS) with a high sensitivity of 97%. The method was based in turn on prior publications of Focke et al., (2008, 2009), who presented an approach for quantitative whole brain analysis of FLAIR scans.

In the present study we modified the FLAIR analysis described by Huppertz et al., thereby allowing for additional intensity analyses of the amygdala. We applied this analysis to patients with antibody-associated LE and HS. The aim was to evaluate (1) the sensitivity of this method in LE and HS, (2) the potential diagnostic benefit of this approach compared to visual analysis by an experienced neuroradiologist, (3) whether the results in LE and HS differ from each other, and (4) whether the results differ within the LE group depending on the associated antibody.

Methods

  1. Top of page
  2. Summary
  3. Methods
  4. Results
  5. Discussion
  6. Conclusion
  7. Acknowledgments
  8. Disclosure of Conflict of Interest
  9. References

Patients and controls

We studied all patients diagnosed with antibody-associated LE and HS presenting at the Department of Epileptology, University of Bonn, from January 1, 2006 to November 30, 2012. This was done retrospectively until 2010 and prospectively thereafter. LE was diagnosed based on the features of (1) limbic signs and symptoms, which manifested themselves in adolescence or adulthood (one or more of the following: seizures of temporal semiology, disturbance of episodic memory, affective disturbances with prominent mood lability, or disinhibition), and (2) presence of serum antibodies associated with LE (onconeural, VGKC, GAD, NMDAR). Patients with clinical and paraclinical evidence of a predominant extralimbic encephalitic manifestation were excluded as the postprocessing is capable of detecting signal changes of only the temporomesial structures.

HS was diagnosed according to (1) MRI criteria after evaluation by an experienced neuroradiologist or (2) histology for those patients who underwent epilepsy surgery. Other potential diagnoses were fully explored and excluded in both groups.

Inclusion criterion for both groups was the presence of a three-dimensional (3D) T1 and 3D FLAIR sequence necessary for the MRI postprocessing (see below). Because the 3D FLAIR sequence is performed routinely in our department since 2010, most patients in both groups were studied prospectively (85% in LE group; 68% in HS group).

The group of controls consisted of 100 healthy subjects with no neurologic disorder and no pathology on brain MRI in the temporomesial region. We did not exclude controls with unspecific extratemporal lesions on MRI (e.g., microangiopathy).

MRI examinations

All included patients underwent routine clinical MRI examinations for the neuroradiologic assessment using a Philips 3 Tesla MRI scanner (Intera, Philips Medical Systems, Amsterdam, The Netherlands) according to a standard protocol (Urbach, 2012). All images were acquired with an angulation oriented along the hippocampus facilitating the detection of temporomesial abnormalities. The clinical scans were evaluated prospectively by an experienced neuroradiologist (HU) for possible epileptogenic lesions. Clinical data (anamnesis, electroencephalography [EEG], and semiology) were provided to the neuroradiologist if available. MRI criteria of HS were atrophy of the temporomesial structures, hyperintensity on T2 and/or FLAIR, and loss of hippocampal lamination; those of LE were temporomesial hyperintensity on T2 and/or FLAIR and volume changes (swelling in most cases).

The 3D T1 and 3D FLAIR volume datasets for the quantitative FLAIR analysis were acquired independently from the clinical scans using a 3 Tesla Siemens scanner (Magnetom Trio, Siemens, Erlangen, Germany). Informed consent was obtained from all patients and healthy subjects participating in the study and the study was approved by the local ethics committee.

MRI postprocessing

Automated quantitative FLAIR analyses were performed by JW on all included patients and controls. The analysis was based largely on the method described recently by Huppertz et al. (2011). However, in contrast to the cited study, our method was based on standard procedures of the freely available software Functional MRI of the Brain Software Library (FSL, Version 4.1, Department of Clinical Neurology, University of Oxford, UK; http://www.fmrib.ox.ac.uk/fsl) (Smith et al., 2004; Jenkinson et al., 2012), whereas Huppertz et al. used the Statistical Parametric Mapping software (SPM5, Wellcome Trust Center for Neuroimaging, London, UK; http://www.fil.ion.ucl.ac.uk/spm). The analysis was fully automated using a bash script and required about 30 min per subject.

Each T1-weighted and FLAIR volume dataset was converted into the compressed Neuroimaging Informatics Technology Initiative (NIfTI-1) format and reoriented to match the orientation of the standard template images in FSL using the fslreorient2std tool. The further analysis included the following steps (see also Fig. 1; the numbers within the figure correspond to the following processing steps):

image

Figure 1. Overview of the MRI processing steps: 1. Skull stripping; 2. Registration of the FLAIR image to the T1-weighted image; 3. Segmentation of the T1 image into gray matter, white matter, and cerebrospinal fluid; 4. Bias correction of the FLAIR image with illustration of the typical bias field of the MRI scanner used in this study; 5. Intensity normalization of the FLAIR image; 6. Subcortical segmentation of the T1 image; 7. Application of the amygdalar and hippocampal masks to the intensity normalized and bias corrected FLAIR image (see text for details).

