Skip to main content
NCBI home page
Neurology logoLink to Neurology
. 2022 Jan 11;98(2):e141–e151. doi: 10.1212/WNL.0000000000013033

Association Between Anatomical Location of Surgically Induced Lesions and Postoperative Seizure Outcome in Temporal Lobe Epilepsy

Ezequiel Gleichgerrcht 1,✉, Daniel L Drane 1, Simon S Keller 1, Kathryn A Davis 1, Robert Gross 1, Jon T Willie 1, Nigel Pedersen 1, Christophe de Bezenac 1, Jens Jensen 1, Ruben Kuzniecky 1, Leonardo Bonilha 1
PMCID: PMC8762583  PMID: 34716254

Abstract

Background and Objectives

To determine the association between surgical lesions of distinct gray and white structures and connections with favorable postoperative seizure outcomes.

Methods

Patients with drug-resistant temporal lobe epilepsy (TLE) from 3 epilepsy centers were included. We employed a voxel-based and connectome-based mapping approach to determine the association between favorable outcomes and surgery-induced temporal lesions. Analyses were conducted controlling for multiple confounders, including total surgical resection/ablation volume, hippocampal volumes, side of surgery, and site where the patient was treated.

Results

The cohort included 113 patients with TLE (54 women; 86 right-handed; mean age at seizure onset 16.5 years [SD 11.9]; 54.9% left) who were 61.1% free of disabling seizures (Engel Class 1) at follow-up. Postoperative seizure freedom in TLE was associated with (1) surgical lesions that targeted the hippocampus as well as the amygdala–piriform cortex complex and entorhinal cortices; (2) disconnection of temporal, frontal, and limbic regions through loss of white matter tracts within the uncinate fasciculus, anterior commissure, and fornix; and (3) functional disconnection of the frontal (superior and middle frontal gyri, orbitofrontal region) and temporal (superior and middle pole) lobes.

Discussion

Better postoperative seizure freedom is associated with surgical lesions of specific structures and connections throughout the temporal lobes. These findings shed light on the key components of epileptogenic networks in TLE and constitute a promising source of new evidence for future improvements in surgical interventions.

Classification of Evidence

This study provides Class II evidence that for patients with TLE, postoperative seizure freedom is associated with surgical lesions of specific temporal lobe structures and connections.

Surgery for temporal lobe epilepsy (TLE) can be curative.1,2 Nonetheless, 30%–40% of patients with TLE continue to experience seizures with loss of awareness after surgery.1,3,4 Seizures with loss of awareness are directly associated with a lower quality of life.5 The reasons for suboptimal surgical outcomes are complex and many studies have investigated presurgical clinical factors that may be associated with seizure freedom. Consistently, good outcomes are most frequent among patients with unilateral signs of underlying hippocampal sclerosis on MRI.6-8 Other factors associated with favorable outcomes include shorter duration of epilepsy, younger age at surgery,9 history of febrile seizures,10 unilateral,8,11 and infrequent12 interictal epileptiform discharges. Nonetheless, even when combined, these predictive factors only partially explain postoperative outcomes, and many patients who are expected to achieve optimal results continue to have seizures after surgery.

Beyond clinical variables, the location and extent of the anatomical resection have been deemed to be a crucial determinant of outcome, with hippocampal removal being necessary for seizure freedom.13-16 However, recent studies have challenged this notion17,18 and other medial temporal structures may be similarly important for outcomes, such as the entorhinal19,20 and piriform21 cortices, which can independently generate or sustain seizures in TLE.22 Furthermore, many recent whole-brain neuroimaging studies have indicated that the structural and functional connectivity of the medial temporal lobe can also support seizures in TLE.23-31

Voxel-based lesion-mapping neuroimaging methods have been extensively employed and refined to assess the quantitative relationship between brain injury and clinical variables outside of epilepsy.32,33 When a lesion to a specific brain region is statistically associated with a clinical variable, it constitutes a rigorous demonstration that the region is critical for that function.32 In this context, a lesion-based approach is optimally suited to elucidate the relationship between finer anatomical landmarks in TLE resections and outcomes. Moreover, our group has expanded conventional lesion-based approaches to include structural connectivity lesion mapping,34 which can greatly elucidate the anatomy of medial temporal network disconnections and their importance towards seizure freedom.

Based on the evidence supporting the role of the medial temporal cortical structures in seizure onset in TLE,19-21 we hypothesized that resection of the entorhinal and amygdala–piriform regions would be associated with favorable postoperative outcomes. We further hypothesized that medial temporal structural and functional disconnections are also independently associated with better surgical outcomes. In this study, we employed a multimodal quantitative lesion–clinical approach to test these hypotheses leveraging a large dataset from multiple epilepsy centers and a combination of voxel-based outcome–resection mapping combined with structural and functional connectivity outcome–lesion mapping.

Methods

Standard Protocol Approval, Registrations, and Patient Consents

Data gathering followed institutional review board (IRB) approval at each site. The majority of patients had been enrolled prospectively through informed consent for the use of anonymized neuroimaging and clinical data. The IRB at the Medical University of South Carolina approved the use of retrospective analysis of anonymized multicentric data.

