Bouchet and Cazauvieilh, in 1825, described grossly visible and palpable abnormalities of hippocampal autopsy specimens obtained from nine patients with “alienation epilepsy.”⁷ This was the first evidence linking the most common form of partial epilepsy to a pathologic substrate. In another autopsy study, Sommer, in 1880, described gliosis and pyramidal cell loss within the hippocampus of epileptic patients and hypothesized that this finding was responsible for their epilepsy.⁷¹ In 1872, Hughlings Jackson suggested that hyperexcitable brain adjacent to temporal lobe lesions was responsible for epilepsy, and in 1888 he pointed out the association of the “dreamy state” and the classic auras of taste, smell, and epigastric sensations with temporal lobe lesions.³⁰ Therefore, well over 100 years ago, a variety of temporal lobe epileptogenic substrates were recognized.

Surgical resection provided the link between medial temporal pathology and seizures by permitting the correlation of seizure control and electrophysiologic evidence of epileptogenicity with the resected tissue’s histopathology. The first operations were pioneered by Horsley¹⁸ in 1886, who performed surgery for epilepsy by excising visible cortical scars. Bailey and Gibbs, in 1951, popularized the resection of the temporal lobe using electrographic data, limiting their resection to the lateral temporal cortex.⁵ Penfield²⁰ also adopted intraoperative electrocorticography (ECoG) to determine the extent of his temporal lobectomy, but, with Baldwin, he noted an increase in successful outcome with more complete resection of the medial structures (amygdala and anterior hippocampus).⁶⁰

Temporal lobe resection in some form continues to be the most commonly performed operation for the medically intractable symptomatic epilepsies. Almost all temporal lobectomies performed before the late 1970s involved a generous lateral temporal neocortical resection, which was either standard or determined by intraoperative ECOG spikes and limited by essential cortex such as language and visual fields. The volume of medial hippocampus, parahippocampus, and amygdala removed was extremely variable and surgeon dependent.

In the late 1970s, based on invasive electrophysiology and pathology, Spencer modified the standard resection for those patients with medial temporal ictal onset, such that all of the medial structures (amygdala, hippocampus, and parahippocampus) were removed via a limited temporal pole resection.⁷⁵ At the same time, Wieser and Yasargil independently devised the more restricted amygdalohippocampectomy for presumptive medial basal epilepsy with or without an associated amygdalohippocampal mass (usually tumor).¹⁰⁰ Although some centers still perform generous lateral resections often based on ECOG, modifications of these more selective medial approaches have been used increasingly over the last decade. The increased use of selective medial resection has been based on the aforementioned convergence of medial electropathophysiology and on additional literature demonstrating more failures of temporal lobe resection when the medial structures are incompletely resected.^51,56,63

Recently, a number of centers have looked at the seizure and neuropsychological outcomes of patients that have undergone transsylvian selective amygdalohippocampectomy.⁵⁷ These results seem to indicate that seizure control is equivalent to that with resections that involve limited resection of the lateral structures.⁵⁷ The neuropsychological results show either equivalence⁴¹ or some trends toward decreased morbidity for functions involving the temporal lobe but increased morbidity for frontal lobe functions, especially when associated with subarachnoid hemorrhage from the sylvian dissection.²⁵

The following discussion is concerned with patients undergoing evaluation for surgical candidacy for treatment of intractable epilepsy. Typically, patients with new-onset seizures and significant MRI findings are treated with paradigms that address their underlying pathology first but may be considered for epilepsy surgery should their epilepsy prove to be intractable despite optimal treatment of their underlying pathology.

Selecting patients with TLE for the appropriate operation begins by first using MRI to classify them into the three anatomic substrate categories noted earlier: (a) hippocampal atrophy (MTLE with presumed MTS), (b) lesions (LRTLE), and (c) normal anatomy (CTLE). Noninvasive tests are then applied to search for either concordance to an anatomic abnormality or a well-delineated temporal lobe EEG localization when the MRI is normal. For example, in a patient with a history of febrile seizures or early injury at age <4 years and unilateral hippocampal atrophy, a selective medial temporal resection might be indicated if the MTLE syndrome was confirmed by video-EEG–recorded appropriate ictal pattern, temporal lobe–specific short-term memory deficit, and unilateral hypometabolic PET scan when available. Finally, although the method and interpretation vary considerably, the intracarotid Amytal procedure (IAP) will usually demonstrate a more specific memory deficit in the suspicious temporal lobe.

