Neuroimaging in the Evaluation of Epilepsy

Naymee J. Velez-Ruiz, MD; Joshua P. Klein, MD, PhD2


Semin Neurol. 2012;32(4):361-373. 

In This Article

Hippocampal Sclerosis

The syndrome of mesial temporal lobe epilepsy (MTLE) is the most common form of symptomatic localization related epilepsy.[14] The associated histopathologic substrate is usually hippocampal sclerosis (HS),[15,16] which can be reliably detected by MRI (Fig. 1). Optimal imaging of the hippocampus requires orthogonal axial and coronal sequences, with axial images in a plane along the long axis of the hippocampus. Classic findings of hippocampal sclerosis are atrophy of the hippocampal gray matter, best shown on inversion recovery sequences, and increased hippocampal signal on T2-weighted images (T2WI). A series of 81 patients that underwent therapeutic temporal lobectomy, with pathologic findings available in all patients, demonstrated that MRI can detect HS with great sensitivity (93%) and specificity (86%).[17] MRI evidence of hippocampal atrophy has been shown to have a positive predictive value of up to 86% for excellent postoperative seizure control after anterior temporal lobectomy.[18] However, in a minority of cases of HS, these qualitative radiologic findings are not present. This dilemma led to the development of quantitative volumetric hippocampal analysis. Volumetric methods correlate well with histopathologically confirmed hippocampal cell loss.[19] Thus, reduced volume by quantitative analysis has been established as a surrogate marker for the presence and severity of hippocampal atrophy. In addition, hippocampal atrophy by MRI volumetry has been demonstrated to be 75% sensitive to, and 64% specific for, ipsilateral medial temporal lobe seizure onset as corroborated by intracranial EEG recordings.[20]

Figure 1.

Hippocampal sclerosis. A 31-year-old right-handed woman had frequent seizures consisting of staring and automatisms, then fist clenching, upper extremity posturing, and reduced level of consciousness. During events, electroencephalogram (EEG) showed rhythmic sharp waves centered over the right anterior temporal lobe electrodes. (A) Coronal T2-fluid attenuated inversion recovery magnetic resonance imaging (FLAIR MRI) showed abnormal hyperintensity in the right hippocampus. (B) Axial T2-FLAIR MRI with fat suppression confirmed abnormal hyperintensity in the right hippocampus. (C) Interictal 18F-fluorodeoxyglucose positron emission tomography (18FDG-PET) showed corresponding hypometabolism (darkness) in the right temporal lobe. (D) The patient underwent right anterior lobectomy, seen on axial T2-FLAIR MRI, with pathologic evidence of marked hippocampal neuronal loss and gliosis. The patient is seizure-free without anticonvulsant medications.

Despite its known utility, hippocampal volumetry has been difficult to incorporate in clinical practice because of the time demands and the technical skills required. As a solution, investigators have substituted time-consuming manual techniques with automated software for generic quantitative morphometrics. Recent studies have shown that these automated techniques can also detect hippocampal asymmetry and lateralize hippocampal atrophy accurately.[21] Other techniques, such as T2 relaxometry and magnetic resonance spectroscopy (MRS), have encountered the same practical difficulties than volumetric analysis and their use for the detection of HS has been limited mostly to investigational purposes. Hippocampal T2 relaxation time is prolonged ipsilateral to HS identified by visual inspection and volumetry.[22] In MRS, comparison with control subjects has shown that the temporal lobe ipsilateral to the seizure focus (with evidence of HS on visual inspection) has a mean reduction in the NAA signal, increases in the Cr and Cho peaks, and an abnormally low NAA/Cho:Cr ratio.[23] However, in most cases, these metabolic abnormalities extend beyond the epileptogenic focus.[24]

Often, in patients with MTLE, there are other MRI findings associated with those typical of HS. These include ipsilateral enlargement of the temporal horn, reduced size of the fornix and the mamillary body, loss of the normal interdigitations of the hippocampal head, hippocampal malrotation, and atrophy of the collateral white matter.[25–27] However, their significance with regards to the diagnosis and prognosis of MTLE and HS is not well understood. With the advent of 7 Tesla (7T) MRI, there has been an increased focus in the study of these associated findings because ultrahigh field strength enables detailed evaluation of normal and abnormal hippocampal folding and rotation. Besides permitting the detection of selectively greater Ammon horn atrophy in MTLE with HS, this technique has also allowed the identification of a paucity of digitations of the hippocampal head as a deformity independent of hippocampal atrophy on the epileptogenic side of patients with MTLE.[28] Malrotation of the hippocampi is clearly visualized as well, but it has been detected in both MTLE and control subjects.[28] Studies with 7T MRI are limited to a small amount of subjects and further investigation of these findings is needed.

Patients can also present with dual pathology on MRI, including hippocampal atrophy in conjunction with an additional structural lesion.[29] The most common of such lesions are the malformations of cortical development, which have been reported in up to 15% of patients with HS.[30] In addition, patients can present with bilateral, symmetric, or asymmetric hippocampal atrophy detected by visual or volumetric analysis.[31] When this is the case, the epileptologist relies on the electroencephalographic documentation of the side of ictal onset and on other functional imaging modalities.