Patients

We included data from 113 patients with TLE from 3 different epilepsy centers: Medical University of South Carolina (MUSC; n = 10), Emory University (Emory; n = 59), and Bonn University (Bonn; n = 44). Patients were included at each site following local IRB protocols. Patients were diagnosed with unilateral TLE following the criteria proposed by the International League Against Epilepsy (ILAE).35 Only patients who had normal MRI or findings suggestive of underlying medial temporal sclerosis were included in the study; that is, we did not include patients who had other lesions (e.g., tumors, cavernoma). All patients were refractory to pharmacologic treatment, prompting presurgical evaluation at their site, and were referred to epilepsy surgery based on expert consensus. The choice between the type of procedure (open resection or laser ablation) depended on a combination of factors including the availability of laser ablation at the time and place of the surgery referral, suiting the size of the procedure to the putative or demonstrated early ictal network, and patient preference. A subset of patients from this study was studied as part of prior multisite studies from our group.26

Surgical Outcome

Seizure control outcomes were quantified at least 1 year after the procedure by an epileptologist based on the Engel outcome scale36 where Engel 1 indicates freedom from disabling seizures (i.e., without loss of consciousness), Engel 2 indicates significant improvement in seizures with persistence of rare disabling seizures, Engel 3 indicates worthwhile improvement, and Engel 4 indicates no changes. If patients underwent more than 1 procedure (e.g., resection following ablation), only data related to the first surgical procedure were included in this study. For the majority of patients, imaging included for the purposes of this analysis was obtained within a 2-month window of the clinical assessment of seizure outcomes.

Neuroimaging Acquisition and Processing

T1-weighted isotropic volumetric images were acquired at each site with the following measures:

  • MUSC: Siemens Skyra 3T scanner, isotropic voxel size 1 mm, 12-channel head coil, repetition time (TR) 2,050–2,250 ms, echo time (TE) 2.5–3 ms, field of view 256–320 mm, flip angle 10°

  • Emory: Siemens Skyra 3T scanner, isotropic voxel size 0.8 mm, 12-channel head coil, TR 2,300 ms, TE 2.75 ms, inversion time (TI) 1,100 ms, flip angle 8°

  • Bonn: Siemens Magnetom Trio 3T scanner, 8-channel head coil, isotropic voxel size of 1 mm, TR 1,300 ms, TE 3.97 ms, TI 650 ms, flip angle 10°

Individual voxels were rearranged by flipping sides so that voxels represented ipsilateral vs contralateral (to the side of surgery) areas rather than a left vs right dichotomy. For example, for a patient with left-sided TLE, voxels on the left hemisphere became ipsilateral, whereas voxels in the right hemisphere were considered contralateral. We arbitrarily chose to represent ipsilateral findings on the left side of the brain.

We employed 3 different and complementary approaches to define the relationship between surgical cavity and clinical outcomes: voxel-based lesion–outcome mapping, which is based on conventional voxel-based lesion–symptom mapping from classical cognitive neuroscience studies,32 as well as structural connectivity lesion–outcome mapping and functional connectivity lesion–outcome mapping. All analyses were performed based on postoperative high-resolution volumetric T1-weighted imaging acquired at least 12 months after surgery to ensure complete or near-complete resolution of edema and gliosis. Structural and functional connectivity analyses were calculated from the T1-defined surgical cavity using high-volume normative databases of brain connectivity, from which projected connectivity loss from the resection was calculated. These steps are summarized in the following.

Definition of the Surgical Cavity and Spatial Normalization

Manual tracing of the surgical cavity on all postoperative T1-weighted scans was performed by a blinded rater trained on neuroanatomy with emphasis on medial temporal lobe anatomy37 using the open source software MRIcroGL (nitrc.org/projects/mricrogl). All surgical delineations were reviewed by the first author (E.G.). We employed an enantiomorphic approach described by Nachev et al.38 implemented into the software SPM39 by the NiiStat toolbox developed by Chris Rorden (nitrc.org/projects/niistat/) to normalize brain images and surgical masks into standard space (see details below).

Assessment of Structural Connectivity

We measured structural connectivity as the predicted structural network damage from the surgical cavity on a normative structural white matter connectogram derived from the diffusion MRI Human Connectome Project template HCP842 with 1 mm resolution,40 a map of diffusion tensor imaging pathways in Montreal Neurological Institute (MNI) 152 space consistent across a large number of healthy adult individuals. For each patient with TLE, we assessed which connectogram fibers would have been affected by the surgical cavity by evaluating the overlap between voxels in the surgical cavity in stereotaxic space and each fiber in the connectogram. The result was a map of lesioned fibers (a “disconnectogram”) per individual, from which a tract density image was created, where each voxel in MNI 152 space represented a continuous integer denoting the number of fibers affected by the surgical resection at that voxel.

Assessment of Functional Connectivity

Similarly, we measured functional connectivity based on the predicted functional network affected by the surgical cavity using a large normative meta-analytical functional MRI database. Specifically, using the surgical cavity in MNI space as the region of interest (ROI), we used the Neurosynth meta-analytical data and code to calculate the ROI's coactivation. Neurosynth is a seed-based approach that generates a whole brain map of areas with reported coactivation with a given ROI across a large database including 14,371 functional MRI studies at the time of this study.41 We employed an false discovery rate–corrected threshold of p < 0.01, resulting in a voxel-wise whole brain map per patient where each voxel indicated the potential loss of connectivity from the surgical cavity represented as a Z score float variable. This step was performed using the Neurosynth coactivation function and in-house developed Python and MATLAB scripts for image coregistration.