Download figure to PowerPoint

  1. Skull stripping of the T1 and FLAIR image using the Brain Extraction Tool (BET).
  2. Linear rigid body registration (six degrees of freedom) of the FLAIR image to the T1-weighted image using FMRIB's Linear Image Registration Tool (FLIRT).
  3. Segmentation of the T1-weighted image into gray matter, white matter, and cerebrospinal fluid using FMRIB's Automated Segmentation Tool with its default settings (FAST).
  4. Segmentation of the FLAIR image using FAST with its default settings. This was done to remove the intensity inhomogeneities in the FLAIR image. The FLAIR segmentation results were discarded because they are less precise than the T1 results.
  5. Intensity normalization of the bias corrected FLAIR image. This was done as described by Huppertz et al. (i.e., the mean intensities of the segmented gray and white matter compartments were used as reference for intensity rescaling) with one difference: the FLAIR image was intensity normalized by setting the whole brain intensity average to an arbitrary value of 100 (and not 1,000). The result of this step is an intensity normalized and bias corrected FLAIR image allowing for interindividual analyses.
  6. Subcortical segmentation of the T1-weighted image for creation of the hippocampal and amygdalar masks using FMRIB's Integrated Registration and Segmentation Tool (FIRST). These masks were created on the basis of each individual's T1 image, thereby obviating the need for a spatial normalization of the T1 and FLAIR image and application of a predefined hippocampal and amygdalar mask as it was done by Huppertz et al.
  7. Application of the hippocampal and amygdalar masks to the coregistered and intensity-normalized FLAIR image using the image calculator tool (fslmaths).
  8. Calculation of the mean intensities of the whole volume of amygdala and hippocampus for each side by means of fslmaths. In addition, according to the study of Huppertz et al., the mean of the values of the upper quartile of the intensities was calculated (not shown in Fig. 1). We did not decide beforehand about the whole-volume or upper-quartile analysis for amygdala and hippocampus. Instead, we applied both approaches to both structures of interest in an explorative way to find out which is best suited to differentiate between patients and controls. In the discussion, we provide a detailed explanation for the findings of this study, demonstrating that the whole-volume analysis is more apt for the amygdala, whereas the upper-quartile analysis seems to be better suited for the hippocampus (in line with Huppertz et al.).

Evaluation of results and statistical analysis

The results of controls and patients are shown in a scatter plot with mean FLAIR intensities of right (y-axis) and left (x-axis) amygdala and hippocampus. A 95% confidence region was calculated from the FLAIR intensities of the controls by means of the freely available MATLAB script ‘error ellipse’ (http://www.mathworks.cn/matlabcentral/fileexchange/4705-errorellipse) by AJ Johnson. A patient was classified as postprocessing positive if falling outside of the 95% confidence region.

Statistical analyses were performed using SPSS 19.0 for Windows (IBM, Armonk, NY, U.S.A.). A probability (p) value ≤ 0.05 was regarded as statistically significant using two-tailed tests.

Results

  1. Top of page
  2. Summary
  3. Methods
  4. Results
  5. Discussion
  6. Conclusion
  7. Acknowledgments
  8. Disclosure of Conflict of Interest
  9. References

Study group

According to the above-mentioned inclusion and exclusion criteria, 39 patients with LE and 63 patients with HS were included in the study. Clinical and demographic data of the control group and the two patient groups are summarized in Table 1. The groups did not differ significantly concerning age at MRI (p = 0.123, Kruskal-Wallis test) and seizure frequency (p = 0.488, Mann–Whitney U-test). Patients with LE had a significant later epilepsy onset compared to those with HS (p < 0.001, Mann–Whitney U-test), and the period from epilepsy onset to MRI was significantly shorter in LE compared to HS (p < 0.001, Mann–Whitney U-test). Tumor searches were negative in all LE patients including the two cases with onconeural antibodies (1 × Hu, 1 × amphiphysin; repeated extensive diagnostics were performed in these two patients).

Table 1. Clinical and demographic data of the three study groups
 ControlsLEHS
  1. Two patients with bilateral HS on MRI underwent a unilateral temporal resection.

  2. a

    p < 0.001 (Mann-Whitney U-test).