18F-FDG PET is routinely used in the presurgical evaluation of MTLE. Interictal studies show hypometabolism in a relatively wide area encompassing the temporal lobe of seizure onset, but the pathophysiology of this phenomenon is incompletely understood. In a meta-analysis by Spencer, the overall sensitivity of interictal FDG-PET to temporal lobe epilepsy as judged by EEG criteria was found to be 84% with a specificity of 86%.[32] Furthermore, when pathologic groups were analyzed, PET measurement of metabolism had even higher sensitivity for HS. However, the extent of the hypometabolism does not correlate with the degree of cell loss or hippocampal atrophy.[33,34] There seems to be an association between hypometabolism and uncontrolled seizures or medical refractoriness.[35–37] In fact, 18F-FDG PET evidence of restricted temporal lobe hypometabolism, by visual inspection correlates with postoperative seizure control after anterior temporal lobectomy.[38] The absence of, or evidence of multilobar, hypometabolism is considered concerning when encountered during presurgical evaluation because it suggests worse surgical outcome.[38,39] Some investigations have shown that compared with normal controls, more than 15% reduction in predominantly lateral temporal metabolism has the best correlation with seizure outcome (despite pathologic abnormalities in medial temporal areas),[40–42] whereas others reported a correlation with hypometabolism of the medial temporal region.[38] Overall, the sensitivity of PET in MTLE is limited by the lack of localization within the temporal lobe. Despite its limitations, FDG PET is considered extremely useful in patients with medically intractable MRI negative TLE, in which the decision to surgically intervene is often taken on the basis of electrographic, metabolic, and neuropsychologic data.[43]

Ictal SPECT is another functional imaging modality utilized in the presurgical evaluation of epilepsy. In patients with MTLE, it shows increased ictal cerebral perfusion in the medial and lateral temporal regions when the tracer injection is appropriately administered. In these patients, ictal SPECT is a very sensitive (90–93%), but not so specific (13–77%) diagnostic tool.[32] Interpretation is difficult because it requires knowledge of seizure type, clinical activity, time of ictal injection in relation to seizure onset and MRI findings. From a practical point of view, in MTLE it adds little useful information beyond what is provided by EEG-video telemetry and structural MRI, especially when there is a lesion.[44] Hence, it usually does not alter the surgical decision and outcome for patients with HS. In fact, its use increases the length of hospital stay and subsequently, the costs and risks of video-EEG monitoring. The use of SPECT is usually reserved for the evaluation of intractable epilepsy of presumed extratemporal origin, and in some cases it is helpful in delineating the epileptogenic zone when planning for implantation of intracranial electrodes.

There has been much debate regarding the utility of MEG in the evaluation of MTLE with HS. When MEG is utilized to determine the electric current dipole in MTLE with HS, the activity is found to be distributed in diverse regions that may be unrelated to the ictal-onset area.[45] Moreover, there is persistent debate as to whether or not MEG recordings in patients with MTLE can detect mesial temporal interictal epileptiform discharges (spikes). Wennberg and colleagues found that intracranially detected spikes localized to the mesial temporal structures could not be detected with extracranial MEG, which suggests that this noninvasive source localization technique may be unable to identify true mesial temporal spikes.[46] In their data, classic anterior or midtemporal spikes recorded with MEG seemed to be generated in anterior and lateral temporal neocortical structures and were neither propagated from nor to the mesial temporal region.[46] Another study found that current dipole orientation and distribution of epileptiform activity correlates with cortical thinning in left MTLE with HS.[47] These results suggest that regardless of the presence of HS, in a subgroup of patients with MTLE a large cortical network is affected.[47] MEG is more commonly used when neocortical epilepsy is suspected because relevant lesions are often small and easily missed during routine MRI review, or too large to be completely resected and the area of interest needs to be defined.

Functional MRI (fMRI) evaluates cerebral blood flow by looking at the difference between venous oxyhemoglobin and deoxyhemoglobin. This is called the blood-oxygenation-level-dependent (BOLD) contrast technique. When certain functional areas in the brain are activated, there is an increase in local metabolism and oxygen consumption. This MR technique may be helpful in the lateralization and localization of language functions when planning for an extended temporal lobectomy. Generally, good concordance exists between fMRI and the WADA test.[48] However, there are certain requirements that need to be met to enable its utility, including a battery of paradigms that can be administered to patients in a clinically feasible time. fMRI is also believed to be a valid tool for assessing verbal memory lateralization in patients with mesial temporal lobe epilepsy.[49,50] However, activation statistic maps produced with fMRI sometimes reveal noisy data that may be difficult to interpret, and the specific paradigms for verbal memory may be difficult to develop. Despite its limitations in the lateralization of verbal memory, fMRI studies have shown that functional adequacy of the left hippocampus is the best predictor of postoperative memory outcome.[51,52] In most centers, this technique is commonly used for the lateralization and localization of language functions, and in some specific cases for the lateralization of verbal memory.[53] However, overall, most centers still rely on the WADA test for the definite lateralization of verbal memory.