Steps A, B, and C are shown for one representative patient in eFigure 1, links.lww.com/WNL/B651.

Presurgical Hippocampal Atrophy as a Covariate Measure

Presurgical hippocampal atrophy is an important confounder in TLE. For example, patients with less prominent degree of preoperative hippocampal atrophy would require more hippocampal tissue resected. As such, an association between higher hippocampal volume resection and outcomes would not be truly related to the resection per se, but merely to the presence or absence of hippocampal atrophy. To account for this important confounder, we used voxel-based morphometry (VBM) to measure presurgical hippocampal volumes/concentration, which were used as control variables in the statistical analyses described above. VBM was computed using the software package SPM with iterative normalization and segmentation using the modulated normalization option.42 The resulting probabilistic gray matter map was segmented into ROIs based on the automatic anatomical labeling (AAL) atlas43 and the hippocampal volumes ipsilateral and contralateral to the side of resection were derived.

Statistical Analyses

This study sought to investigate the association of postoperative seizure outcomes with surgical lesions of specific temporal lobe structures and connections (Class II evidence). In order to achieve this, the following analyses were conducted.

Lesion and Disconnection

Voxel-Based Outcome–Resection Mapping

Each surgical resection mask is a binary map, where a voxel is coded as 1 if lesioned or 0 if intact. We performed a voxel-based general linear model with surgical outcome as the ranked dependent variable, and with independent variables including the presence or absence of lesion, the total surgical resection volume, the hippocampal volumes, the side of surgery, and the site where the patient was treated (the latter two encoded as dummy variables). The β and p values indicating the relationship between resection and outcomes were computed for each voxel.

Structural Connectivity Outcome–Lesion Mapping

Different from voxel-based outcome–lesion mapping, the individual structural disconnectogram is a continuous voxel-based map, that is, each voxel is not a binary mask but rather encodes a continuous value related to the number of affected fibers. For this reason, we employed voxel-wise partial rank-based correlations (Spearman) to assess the relationship between surgical outcome as a ranked variable and number of fibers affected per voxel, controlling for the total surgical resection volume, the hippocampal volumes, the side of surgery, and site where the patient was treated. The voxel-wise R and p values were computed.

Functional Connectivity Outcome–Lesion Mapping

Similar to structural disconnectograms, functional maps also represented continuous voxel-wise values, in this case, Z score coactivation values. We therefore employed a comparable approach: voxel-wise partial rank-based correlations (Spearman) measuring the relationship between surgical outcome as a ranked variable and voxel-wise coactivation Z score, again controlling for the total surgical resection volume, the hippocampal volumes, the side of surgery, and site where the patient was treated. The voxel-wise R and p values were computed.

Image Processing Pipeline: Surgical Resection and Related Disconnections

We employed an enantiomorphic approach described by Nachev et al.38 implemented into the software SPM39 by the NiiStat toolbox developed by Chris Rorden (nitrc.org/projects/niistat/) to spatially normalize the individual surgical cavities to the MNI 152 stereotaxic standard space. This approach applies segmentation/normalization routines from SPM to estimate a nonlinear normalization matrix for the postoperative image in native space with the surgical cavity being replaced by the homologous contralateral nonlesioned tissue. The normalization matrix is then applied to the original postoperative image resulting in a T1-weighted scan in standard space without the tissue deformation that would have ensued by nonlinear normalization estimation using the cavity. The same normalization matrix was applied to the surgical mask in native space, yielding a surgical mask in stereotaxic space, which is comparable across individuals by being in the same space.

Strategies to Avoid Statistical Type I and II Errors

A common limitation of massive univariate voxel-based approaches is the high probability of type I (false negative) and type II (false positive) errors. This pitfall is greatly compounded when multiple correction variables are included in the voxel-wise models, that is, where the statistical analysis for each voxel includes multiple control variables to determine the more subtle association between one variable (resection or disconnection at that voxel) with outcomes while controlling for many other variables. This process is repeated for a very large number of voxels, and further multiple comparison corrections become disproportionately stringent and greatly increase the chances of type I errors. In this study, we opted to maintain the use of multiple controlling variables per voxel (i.e., resection volume, hippocampi volumes, site, and side of surgery) because these are fundamentally important to determine the true relationship between resection and outcomes while accounting for clinically relevant confounders. Nonetheless, the penalty of such an approach is the massive reduction in statistical power. To balance this bias and prevent type II errors, we adopted a voxel-wise cluster correction for multiple comparisons where only voxels with p < 0.05 located in clusters with at least 16 mm3 were considered significant. Moreover, we also prevented type II errors by performing analyses using 3 different imaging modalities and assessing the anatomical concordance between the results as important corroboration of true positive findings.

Data Availability

Anonymized data inclusion lesion masks and behavioral outcomes are available from the authors upon written request.

