Reading no further than the title of this chapter has already presented you with a conundrum: how exactly does one define “complex partial seizures of temporal lobe origin”? It’s actually two difficult questions rolled into one seemingly innocuous phrase. First, a complex partial seizure (CPS) is defined as a seizure of focal onset accompanied by some impairment of awareness. Although this construction has a fair deal of practical value, particularly vis-à-vis driving safety, impairment of awareness can be very difficult to ascertain clinically.

Yet the second part is even more vexing. In theory, any seizure arising from a generator in the temporal lobe would qualify as a seizure of “temporal lobe origin.” But because our knowledge of epileptic circuits is paltry at best, this easy concept becomes exceedingly difficult to apply. What about a patient would unambiguously signify that his or her seizures were of temporal lobe origin? Frequent interictal spikes with a temporal maximum? Ictal recordings with origin in the temporal lobes? The history and semiology? The findings on imaging studies? In point of fact, it is well known that none of these are unambiguous indicators of seizure generation in the temporal lobes. The only way to truly confirm temporal lobe origin of seizures is to surgically resect the presumptive offending tissue and see whether seizures are abolished.

Of course, the “surgical standard” has a very high specificity but a very low sensitivity: it is applicable only to those with intractable epilepsy, and only to that subset who present for surgical evaluation and who are then found to have adequate concordant data, and low enough functional risk, to qualify for the resection. Thus, any conclusions about this population may or may not be applicable to the broader spectrum of patients with seizures of temporal lobe origin. With that caveat in mind, the following discussion will largely be drawn from the seizure-free surgical population, not only because of the lack of ambiguity, but because the large majority of the literature stems from this population. In the end, this may be the most relevant population to discuss in this volume, as the group of patients with temporal lobe seizures who are undergoing video-electroencephalographic (EEG) monitoring (VEM) will predominantly be those for whom surgery is being considered.

Little information is available regarding the incidence and prevalence of temporal lobe CPSs. This makes sense in light of the issue mentioned in the introduction: the tertiary care “surgical standard” for verification of temporal lobe origin is completely at odds with the broad-based community sample that is required for epidemiologic estimates. The best numbers come from Zarrelli et al.,¹ who found that of 156 patients with epilepsy, 21 (13.5%) had symptomatic or cryptogenic temporal lobe epilepsy (TLE), corresponding to an incidence of 6.8 cases per 100,000 person-years. Unfortunately, this estimation is heavily compromised by the fact that the period of observation predated the existence of MRI scans, and by the (probably related) fact that the majority of cases of focal epilepsy were classified as “unlocalized.” Thus, this should be considered a low-end estimate of both general population incidence and percentage representation among epilepsy cases. The gender distribution is even.

The typical natural history of mesial TLE due to hippocampal sclerosis (HS) consists of an early life “hit” of some kind, generally occurring before the age of 3 years.^2,3 Most commonly, this is a complex febrile seizure (prolonged in duration, or with clear focal features), but head trauma or central nervous system infection with acute symptomatic generalized tonic-clonic seizures may also be seen as precipitants. There is then a latent period, which averages 7 or 8 years, before the development of spontaneous, stereotyped complex partial seizures.² When TLE occurs due to causes other than HS (see discussion below), the age of onset can span the gamut from childhood to old age, although restricting the inquiry to the successful surgical population skews this toward the young for obvious reasons. Aside from mesial TLE due to HS, no characteristic syndrome has been identified.

Trying to gauge the natural history of temporal lobe complex partial seizures is particularly warped by the use of the surgical standard for disease definition. Those who are rendered free of CPS by temporal lobe resection are, by design, those for whom seizures persist despite medical treatment. The course of such patients may be that of intractability from the outset, but many patients also have periods of prolonged remission punctuated by recurrence of resistant seizures.⁴

Causes and Risk Factors

There are a host of potential causes for temporal lobe CPS. The most important of these are HS, cortical dysplasias, tumors, vascular malformations, infections, and head trauma.