A patient with a temporal lobe lesion also requires EEG concordance. Neuropsychological, PET, and IAP testing can range considerably from normal to various degrees of abnormality. These latter three tests are not well categorized regarding usefulness in deciding about a resection or its volume. Nonetheless, the concordance of scalp EEG and anatomic substrate directs the surgeon to a specific region. The variability of other tests can result from lesion location, extent and age of an injury or lesion, and the central nervous system response to the same. The IAP’s major role has been to assure the physician that memory is adequate in the temporal lobe opposite to the one to be surgically addressed. Outcome studies with substantial numbers of patients are needed to clarify how cognitive testing, IAP, PET, and ictal and interictal SPECT can be used in modifying decisions based on the major criteria of MRI substrate and EEG localization.

In cases in the third category (CTLE), in which hippocampal atrophy or T2 signal change does not clearly predict MTS and the MRI appears normal, all EEG criteria must be typical of MTLE and the PET scan likewise positive for selective medial temporal resection to be performed in some centers. Patients in this group with a normal MRI often undergo chronic invasive electrophysiologic recording of seizures. Almost all centers agree that some form of invasive recording must be used if the preoperative evaluation criteria are discordant with one another; particularly when any imaging abnormality (MRI, PET, or SPECT) conflicts with the EEG. The temporal lobe resection region will then be tailored to the invasive ictal findings.

Chronic invasive electrophysiologic monitoring shows that the majority of ictal events observed in TLE originate in the medial temporal lobe, often within the hippocampus.^78,79,82,83 Electrographic ictal localization with invasive studies, such as depth and subdural electrodes, is accepted as the most accurate method for identifying the region of seizure onset.^14,23,79 Use of interictal spikes was advocated historically and is still advocated by some to define the cortical resection.⁵⁵ However, the location of interictal spikes has only been weakly correlated with the site of ictal onset or any pathology within the resected tissue.^23,72

Considering that interictal spikes may represent synaptically mediated events propagated from distant abnormal sites, this finding is not unexpected.¹⁰³ Epileptiform spiking can be nonspecific because it has been described in nonepileptic individuals.⁸ Ictal electrophysiologic data remain the standard for resective procedures for TLE. The disconnection of afferent and efferent networks also helps to explain how varied temporal lobectomy approaches across the spectrum from standardized to tailored resections can have similar outcome efficacy.

Discordant electrophysiology and imaging also may be seen in patients with LRTLE and may lead to invasive recordings to clarify ictal onset. When an MRI lesion is present, the necessity for invasive EEG is substrate dependent. It is infrequently necessary when the lesion is a neoplasm, circumscribed DA, or CA, unless there is ictal scalp onset clearly from another lobe or the opposite hemisphere. Other MRI lesions may not be associated so robustly with just perilesional epileptogenicity because a microscopic substrate may extend beyond the imaged abnormality. These patients with LRTLE require invasive recording to define the necessary and sufficient temporal lobe volume that must be removed. Common lesions in this category are most of the DAs, such as cortical dysplasia and gyration defects, and the indeterminate substrates, including traumatic and infectious gliosis and neuronal loss. Separate consideration for invasive monitoring may be necessary to stimulate and localize certain essential cortical regions such as language and sensorimotor areas when awake craniotomies are not possible, such as in children or adults with limited ability to undergo awake procedures.²³

In MTLE that meets evaluation criteria, neuropathologic, neuroanatomic, and neurophysiologic^10,33,102 findings all identify abnormalities in the medial temporal structures, particularly
the hippocampus. Therefore, an extensive medial resection removes the majority of this epileptogenic substrate. Accumulated evidence suggests that the medial structures should be resected extensively in this syndrome. Depth electrodes placed parallel and through the long axis of the hippocampus have shown that electrical abnormalities are usually found throughout the body of the hippocampus, and less often is there a distinct segment of the hippocampus that initiates the seizure. In approximately 20% of patients with unilateral TLE studied with depth electrodes, abnormal electrical foci in the posterior portions of the hippocampus were not routinely resected by traditional standard lobectomies.^78,83 MRI most frequently shows atrophy of the entire hippocampus.⁴⁴ And although investigators have described a decreasing gradient of neuronal cell loss from anterior to posterior hippocampus, pathology is generally seen throughout.⁴ The need to include the posterior portions of the hippocampus in the medial resection was addressed by designing a more radical hippocampus and parahippocampectomy that preserved the lateral temporal cortex—the anterior medial temporal resection (AMTR). Finally, Wyler et al. reported a prospective, randomized study of patients undergoing surgery for MTLE, with ictal onset verified in each case by subdural recordings.¹⁰⁵ They compared the AMTR to a more limited medial resection in this group. Complete seizure control was seen in 69% of the AMTR patients and in only 38% of the patients with a partial hippocampectomy. Other surgical options are also reviewed in what follows, representing variations of opinion of the extent of medial resection necessary and sufficient for success, and representing different approaches to minimizing the functional disruption of overlying cortex. They all share the common goal of medial temporal resection, specifically of the hippocampus.