Results

Cohort Characteristics and Surgical Outcomes

The cohort's demographic and clinical profile is summarized in Table 1. Of the 113 patients included, 61 (54%) were female and 98 (86%) were right-handed. Mean age at seizure onset was 16.5 years (SD 11.9). Side of surgery was left in 54.9% of cases. A total of 62 (54.9%) patients underwent resection and 51 (45.1%) underwent stereotactic laser ablation at an average age of 38.8 years (SD 12.9). At least 12 months postoperatively, 69 (61.1%) patients were free of disabling seizures (Engel Class 1). Of those who experienced postoperative seizures, 19 (16.8%) were Engel Class 2, 20 (17.7%) were Engel Class 3, and 5 (4.4%) were Engel Class 4. There were no significant differences across sites except for the proportion of each surgical approach (χ2 = 69.5, df = 2, p < 0.001), with Emory performing mostly laser ablations (81.3% of their cohort) and Bonn performing exclusively resections. Of note, neither the site (χ2 = 6.5, df = 6, p = 0.37) nor the type of surgery was significantly associated with differences in outcomes (χ2 = 1.85, df = 3, p = 0.60).

Table 1.

Demographic and Clinical Profile

graphic file with name NEUROLOGY2021172719T1.jpg

Open in a new tab

Neuroimaging Results

The overlay of all surgical cavities in stereotaxic space is shown in Figure 1. Lesions were flipped to be represented as ipsilateral to the area of TLE onset, represented on the figure over the left side according to radiologic convention. eFigure 2 (links.lww.com/WNL/B651) demonstrates the association between preoperative gray matter volume and area of resection, demonstrating that there are larger resections associated with more prominent presurgical atrophy particularly in the hippocampus but less so in the surrounding tissue.

Figure 1. Lesion Overlap Across Participants.

Figure 1

Open in a new tab

For statistical analysis, left- and right-sided temporal lobe epilepsy were condensed into a single group by flipping their voxels into ipsilateral or contralateral. Lesions are represented here for both groups as ipsilateral arbitrarily on the left side (radiologic convention).

Voxel-Based Outcome–Resection Mapping

Voxel-based outcome–resection mapping revealed 2 clusters associated with better outcomes (Figure 2 and Table 2). More specifically, resection of voxels in the hippocampus and the amygdala–piriform cortex complex (cluster 1) as well as in the parahippocampal gyrus (entorhinal cortex) extending into the rhinal sulcus (cluster 2) were associated with better outcomes, controlling for the total volume of resection/ablation, side of surgery, treatment center, and presurgical hippocampal volumes.

Figure 2. Summary Results From Voxel-Based Outcome–Resection Mapping.

Figure 2

Open in a new tab

The results are corrected for multiple comparisons using a cluster-based threshold of p < 0.05. The ipsilateral and contralateral sides are identified on the first axial slice for orientation. For each modality, significant cluster details are in Table 1. Color bars represent p values at each voxel.

Table 2.

Anatomic Locations and Peak Stereotaxic Coordinates Related to Each Lesion-Symptom Model

graphic file with name NEUROLOGY2021172719T2.jpg

Open in a new tab

Structural Connectivity Outcome–Lesion Mapping

Structural connectivity outcome–lesion mapping demonstrated a single cluster where structural disconnections were associated with better outcomes (Figure 3 and Table 2). Disconnection of fibers in the uncinate fasciculus, anterior commissure, and fornix (cluster 1) were associated with better outcomes. This association was also observed after controlling for the total volume of resection/ablation, side of surgery, treatment center, and presurgical hippocampal volumes.

Figure 3. Summary Results From Structural Connectivity Outcome–Lesion Mapping.

Figure 3

Open in a new tab

The results are corrected for multiple comparisons using a cluster-based threshold of p < 0.05. The ipsilateral and contralateral sides are identified on the first axial slice for orientation. For each modality, significant cluster details are in Table 1. Color bars represent p values at each voxel.

Functional Connectivity Outcome–Lesion Mapping

Functional connectivity outcome–lesion mapping revealed 4 clusters where functional disconnection was associated with better outcomes (Figure 4 and Table 2). Functional disconnection of the contralateral superior frontal gyrus and contralateral middle frontal gyrus (cluster 1), bilateral gyri rectus (cluster 2), ipsilateral middle and superior gyri of the temporal pole (cluster 3), and the ipsilateral middle orbitofrontal cortex (cluster 4) were all associated with better outcomes, controlling for the total volume of resection/ablation, side of surgery, treatment center, and presurgical hippocampal volumes.

Figure 4. Summary Results From Functional Connectivity Outcome–Lesion Mapping.

Figure 4

Open in a new tab

The results are corrected for multiple comparisons using a cluster-based threshold of p < 0.05. The ipsilateral and contralateral sides are identified on the first axial slice for orientation. For each modality, significant cluster details are in Table 1. Color bars represent p values at each voxel.

Concordant Anatomical Pattern Between Connectivity Modalities

Structural and functional connectivity outcome–lesion mapping are based on different brain tissues, that is, white and gray matter, respectively. For this reason, it is expected that there is incomplete overlap between their statistical maps. It is nonetheless possible to evaluate the concordance between the results by qualitatively inspecting the projection of disconnected fibers towards disconnected cortex. This is demonstrated in Figure 5, where a clear pattern reveals the correspondence between disconnected anterior temporal fibers (corresponding to cluster 1 from structural connectivity outcome lesion mapping) with temporal pole and inferior frontal functional disconnection (clusters 2 and 3 from functional connectivity outcome lesion mapping). Functional disconnection of the orbitofrontal cortex was not associated with significantly disconnected structural fibers; nonetheless, there is a clear and unique projected structural pathway towards the orbitofrontal cortex that might be significant in larger cohorts with increased power.