N (male)100 (47)39 (17)63 (28)
Mean age at MRI, years (range)37.7 (17.2–75.2)40.0 (17.6–76.3)41.6 (8.6–73.0)
Mean age at epilepsy onset, years (range)37.6 (12.9–72.6)a18.9 (0.2–61.1)
MRI after epilepsy onset, mean, years (range)2.4 (0.1–5.9)a22.7 (1.4–50.1)
Mean seizure frequency, seizures per month (range)12.5 (0–300)10.9 (0–300)
Antibody   
VGKC 11 
GAD 18 
NMDAR 8 
Onconeural 2 
Side of HS (histologically confirmed)   
Right  28 (19)
Left  28 (17)
Bilateral  7 (2)

FLAIR analysis in LE and HS

With respect to the amygdala, 31 (79%) of 39 patients with LE and 17 (27%) of 63 patients with HS fell outside of the confidence region (Fig. 2A). This difference was statistically significant (p < 0.001, Fisher's exact test). Five patients (13%) with LE fell into the upper right quadrant of the scatter plot indicating bilateral amygdalar abnormality.

image

Figure 2. Scatter plots showing mean FLAIR intensities of the amygdala (A) and the upper quartile of the hippocampus (B) of patients and controls. The oblique dotted line represents the main diagonal of the confidence region, and the upper right quadrant marks the area that indicates a bilateral abnormality. LE patients predominantly fell outside of the amygdalar confidence region, whereas a predominant hippocampal abnormality was found in the HS cases. Arrowhead in (B) points to the only patient with HS whose MRI was diagnosed as normal by the neuroradiologist (i.e., HS was diagnosed based on histology after epilepsy surgery). In this case the FLAIR analysis did not indicate an abnormality as well.

Download figure to PowerPoint

Referring to the mean intensity of the whole hippocampal volume, we found relatively low detection rates for both groups (LE = 28%, HS = 46%; not shown in Fig. 2). However, when evaluating the mean of the upper quartile of the hippocampal intensities, 60 (95%) of 63 patients with HS fell outside of the confidence region. Right and left HS could be separated invariably by the main diagonal of the confidence region, and all seven patients with bilateral HS fell into the upper right quadrant of the scatter plot indicating bilateral hippocampal abnormality. The portion of patients with LE falling outside of the confidence region also increased (38%) when referring to the mean of the upper quartile of the hippocampal intensities but remained significantly smaller in comparison to the HS group (p < 0.001, Fisher's exact test; Fig. 2B). Confining the analysis to the mean of the upper quartile for the amygdala resulted in a considerably lower detection rate for both LE (38%) and HS (13%) compared to the mean of the complete volume (not shown in Fig. 2).

In the following, all results of the amygdala refer to the mean of the complete volume and all results of the hippocampus refer to the mean of the upper quartile.

When combining the analysis for amygdala and hippocampus, 33 (85%) of 39 patients with LE fell outside of the confidence region (i.e., abnormality of amygdala OR hippocampus). This means that only two patients with LE had an isolated hippocampal abnormality. In the HS group the detection rate did not improve when combining the results in comparison to the hippocampal analysis alone (i.e., no patient with HS showed an isolated amygdalar abnormality).

Patients with LE had significantly higher absolute intensity values of the amygdala compared to the HS group (mean 119.4 vs. 117.5; p < 0.001, Mann-Whitney U-test). Conversely, patients with HS had significantly higher hippocampal intensities in comparison to the LE patients (mean 129.6 vs. 127.3; p < 0.001, Mann-Whitney U-test). The control group had significantly lower values for both amygdala and hippocampus compared to both patient groups (amygdala mean 116.3, hippocampus mean 124.1; p < 0.001, Mann-Whitney U-test).

LE subgroups

We found no significant difference in the detection rate in the amygdalar and hippocampal analysis between the four LE subgroups defined by the antibody findings (p = 0.093 for amygdala; p = 0.057 for hippocampus, Pearson's chi-square test). However, patients with GAD associated LE (GAD-LE) more often fell outside of the amygdalar (94%) and hippocampal (64%) confidence region compared to VGKC-LE (64% amygdala, 27% hippocampus) and NMDAR associated LE (NMDAR-LE; 63% amygdala, 13% hippocampus; Fig. 3). In addition, GAD-LE patients had significantly higher absolute intensity values for the hippocampus (mean 128.1) compared to both VGKC-LE patients (mean 124.5; p = 0.008, Mann-Whitney U-test) and NMDAR-LE patients (mean 126.9; p = 0.006, Mann-Whitney U-test). There was no significant difference concerning the absolute intensity values for the amygdala between the LE subgroups (p = 0.081, Kruskal-Wallis test).