HS is a unique pathological lesion responsible for TLE and is not to be confused with the “hippocampal sclerosis” sometimes mentioned in the dementia literature, which refers to preferential neuronal loss within the CA1 region of the hippocampus extending into the subiculum and not accompanied by epilepsy.⁵ It is also not to be confused with the hippocampal damage that accrues with anoxic injury, sometimes called end folium sclerosis because of its preferential involvement of the hilar (CA4) region. In contrast, HS produces a distinct pattern of neuronal cell loss and gliosis that involves the CA1, CA3, and CA4 hippocampal regions with relative and noteworthy sparing of CA2.⁶ On magnetic resonance imaging (MRI) done with the appropriate sequences, this manifests as reduced volume and increased signal in the involved hippocampus (or, occasionally, both hippocampi). Why this particular pathologic condition—or, for that matter, any pathologic condition—should produce spontaneous seizures remains wholly mysterious; many have postulated that the sprouting of mossy fibers in the dentate gyrus may have a causative role in the seizure generation of HS, but this remains inadequately proven.⁶

Cortical dysplasias are an important cause of temporal lobe CPS and may in fact be even more common than we realize, as the majority are quite subtle or MRI occcult.⁷ These abnormal (but not neoplastic) cells appear to be electrically irritable either at the cellular level, the network level, or both.⁸ There are several kinds, classified by their cellular architecture.⁹ They are often multifocal, which can make surgical localization and cure a challenge. Many patients who have apparently cryptogenic TLE likely have cortical dysplasias that are imaging occult.¹⁰ One report suggests that these lesions have a predilection for the bottoms of deep sulci,¹¹ which can make identification of abnormal electrical activity difficult even with the use of intracranial electroencephalography (EEG).

Brain tumors represent another important substrate for TLE. These can range from highly malignant gliomas to extremely indolent lesions, such as gangliogliomas and dysembryoplastic neuroepithelial tumors, that straddle the fine line separating dysplasia from neoplasia. In fact, the latter tumor types and others classified as World Health Organization grade I lesions are sometimes found with adjacent areas of dysplasia.¹² There may be something fundamentally biologically different about the tumors that produce chronic epilepsy, as even when the tumors are grade II or III lesions, survival is often prolonged well beyond what would be expected.¹³ This is one important reason why any patient with epilepsy should undergo MRI at least once, even if seizures have been present for many years.

The other type of foreign tissue lesion that can be responsible for temporal lobe CPS is a vascular malformation. Both cavernous angiomas and arteriovenous malformations are potential culprits,¹⁴ presumably due to seepage of blood that acts as an electrical irritant to the immediately surrounding brain tissue (even in the absence of frank hemorrhage). For this reason, resection of these lesions for epilepsy is generally performed with a margin.

The temporal pole is one of the regions most commonly affected by head trauma, with CPS a common concomitant. Although the epidemiologic data are not specific to temporal lobe seizures, at least two large-scale studies have found that the incidence of epilepsy overall remains markedly elevated for over a decade following severe traumatic brain injury, and for at least several years, if not longer, following even mild traumatic injury.^15,16 With head injury, as with foreign tissue lesions, the question often arises as to whether the seizures are generated by the heart of the damaged region or at the interface between the gliosis and the apparently undamaged brain. Yet neuropathologic studies of resected posttraumatic TLE patients reveal both neocortical pathology and substantial hippocampal hilar cell loss, suggesting that the hippocampus itself might indeed be a part of the generator in many such cases.¹⁷

Lastly, infections of various kinds can precipitate CPS from the temporal lobe. Encephalitis, especially that caused by herpes simplex, is the most common culprit, followed by bacterial meningitis; in either case, the seizures may arise unilaterally or bilaterally. In countries without advanced economies, cysticercosis is a common cause of CPS, and the characteristic lesions can be present in any location, including the temporal lobes.

Ictal Behavior

The advent of time-locked video-EEG technology dramatically expanded our ability to correlate the behavioral aspects of seizures with ictal EEG changes. The capacity for careful review of ictal phenomena has led to the identification of typical and atypical semiologic findings for certain types of epilepsy and the documentation of a number of valuable lateralizing signs (semiologic features that point to the hemisphere of onset). Many of the different semiologic features do not occur in isolation and can overlap with each other as the seizure evolves. It is therefore important to describe the predominant feature or features that occur during the seizure evolution.