Figure 5. Overlay Between Statistical p Value Maps From Structural and Functional Connectivity Outcome–Lesion Mapping (p < 0.5).

Figure 5

Open in a new tab

The crosshairs indicate the locations with significant structural or functional disconnection associated with outcomes. Note the correspondence between fiber projections (blue) and cortical disconnections (red).

Discussion

This study employed a multimodal lesion mapping approach to evaluate the relationship between surgical outcomes in TLE and resection/ablation location or its ensuing disconnection from the brain network. Controlling for presurgical medial TLE, we observed that better outcomes were associated with resections involving the hippocampal formation and the entorhinal–rhinal and amygdala–piriform cortices, as well as temporal pole and temporal–frontal structural and functional disconnections. The results reconcile previous seemingly inconsistent findings from the literature and can provide insight into factors related to surgical outcomes. They also underscore the value and caveats of voxel-based approaches for this type of investigation. These are discussed in further detail below.

Many studies have underscored the importance of complete hippocampal resection for seizure freedom.13,15,44 Nonetheless, there are anecdotal reports of optimal outcomes with incomplete resection,44,45 and a recent elegant study challenged the importance of hippocampal resection in nonlesional cases by demonstrating overall favorable outcomes in patients with hippocampal-sparing surgery.17 Our analysis shows that there is indeed a tendency to perform more extensive resections/ablations when the hippocampal volume is more atrophied preoperatively (as seen in eFigure 1, links.lww.com/WNL/B651). Hence, we sought to understand the anatomy that supports seizure outcomes beyond hippocampal resection. This stems from the observation of patients who continue to have seizures despite surgical lesioning of the hippocampus as well as intracranial EEG studies demonstrating that the entorhinal and perirhinal cortices are likely involved in seizure onset in a proportion of patients.46-48 Our results indicate that resection of the anterior hippocampus–amygdala–piriform region or the entorhinal–perirhinal regions is associated with better outcomes, thus reconciling these observations. We did not specifically test whether the resection of the rhinal or amygdala–piriform cortices are the crucial determinant for patients with hippocampal-sparing surgery, which may be a focus of a future study. Nonetheless, our findings suggest that resection of these regions is associated with better outcomes. Further studies are needed to determine whether resection/ablation of these regions can yield surgical freedom independently of each other or whether their combined removal is associated with better outcomes. It should be noted that we did not observe a relationship between hippocampal body resection and better outcomes, but this was likely an artifact related to the use of hippocampal volumes as a control covariate; that is, patients with less atrophied hippocampus would have large hippocampal body resections. Our findings also corroborate converging evidence from prior independent studies showing favorable outcomes seen with surgical lesioning of the entorhinal19,49 and piriform21 cortices independent of lesion volume, controlling for potential confounders such as total surgical resection volume, hippocampal volumes, side of surgery, and site performing the surgery.

Gray matter disconnection is an important factor in various conditions in neurology50,e1 such as stroke,e2,e3 dementia,e4 and traumatic brain injury.e5 Yet disconnection in epilepsy remains underappreciated as a potential factor associated with clinical phenotypes, including surgical outcomes. Essentially, cortical disconnection stipulates that seemingly intact gray matter has impaired function despite structural preservation directly as a result of reduced white matter connections.e1 For example, we have recently demonstrated that cortical disconnection of the left cingulum, bilateral amygdala, and left thalamus was associated with worse memory performance even after controlling for medial temporal volume loss, confirming the functional consequences of disconnection beyond the degree of gray matter loss.e6 Our present findings demonstrate that disconnection may also play a key role in seizure outcomes following epilepsy surgery. Specifically, disconnecting the anterior temporal lobes from the medial and lateral orbitofrontal cortex (by disrupting the uncinate fasciculus), the anterior and ventral temporal lobes including the amygdala from their contralateral homologs (by disrupting the anterior commissure), and the medial temporal structures from the remainder of the limbic system (by disrupting the fornix) were independently associated with better outcome. These findings support prior evidence from correlational analyses showing distinct changes in diffusion scalar maps associated with seizure freedom vs refractoriness after surgerye7-e9 as well as previous observations that a larger resection of uncinate fibers is associated with better ouctomes,e10 in line with authors advocating for disconnection surgery for TLE.e11,e12

Our results also indicate that functional disconnection of regions in the frontal (superior and middle frontal gyri, orbitofrontal region) and temporal (superior and middle pole) lobes are likely crucial for postoperative seizure freedom. Concordant anatomical patterns between connectivity modalities suggest these functionally disconnected regions could be related to the structural disconnections identified above.

Voxel lesion–symptom mapping was introduced by Bates et al.32 to statistically assess which brain regions are not just associated with but are essential for a given function or behavior. Few in vivo methods confirm the essential regions, because other imaging modalities, for example, fMRI, cannot dissociate between areas that are involved or are required for a function. Since then, lesion mapping has been widely used to determine the crucial anatomy of higher cognitive functionse13-e15 and has been refined to take into account caveats related to spatial normalization of brains with lesionse16 with implementation of statistical approaches to avoid statistical errors.e17 Lesion variability is one of the most desirable characteristics for datasets in lesion–symptom mapping studiese18 as interindividual differences in resection sites and outcomes provide statistical power to define voxel-based differences. The combination of temporal lobe open resections with laser ablation from multiple sites was purposefully chosen for this study for this reason. Lesion–symptom mapping approaches allow for a systematic and objective statistical approach to confirm anecdotal observations from lesion case series.