image

Figure 3. Scatter plots of the LE group subclassified by the antibody. Characters outlined in black indicate cases that were diagnosed as normal/unspecific after visual assessment by a neuroradiologist. Arrowheads point to the two LE patients with an isolated hippocampal abnormality. Patient examples illustrate two cases with clearly hyperintense right amygdala (patient 1) and left hippocampus (patient 4), and four cases with subtle MRI changes that fell outside of the amygdalar (patients 2 and 5) and hippocampal confidence region (patients 3 and 6). Patient 2 shows a slight hyperintensity of the left amygdala, and both EEG and neuropsychology indicated a left mesiotemporal focus. Patient 3 had seizures of right temporal origin and neuropsychology showed bitemporal deficits. On MRI, a subtle hyperintensity of the right hippocampus can be seen, which is confirmed by the results of the FLAIR analysis. Patient 5 had right temporal epileptiform discharges on EEG and a slight volume increase and hyperintensity of the right amygdala on MRI, and patient 6 had left temporal epileptiform discharges on EEG compatible with the results of the FLAIR analysis.

Download figure to PowerPoint

Diagnostic benefit

After visual assessment of the clinical MRI scans by an experienced neuroradiologist, 19 patients with LE (49%) were diagnosed as normal or exhibiting unspecific changes that were not considered epileptogenic. As mentioned earlier, the FLAIR analysis was negative in only six patients (15%). This difference was statistically significant (p = 0.001, McNemar test) indicating a significant diagnostic benefit of the MRI postprocessing in LE. Three LE cases falling outside of the confidence region that were diagnosed as normal/unspecific after visual assessment are illustrated in Fig. 3.

For the evaluation of the diagnostic value of the MRI postprocessing in HS, we restricted the analysis to the histologically confirmed patients (N = 38). All but one (97%) of these could be detected by visual analysis. The FLAIR analysis was positive in 36 cases (95%; p = 1.000, McNemar test). The only patient who was negative after visual assessment could not be detected by the FLAIR analysis as well (arrowhead in Fig. 2).

Discussion

  1. Top of page
  2. Summary
  3. Methods
  4. Results
  5. Discussion
  6. Conclusion
  7. Acknowledgments
  8. Disclosure of Conflict of Interest
  9. References

In this study, we evaluated a fully automated MRI postprocessing analysis tool in patients with LE and HS. The approach is a further development of a recently described method allowing for FLAIR signal analyses of amygdala and hippocampus. To our knowledge this is the first study quantifying signal changes in LE in an observer-independent and time-efficient manner. Main findings are discussed in the following.

Methodologic considerations

We found a considerably higher detection rate, particularly in HS (to a far lesser extent also in LE) when confining the hippocampal analysis to the mean of the upper quartile of the voxel intensities. The percentile of 25% was chosen, corresponding to the study of Huppertz et al. (2011), as it has been shown to be efficient in detecting HS. However, we did not observe this effect for the amygdala. A possible explanation for this finding could be that the segmentation of the hippocampal mask may be imprecise, especially in severely atrophied hippocampi. The hippocampus is surrounded mainly by cerebrospinal fluid, which is dark and therefore shows low intensity values on FLAIR images. Although cerebrospinal fluid is masked during the postprocessing, voxels located at the outside margin of the hippocampus at the transition between tissue and cerebrospinal fluid exhibit relatively low intensity values, probably owing to partial volume effects. These low intensity voxels may lead to a substantial reduction of the mean intensity value. Because of the hippocampal atrophy and the more imprecise segmentation, the relative share of these low intensity voxels is higher in HS compared to the healthy controls and the LE group and therefore leads to a (false) greater reduction of the mean intensity in HS. Confining the analysis to the upper quartile excludes these low intensity voxels from the analysis, resulting in a better discrimination especially of the HS group from the controls. In contrast to the hippocampus, the amygdala is surrounded by cerebrospinal fluid to a far lesser extent. Furthermore, amygdalar volume usually is increased, especially in the early stage of LE (Urbach et al., 2006; Bien et al., 2007), and severe amygdalar atrophy is uncommon even in HS (Bernasconi et al., 2003; Mitsueda-Ono et al., 2011). These facts may explain why we did not observe the above-mentioned effect for the amygdala.

Results in LE and HS

The FLAIR analysis detected HS with a high sensitivity of 95%, which is comparable to the previous results reported by Huppertz et al. (2011), who found a sensitivity of 97%. In line with the cited study, our method could invariably separate right and left HS and allowed for the identification of bilateral HS in all cases. Furthermore, taking the analysis of the amygdala into account, our results indicate that HS is an entity that primarily affects the hippocampus, as we found an amygdalar abnormality in only 27% of the HS patients. No HS patient showed an isolated abnormality of the amygdala. Nevertheless, patients with HS showed significantly higher absolute intensity values of the amygdala compared to the controls. This indicates that the amygdala is also affected in HS but to a considerably lower extent compared to the hippocampus. These results are in accordance with previous studies that investigated signal changes of the amygdala in patients with HS by means of T2 relaxometry. Bartlett et al., (2002) found an abnormality of the T2 relaxation time of the amygdala in 44% of a cohort of 25 patients with HS compared to a control group. Two previous studies by Briellmann et al. reported about a slight increase of the T2 relaxation times of the amygdala in HS patients compared to healthy controls (about 4% higher values) (Briellmann et al., 2004, 2007). In contrast to this, the values of the sclerotic hippocampus were 20% higher than those of the control group, thus supporting the findings of the present study.