Cluster correction of voxel-based lesion analyses is a well-established method to control for multiple comparisonse19-e21; however, massive univariate analyses carry the risk of type I errors and these should be taken in context. Second, the co-occurrence of parahippocampal and hippocampal lesions in traditional approaches (i.e., nonhippocampal sparing) limits the ability to dissociate the differential role of each structure in outcomes, a well-established phenomenon in lesion–symptom mapping studies.e18 Third, the regions of interest in the AAL atlas employed may be too large for a fine-grained characterization of subtle anatomical variability. For example, the AAL “amygdala” ROI actually extends to the piriform cortex—a reason why we have chosen to refer to this as the amygdala–piriform complex. Similarly, the AAL “fusiform” ROI has more prominent involvement of the entorhinal and perirhinal regions. Thus, it will be critical for future studies following this approach to apply atlases with a focus on mesial temporal subfields.37 Fourth, we employed indirect measures of disconnection instead of de facto disconnections obtained from individual diffusion and functional MRI. The advantage of the indirect approach lies in the larger number of individuals from a connectome template, yielding a realistic approximation of white matter anatomy and functional synchronicity,e22 but it does not take into account individual-specific changes related to chronic white matter disease.23 Along the same lines, it is important to recognize that left- and right-sided TLE may have subtle differences in the pattern of aberrant structures and pathways; in addition, their surgical lesions may differ (e.g., right-sided lesions tend to be more extensive). Larger cohorts may allow enough power to explore independent lesion–symptom mapping for each side independently. Finally, seizures can relapse after more than 1 year postsurgery. We employed the 1-year cutoff routinely used in outcome analysis of epilepsy surgery,e23 but this study did not assess longer-term outcomes, which should be of interest for future studies in this field.

To conclude, we found that postoperative seizure freedom in TLE is associated with (1) surgical lesions that target the hippocampus as well as the amygdala–piriform cortex complex and entorhinal cortex; (2) disconnection of temporal, frontal, and limbic regions through loss of white matter tracts within the uncinate fasciculus, anterior commissure, and fornix; and (3) functional disconnection of the frontal (superior and middle frontal gyri, orbitofrontal region) and temporal (superior and middle pole) lobes may be associated with better postoperative seizure freedom. These findings shed light on the key components of epileptogenic networks in TLE and constitute a promising source of new evidence for future improvements in surgical interventions. Seizure freedom is an undoubtedly important outcome of epilepsy surgery. However, destructive or resective procedures need to be balanced with minimally invasive and highly targeted approaches to diminish morbidity, including visual field defects as well as language and memory impairments. Future studies that focus on the trade-off between these detrimental consequences of surgery and seizure outcomes will be critical for personalized medicine in epilepsy.

Glossary

AAL

automatic anatomical labeling

ILAE

International League Against Epilepsy

IRB

institutional review board

MNI

Montreal Neurological Institute

MUSC

Medical University of South Carolina

ROI

region of interest

TE

echo time

TI

inversion time

TLE

temporal lobe epilepsy

TR

repetition time

VBM

voxel-based morphometry

Appendix. Authors

Appendix.

Open in a new tab

Footnotes

Class of Evidence: NPub.org/coe

Study Funding

This study was supported by grants from the National Institute of Neurologic Disorders and Stroke (1R01NS110347-01A).

Disclosure

The authors report no disclosures relevant to the manuscript. Go to Neurology.org/N for full disclosures.

References

  • 1.Engel J Jr., McDermott MP, Wiebe S, et al. Early surgical therapy for drug-resistant temporal lobe epilepsy: a randomized trial. JAMA. 2012;307(9):922-930. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Wiebe S, Jette N. Epilepsy surgery utilization: who, when, where, and why? Curr Opin Neurol. 2012;25(2):187-193. [DOI] [PubMed] [Google Scholar]
  • 3.Tellez-Zenteno JF, Dhar R, Wiebe S. Long-term seizure outcomes following epilepsy surgery: a systematic review and meta-analysis. Brain. 2005;128(pt 5):1188-1198. [DOI] [PubMed] [Google Scholar]
  • 4.Polkey CE. Clinical outcome of epilepsy surgery. Curr Opin Neurol. 2004;17(2):173-178. [DOI] [PubMed] [Google Scholar]
  • 5.Elliott JO, Richardson VE. The biopsychosocial model and quality of life in persons with active epilepsy. Epilepsy Behav. 2014;41:55-65. [DOI] [PubMed] [Google Scholar]
  • 6.Arruda F, Cendes F, Andermann F, et al. Mesial atrophy and outcome after amygdalohippocampectomy or temporal lobe removal. Ann Neurol. 1996;40:446-450. [DOI] [PubMed] [Google Scholar]
  • 7.Berkovic SF, McIntosh AM, Kalnins RM, et al. Preoperative MRI predicts outcome of temporal lobectomy: an actuarial analysis. Neurology. 1995;45(7):1358-1363. [DOI] [PubMed] [Google Scholar]
  • 8.Radhakrishnan K, So EL, Silbert PL, et al. Predictors of outcome of anterior temporal lobectomy for intractable epilepsy: a multivariate study. Neurology. 1998;51(2):465-471. [DOI] [PubMed] [Google Scholar]
  • 9.Specht U, May T, Schulz R, et al. Cerebellar atrophy and prognosis after temporal lobe resection. J Neurol Neurosurg Psychiatry. 1997;62(5):501-506. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Abou-Khalil B, Andermann E, Andermann F, Olivier A, Quesney LF. Temporal lobe epilepsy after prolonged febrile convulsions: excellent outcome after surgical treatment. Epilepsia. 1993;34(5):878-883. [DOI] [PubMed] [Google Scholar]
  • 11.Schulz R, Luders HO, Hoppe M, Tuxhorn I, May T, Ebner A. Interictal EEG and ictal scalp EEG propagation are highly predictive of surgical outcome in mesial temporal lobe epilepsy. Epilepsia. 2000;41(5):564-570. [DOI] [PubMed] [Google Scholar]
  • 12.Krendl R, Lurger S, Baumgartner C. Absolute spike frequency predicts surgical outcome in TLE with unilateral hippocampal atrophy. Neurology. 2008;71(6):413-418. [DOI] [PubMed] [Google Scholar]
  • 13.Bonilha L, Kobayashi E, Mattos JP, Honorato DC, Li LM, Cendes F. Value of extent of hippocampal resection in the surgical treatment of temporal lobe epilepsy. Arq Neuropsiquiatr. 2004;62(1):15-20. [DOI] [PubMed] [Google Scholar]
  • 14.Kanner AM, Kaydanova Y, deToledo-Morrell L, et al. Tailored anterior temporal lobectomy. Relation between extent of resection of mesial structures and postsurgical seizure outcome. Arch Neurol. 1995;52(2):173-178. [DOI] [PubMed] [Google Scholar]
  • 15.Josephson CB, Dykeman J, Fiest KM, et al. Systematic review and meta-analysis of standard vs selective temporal lobe epilepsy surgery. Neurology. 2013;80(18):1669-1676. [DOI] [PubMed] [Google Scholar]
  • 16.Sheikh SR, Nair D, Gross RE, Gonzalez-Martinez J. Tracking a changing paradigm and the modern face of epilepsy surgery: a comprehensive and critical review on the hunt for the optimal extent of resection in mesial temporal lobe epilepsy. Epilepsia. 2019;60(18):1768-1793. [DOI] [PubMed] [Google Scholar]
  • 17.Morita-Sherman M, Louis S, Vegh D, et al. Outcomes of resections that spare vs remove an MRI-normal hippocampus. Epilepsia. 2020;61(11):2545-2557. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Schramm J. Temporal lobe epilepsy surgery and the quest for optimal extent of resection: a review. Epilepsia. 2008;49(8):1296-1307. [DOI] [PubMed] [Google Scholar]
  • 19.Bonilha L, Yasuda CL, Rorden C, et al. Does resection of the medial temporal lobe improve the outcome of temporal lobe epilepsy surgery? Epilepsia. 2007;48(3):571-578. [DOI] [PubMed] [Google Scholar]
  • 20.Siegel AM, Wieser HG, Wichmann W, Yasargil GM. Relationships between MR-imaged total amount of tissue removed, resection scores of specific mediobasal limbic subcompartments and clinical outcome following selective amygdalohippocampectomy. Epilepsy Res. 1990;6(1):56-65. [DOI] [PubMed] [Google Scholar]
  • 21.Galovic M, Baudracco I, Wright-Goff E, et al. Association of piriform cortex resection with surgical outcomes in patients with temporal lobe epilepsy. JAMA Neurol. 2019;76(6):690-700. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Bonilha L, Martz GU, Glazier SS, Edwards JC. Subtypes of medial temporal lobe epilepsy: influence on temporal lobectomy outcomes? Epilepsia. 2012;53(1):1-6. [DOI] [PubMed] [Google Scholar]
  • 23.Taylor PN, Sinha N, Wang Y, et al. The impact of epilepsy surgery on the structural connectome and its relation to outcome. NeuroImage Clin. 2018;18:202-214. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.He X, Doucet GE, Pustina D, Sperling MR, Sharan AD, Tracy JI. Presurgical thalamic “hubness” predicts surgical outcome in temporal lobe epilepsy. Neurology. 2017;88(24):2285-2293. [DOI] [PubMed] [Google Scholar]
  • 25.Gleichgerrcht E, Munsell B, Bhatia S, et al. Deep learning applied to whole-brain connectome to determine seizure control after epilepsy surgery. Epilepsia. 2018;59(9):1643-1654. [DOI] [PubMed] [Google Scholar]