The FLAIR analysis was also capable of detecting LE with a high sensitivity of 85%. Furthermore, our results indicate that LE, in contrast to HS, is a disease that primarily affects the amygdala, as we found an amygdalar abnormality in 79%, whereas a hippocampal abnormality was present in only 38% of the LE patients. Consistent with this hypothesis, LE patients had significantly higher absolute intensity values of the amygdala compared to those with HS, whereas we found opposite results for the hippocampus (i.e., significantly higher hippocampal values in HS compared to LE). We did not find any studies in the literature that quantified signal changes of the temporomesial structures on MRI in LE (for example, by means of T2 relaxometry). However, there is evidence of an amygdalar affection in LE based on both histopathologic and MRI studies. Histopathologic changes comprise perivascular and parenchymal infiltration of B and T lymphocytes and microglial activation (Vincent et al., 2004; Khan et al., 2009; Bien et al., 2012). MRI studies found a hyperintensity and swelling of the amygdala in the majority of patients with LE by conventional visual inspection (Provenzale et al., 1998; Urbach et al., 2006; Demaerel et al., 2011; Quek et al., 2012). The finding of a predominant affection of the amygdala in LE may explain that these patients suffer from affective disturbances more frequently compared to those with HS, as the amygdala is known to play a key role in the spectrum of mood disorders (Blond et al., 2012; Strakowski et al., 2012).

Because a considerable share of the LE group was diagnosed and imaged during the acute/subacute disease stage (33% within first 12 months after disease onset) and, in contrast to this, the MRI of the HS group was performed in a chronic setting, recent seizure activity can be a confounder of our results as it may lead to signal changes, especially of the temporomesial structures (Chan et al., 1996). However, there was no difference in the seizure frequency between these two groups at the time of the MRI, militating against a relevant influence of the seizure activity on the results presented here.

LE subgroups

Our comparison of the LE subgroups found significantly higher absolute intensity values for the hippocampus in GAD-LE compared to VGKC-LE and NMDAR-LE. These findings may be an indicator of a more chronic disease course in GAD-LE. Complementing these results, Malter et al. (2010) and Quek et al. (2012) reported about a worse mnestic and seizure outcome in GAD-LE compared to VGKC-LE, which may be attributable to a more severe and chronic affection of the hippocampus. Outcome in NMDAR-LE is good in most cases (Novillo-Lopez et al., 2008; Pruss et al., 2010; McCoy et al., 2011; Titulaer et al., 2013) and seems to be superior to GAD-LE, although studies comparing these two disease entities are lacking.

Diagnostic benefit

Although we found a high sensitivity of 95% in detecting HS, there was no diagnostic benefit to the automated FLAIR analysis in comparison to visual assessment by a specialized neuroradiologist, who detected 97% of the histologically confirmed HS cases. These results indicate that visual detection of HS on high quality 3 Tesla MRI, as it was performed in this study, is unproblematic in most cases, assuming that the evaluation is done by an expert reader. This finding may be a result of the well-defined morphology of HS on MRI (atrophy and hyperintensity of the hippocampus). Furthermore, as HS is unilateral in most cases, the comparison with the contralateral hippocampus additionally facilitates the visual detection. Nevertheless, FLAIR analysis may be useful in excluding an affection of the contralateral hippocampus in visually diagnosed unilateral HS especially when epilepsy surgery is considered, thereby allowing for a better estimation of the chances for a sustained postoperative seizure freedom. Further potential benefits of this method in HS are discussed in the publication of Huppertz et al. (2011).

We found a significant diagnostic benefit of the FLAIR analysis compared to conventional visual assessment in LE. The sensitivity of the visual analysis in LE was considerably lower compared to HS, indicating that the visual detection and correct interpretation of this entity on MRI is more challenging for the neuroradiologist. This finding may be due to the more variable presentation of this disease on MRI. Abnormalities are often bilateral hampering the comparison with the contralateral (in HS mostly normal) side. Signal alterations may affect amygdala and/or hippocampus and may frequently be subtle, whereas HS in most cases is restricted to hippocampal signal changes and a marked hyperintensity is present in the majority of cases. Volume can be increased, decreased (especially in later disease stages), or even normal in LE, whereas marked atrophy can easily be detected in many HS cases. All these points make LE more challenging for the neuroradiologist and therefore impede the correct interpretation. Consistent with our findings, Irani et al. (2011) reported about 26 patients with VGKC-LE of whom 46% had a normal MRI after visual assessment. The FLAIR analysis applied in this study seems to be a powerful tool as it increases significantly the diagnostic sensitivity of MRI in LE. Furthermore, it may lead to an earlier identification of this disease because pathologic MRI features can be used as markers prompting antibody determination. This may lead to a better outcome owing to earlier initiation of antiinflammatory therapy. In addition, the FLAIR analysis may also be suitable for monitoring the temporal evolution in LE and facilitate the evaluation of the effects of anti-inflammatory therapy. Further studies are needed to prove these hypotheses.

Limitations

As reported by Huppertz et al. (2011), the method applied in this study is scanner-specific and sequence-specific. This means that each scanner requires its own separate norm database of healthy controls.

Furthermore, the analysis is not specific to HS and LE, as other diseases leading to signal changes of the temporomesial structures may be detected by this method (e.g., tumor, dysplasia).

Signal changes remote from the temporomesial structures cannot be detected with our approach. This was the case in 13% of our LE group according to visual assessment (50% onconeural, 0% VGKC, 13% NMDAR, 17% GAD). Studies report extralimbic signal changes in 25–39% in classical paraneoplastic LE (Gultekin et al., 2000; Lawn et al., 2003), whereas this seems to be less frequent in nonparaneoplastic LE associated with GAD (0–20%) or VGKC (0–14%), which complements our results (Vincent et al., 2004; Urbach et al., 2006; Malter et al., 2010; Irani et al., 2011; Quek et al., 2012). Referring to NMDAR encephalitis in general (limbic and extralimbic manifestations), 35–50% show extralimbic lesions on MRI (Dalmau et al., 2007, 2008). However, when restricting the evaluation to NMDAR-LE, extralimbic manifestations seem to be considerably less common (0–33%) (Dalmau et al., 2007; Novillo-Lopez et al., 2008; Johnson et al., 2010; McCoy et al., 2011).

Finally, our results are limited to a subset of antibody-associated LE, given that no patients with antibodies to alpha-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA), γ-aminobutyric acid (GABA)B and metabotropic glutamate receptors were admitted to our hospital in the time of this study.

Conclusion

  1. Top of page
  2. Summary
  3. Methods
  4. Results
  5. Discussion
  6. Conclusion
  7. Acknowledgments
  8. Disclosure of Conflict of Interest
  9. References

The FLAIR analysis applied in this study offers a fully automated, observer-independent, and time-efficient method for quantifying signal changes of the amygdala and hippocampus in patients with LE and HS. Our results indicate a significant diagnostic benefit for this method compared to conventional visual analysis in LE. Furthermore, our approach gives an interesting insight into the distinct properties of LE and HS on MRI, indicating a predominant affection of the amygdala in LE, whereas the affection of the hippocampus is far less pronounced than in HS.

Acknowledgments

  1. Top of page
  2. Summary
  3. Methods
  4. Results
  5. Discussion
  6. Conclusion
  7. Acknowledgments
  8. Disclosure of Conflict of Interest
  9. References

JW and JCS were supported by the Gerok Program of the BONFOR commission, University of Bonn. JCS was furthermore supported by the Research Cluster on Mesial Temporal Lobe Epilepsy (SFB TR3) by Deutsche Forschungsgemeinschaft (DFG). BW received a grant as part of the Heisenberg Program of DFG (WE 4427/3-1). HJH is supported by the Swiss Epilepsy Foundation. The study was additionally partly supported by the SFB TR3 projects A1 and A8 of the DFG.

Disclosure of Conflict of Interest

  1. Top of page
  2. Summary
  3. Methods
  4. Results
  5. Discussion
  6. Conclusion
  7. Acknowledgments
  8. Disclosure of Conflict of Interest
  9. References

Prof. Elger received grants from the DFG and from the speaker's bureaus of UCB and Desitin. The remaining authors report no potential conflicts of interest. 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.