  • 26.Gleichgerrcht E, Keller SS, Drane DL, et al. Temporal lobe epilepsy surgical outcomes can be inferred based on structural connectome hubs: a machine learning study. Ann Neurol. 2020;88(5):970-983. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Bonilha L, Helpern JA, Sainju R, et al. Presurgical connectome and postsurgical seizure control in temporal lobe epilepsy. Neurology. 2013;81(19):1704-1710. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Bonilha L, Jensen JH, Baker N, et al. The brain connectome as a personalized biomarker of seizure outcomes after temporal lobectomy. Neurology. 2015;84(18):1846-1853. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Morgan VL, Abou-Khalil B, Rogers BP. Evolution of functional connectivity of brain networks and their dynamic interaction in temporal lobe epilepsy. Brain Connect. 2015;5(1):35-44. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Morgan VL, Englot DJ, Rogers BP, et al. Magnetic resonance imaging connectivity for the prediction of seizure outcome in temporal lobe epilepsy. Epilepsia. 2017;58(7):1251-1260. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Morgan VL, Rogers BP, Gonzalez HFJ, Goodale SE, Englot DJ. Characterization of postsurgical functional connectivity changes in temporal lobe epilepsy. J Neurosurg. 2020;133(2):1-11. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Bates E, Wilson SM, Saygin AP, et al. Voxel-based lesion–symptom mapping. Nat Neurosci. 2003;6(5):448-450. [DOI] [PubMed] [Google Scholar]
  • 33.Chechlacz M, Rotshtein P, Hansen PC. Mapping functional brain organization: rethinking lesion symptom mapping and advanced neuroimaging methods in the understanding of human cognition. Neuropsychologia. 2018;115:1-4. [DOI] [PubMed] [Google Scholar]
  • 34.Gleichgerrcht E, Fridriksson J, Rorden C, Bonilha L. Connectome-based lesion–symptom mapping (CLSM): a novel approach to map neurological function. NeuroImage Clin. 2017;16(C):461-467. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Berg AT, Berkovic SF, Brodie MJ, et al. Revised terminology and concepts for organization of seizures and epilepsies: report of the ILAE Commission on Classification and Terminology, 2005-2009. Epilepsia. 2010;51(4):676-685. [DOI] [PubMed] [Google Scholar]
  • 36.Engel J, Jr., Wiebe S, French J, et al. Practice parameter: temporal lobe and localized neocortical resections for epilepsy. Epilepsia. 2003;44(6):741-751. [DOI] [PubMed] [Google Scholar]
  • 37.Bonilha L, Kobayashi E, Cendes F, Min Li L. Protocol for volumetric segmentation of medial temporal structures using high-resolution 3-D magnetic resonance imaging. Hum Brain Mapp. 2004;22(2):145-154. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Nachev P, Coulthard E, Jager HR, Kennard C, Husain M. Enantiomorphic normalization of focally lesioned brains. NeuroImage. 2008;39(3):1215-1226. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39.Friston KJ, Ashburner J, Kiebel SJ, Nichols TE, Penny WD. Statistical Parametric Mapping: The Analysis of Functional Brain Images. Elsevier; 2006. [Google Scholar]
  • 40.Yeh FC, Panesar S, Fernandes D, et al. Population-averaged atlas of the macroscale human structural connectome and its network topology. NeuroImage. 2018;178:57-68. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41.Yarkoni T, Poldrack RA, Nichols TE, Van Essen DC, Wager TD. Large-scale automated synthesis of human functional neuroimaging data. Nat Methods. 2011;8(8):665-670. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42.Ashburner J, Friston KJ. Voxel-based morphometry: the methods. NeuroImage. 2000;11(6):805-821. [DOI] [PubMed] [Google Scholar]
  • 43.Tzourio-Mazoyer N, Landeau B, Papathanassiou D, et al. Automated anatomical labeling of activations in SPM using a macroscopic anatomical parcellation of the MNI MRI single-subject brain. NeuroImage. 2002;15(1):273-289. [DOI] [PubMed] [Google Scholar]
  • 44.Wyler AR, Hermann BP, Somes G. Extent of medial temporal resection on outcome from anterior temporal lobectomy: a randomized prospective study. Neurosurgery. 1995;37(5):982-990; discussion 990-981. [DOI] [PubMed] [Google Scholar]
  • 45.Bannout F, Harder S, Lee M, et al. Epilepsy surgery for skull-base temporal lobe encephaloceles: should we spare the Hippocampus from resection? Brain Sci. 2018;8(3):42. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 46.McIntyre DC, Kelly ME. The parahippocampal cortices and kindling. Ann NY Acad Sci. 2000;911(1):343-354. [DOI] [PubMed] [Google Scholar]
  • 47.Piredda S, Gale K. A crucial epileptogenic site in the deep prepiriform cortex. Nature. 1985;317(6038):623-625. [DOI] [PubMed] [Google Scholar]
  • 48.Vaughan DN, Jackson GD. The piriform cortex and human focal epilepsy. Front Neurol. 2014;5:259. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 49.Keller SS, Richardson MP, Schoene-Bake JC, et al. Thalamotemporal alteration and postoperative seizures in temporal lobe epilepsy. Ann Neurol. 2015;77(5):760-774. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 50.Schmahmann JD, Smith EE, Eichler FS, Filley CM. Cerebral white matter: neuroanatomy, clinical neurology, and neurobehavioral correlates. Ann NY Acad Sci. 2008;1142(1):266-309. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • eReferences e1–e23: links.lww.com/WNL/B651

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Data Availability Statement

Anonymized data inclusion lesion masks and behavioral outcomes are available from the authors upon written request.

Articles from Neurology are provided here courtesy of American Academy of Neurology

RESOURCES