References

  1. Top of page
  2. Summary
  3. Methods
  4. Results
  5. Discussion
  6. Conclusion
  7. Acknowledgments
  8. Disclosure of Conflict of Interest
  9. References
  • Bartlett PA, Richardson MP, Duncan JS. (2002) Measurement of amygdala T2 relaxation time in temporal lobe epilepsy. J Neurol Neurosurg Psychiatry 73:753755.
  • Bernasconi N, Bernasconi A, Caramanos Z, Antel SB, Andermann F, Arnold DL. (2003) Mesial temporal damage in temporal lobe epilepsy: a volumetric MRI study of the hippocampus, amygdala and parahippocampal region. Brain 126:462469.
  • Bien CG, Urbach H, Schramm J, Soeder BM, Becker AJ, Voltz R, Vincent A, Elger CE. (2007) Limbic encephalitis as a precipitating event in adult-onset temporal lobe epilepsy. Neurology 69:12361244.
  • Bien CG, Vincent A, Barnett MH, Becker AJ, Blumcke I, Graus F, Jellinger KA, Reuss DE, Ribalta T, Schlegel J, Sutton I, Lassmann H, Bauer J. (2012) Immunopathology of autoantibody-associated encephalitides: clues for pathogenesis. Brain 135:16221638.
  • Blond BN, Fredericks CA, Blumberg HP. (2012) Functional neuroanatomy of bipolar disorder: structure, function, and connectivity in an amygdala-anterior paralimbic neural system. Bipolar Disord 14:340355.
  • Briellmann RS, Jackson GD, Pell GS, Mitchell LA, Abbott DF. (2004) Structural abnormalities remote from the seizure focus: a study using T2 relaxometry at 3 T. Neurology 63:23032308.
  • Briellmann RS, Hopwood MJ, Jackson GD. (2007) Major depression in temporal lobe epilepsy with hippocampal sclerosis: clinical and imaging correlates. J Neurol Neurosurg Psychiatry 78:12261230.
  • Brierley JB, Corsellis JAN, Hierons R, Nevin S. (1960) Subacute encephalitis of later adult life mainly affecting the limbic areas. Brain 83:357368.
  • Chan S, Chin SS, Kartha K, Nordli DR, Goodman RR, Pedley TA, Hilal SK. (1996) Reversible signal abnormalities in the hippocampus and neocortex after prolonged seizures. AJNR 17:17251731.
  • Dalmau J, Rosenfeld MR. (2008) Paraneoplastic syndromes of the CNS. Lancet Neurol 7:327340.
  • Dalmau J, Tuzun E, Wu HY, Masjuan J, Rossi JE, Voloschin A, Baehring JM, Shimazaki H, Koide R, King D, Mason W, Sansing LH, Dichter MA, Rosenfeld MR, Lynch DR. (2007) Paraneoplastic anti-N-methyl-D-aspartate receptor encephalitis associated with ovarian teratoma. Ann Neurol 61:2536.
  • Dalmau J, Gleichman AJ, Hughes EG, Rossi JE, Peng X, Lai M, Dessain SK, Rosenfeld MR, Balice-Gordon R, Lynch DR. (2008) Anti-NMDA-receptor encephalitis: case series and analysis of the effects of antibodies. Lancet Neurol 7:10911098.
  • Demaerel P, Van Dessel W, Van Paesschen W, Vandenberghe R, Van Laere K, Linn J. (2011) Autoimmune-mediated encephalitis. Neuroradiology 53:837851.
  • Focke NK, Symms MR, Burdett JL, Duncan JS. (2008) Voxel-based analysis of whole brain FLAIR at 3T detects focal cortical dysplasia. Epilepsia 49:786793.
  • Focke NK, Bonelli SB, Yogarajah M, Scott C, Symms MR, Duncan JS. (2009) Automated normalized FLAIR imaging in MRI-negative patients with refractory focal epilepsy. Epilepsia 50:14841490.
  • Gultekin SH, Rosenfeld MR, Voltz R, Eichen J, Posner JB, Dalmau J. (2000) Paraneoplastic limbic encephalitis: neurological symptoms, immunological findings and tumour association in 50 patients. Brain 123(Pt 7):14811494.
  • Huppertz HJ, Wagner J, Weber B, House P, Urbach H. (2011) Automated quantitative FLAIR analysis in hippocampal sclerosis. Epilepsy Res 97:146156.
  • Irani SR, Michell AW, Lang B, Pettingill P, Waters P, Johnson MR, Schott JM, Armstrong RJ, Zagami A, Bleasel A, Somerville ER, Smith SM, Vincent A. (2011) Faciobrachial dystonic seizures precede Lgi1 antibody limbic encephalitis. Ann Neurol 69:892900.
  • Jenkinson M, Beckmann CF, Behrens TE, Woolrich MW, Smith SM. (2012) Fsl. Neuroimage 62:782790.
  • Johnson N, Henry C, Fessler AJ, Dalmau J. (2010) Anti-NMDA receptor encephalitis causing prolonged nonconvulsive status epilepticus. Neurology 75:14801482.
  • Khan NL, Jeffree MA, Good C, Macleod W, Al-Sarraj S. (2009) Histopathology of VGKC antibody-associated limbic encephalitis. Neurology 72:17031705.
  • Lawn ND, Westmoreland BF, Kiely MJ, Lennon VA, Vernino S. (2003) Clinical, magnetic resonance imaging, and electroencephalographic findings in paraneoplastic limbic encephalitis. Mayo Clin Proc 78:13631368.
  • Malter MP, Helmstaedter C, Urbach H, Vincent A, Bien CG. (2010) Antibodies to glutamic acid decarboxylase define a form of limbic encephalitis. Ann Neurol 67:470478.
  • McCoy B, Akiyama T, Widjaja E, Go C. (2011) Autoimmune limbic encephalitis as an emerging pediatric condition: case report and review of the literature. J Child Neurol 26:218222.
  • Mitsueda-Ono T, Ikeda A, Inouchi M, Takaya S, Matsumoto R, Hanakawa T, Sawamoto N, Mikuni N, Fukuyama H, Takahashi R. (2011) Amygdalar enlargement in patients with temporal lobe epilepsy. J Neurol Neurosurg Psychiatry 82:652657.
  • Niehusmann P, Dalmau J, Rudlowski C, Vincent A, Elger CE, Rossi JE, Bien CG. (2009) Diagnostic value of N-methyl-D-aspartate receptor antibodies in women with new-onset epilepsy. Arch Neurol 66:458464.
  • Novillo-Lopez ME, Rossi JE, Dalmau J, Masjuan J. (2008) Treatment-responsive subacute limbic encephalitis and NMDA receptor antibodies in a man. Neurology 70:728729.
  • Provenzale JM, Barboriak DP, Coleman RE. (1998) Limbic encephalitis: comparison of FDG PET and MR imaging findings. Am J Roentgenol 170:16591660.
  • Pruss H, Dalmau J, Harms L, Holtje M, Ahnert-Hilger G, Borowski K, Stoecker W, Wandinger KP. (2010) Retrospective analysis of NMDA receptor antibodies in encephalitis of unknown origin. Neurology 75:17351739.
  • Quek AM, Britton JW, McKeon A, So E, Lennon VA, Shin C, Klein C, Watson RE Jr, Kotsenas AL, Lagerlund TD, Cascino GD, Worrell GA, Wirrell EC, Nickels KC, Aksamit AJ, Noe KH, Pittock SJ. (2012) Autoimmune epilepsy: clinical characteristics and response to immunotherapy. Arch Neurol 69:582593.
  • Smith SM, Jenkinson M, Woolrich MW, Beckmann CF, Behrens TE, Johansen-Berg H, Bannister PR, De Luca M, Drobnjak I, Flitney DE, Niazy RK, Saunders J, Vickers J, Zhang Y, De Stefano N, Brady JM, Matthews PM. (2004) Advances in functional and structural MR image analysis and implementation as FSL. Neuroimage 23(Suppl. 1):S208S219.
  • Strakowski SM, Adler CM, Almeida J, Altshuler LL, Blumberg HP, Chang KD, DelBello MP, Frangou S, McIntosh A, Phillips ML, Sussman JE, Townsend JD. (2012) The functional neuroanatomy of bipolar disorder: a consensus model. Bipolar Disord 14:313325.
  • Titulaer MJ, McCracken L, Gabilondo I, Armangue T, Glaser C, Iizuka T, Honig LS, Benseler SM, Kawachi I, Martinez-Hernandez E, Aguilar E, Gresa-Arribas N, Ryan-Florance N, Torrents A, Saiz A, Rosenfeld MR, Balice-Gordon R, Graus F, Dalmau J. (2013) Treatment and prognostic factors for long-term outcome in patients with anti-NMDA receptor encephalitis: an observational cohort study. Lancet Neurol 12:157165.
  • Urbach H. (2012) High-field magnetic resonance imaging for epilepsy. Neuroimaging Clin N Am 22:173189, ix-x.
  • Urbach H, Soeder BM, Jeub M, Klockgether T, Meyer B, Bien CG. (2006) Serial MRI of limbic encephalitis. Neuroradiology 48:380386.
  • van Vliet J, Mulleners W, Meulstee J. (2012) EEG leading to the diagnosis of limbic encephalitis. Clin EEG Neurosci 43:161164.
  • Vincent A, Buckley C, Schott JM, Baker I, Dewar BK, Detert N, Clover L, Parkinson A, Bien CG, Omer S, Lang B, Rossor MN, Palace J. (2004) Potassium channel antibody-associated encephalopathy: a potentially immunotherapy-responsive form of limbic encephalitis. Brain 127:701712.
  • Vincent A, Bien CG, Irani SR, Waters P. (2011) Autoantibodies associated with diseases of the CNS: new developments and future challenges. Lancet Neurol 10:759772.