The common auras associated with temporal lobe epilepsy have already been discussed in Chapter 8. In brief, auras very commonly precede CPS of temporal lobe origin; the abdominal or epigastric aura is very common for anterior temporal foci,^18,19 whereas auditory auras are more common with posterior or lateral foci.²⁰

The typical semiologic features of temporal lobe epilepsy have been well described.^18,21,22 When reviewing seizures, behavioral arrest is often the first symptom of the complex partial phase of the seizure.^18,23,24 Careful and detailed testing is required during the seizure and the postictal period to confirm that the patient truly has an alteration of awareness. Behavioral arrest is characterized by the abrupt cessation or alteration of the patient’s preictal behavior and is usually associated with staring and a lack of movement. Often the patient will visually attend to items or people, and this can be confused with retained awareness. Hoffmann and colleagues²³ reported that behavioral arrest at seizure onset was noted in approximately 25% of left temporal seizures (mesial and lateral) compared with 3% for right mesial and 10% for right lateral temporal seizures. Behavioral arrest is often very easy to identify during the awake state, but it may be effectively occult when seizures arise from sleep. It is typically followed by oroalimentary automatisms, manual automatisms, and more generalized body movements. This sequence of semiologic events is noted in approximately 69 to 73% of temporal lobe seizures.^18,25 Oroalimentary automatisms consist of rhythmic chewing, swallowing, and/or lip smacking; simple manual automatisms consist of semipurposeful picking or fumbling movements of the hands. The semiology described above is the most common temporal lobe seizure semiology, but it is also common to have temporal lobe seizures with “bland” staring and a paucity of movement—essentially a behavioral arrest without the ensuing automatisms. These seizures (sometimes called dialeptic seizures) are very challenging for epilepsy monitoring unit staff to recognize if the patient does not report an aura, and they are often discovered when the raw EEG is being screened.

Mesial versus Lateral Semiology

Several investigators have attempted to use semiology to differentiate mesial from lateral temporal lobe epilepsy. Maillard and colleagues²⁰ used invasive stereo-EEG to subdivide patients into three groups: medial, medial-lateral, and lateral ictal onset. There was no statistical difference among the groups with respect to the occurrence of loss of consciousness during a seizure; however, loss of consciousness within 10 seconds of onset was seen more frequently with lateral temporal involvement. Oroalimentary automatisms and seizure duration longer than 60 seconds were noted more frequently if the medial temporal lobe was involved. Lastly, the lateral temporal group was more likely to have recurrent seizures with secondary generalization. Early oroalimentary automatisms with mesial temporal lobe seizures were reported by Gil-Nagel and Risinger,²⁶ and O’Brien and colleagues²⁷ found no difference with respect to oroalimentary automatisms, seizure duration, or frequency of secondary generalization. This study was limited because the patients were segregated based on MRI lesion and concordant EEG data rather than seizure freedom after surgery. Early clonic activity is more common in lateral temporal lobe epilepsy,^19,27 whereas contralateral dystonic hand posturing (see the description below) is suggestive of mesial temporal onset.^19,27,28 These subtle differences in seizure semiology are hypothesized to be secondary to preferential spread patterns due to variations in synaptic connectivity, although the precise nature of these hypothetical spread patterns remains obscure.

Lateralizing Signs

Unilateral dystonic limb posturing has been found to be highly predictive of seizure onset in the contralateral hemisphere. The terms contralateral and ipsilateral are used with respect to the hemisphere of seizure onset. During the dystonic movement the limb attains a fixed posture, often with the fingers extended, flexion of the wrist and elbow, and rotation of the forearm, which helps to differentiate it from unilateral tonic limb posturing (which is not a reliable lateralizing sign). This sign has been noted in 18 to 35% of patients^29–31 and 27% of seizures³² in patients with temporal lobe epilepsy undergoing video-EEG monitoring (VEM). The positive predictive value is ~90 to 92%.^30,31 When unilateral dystonic posturing is associated with contralateral automatisms, it becomes both a lateralizing and a localizing sign, being strongly suggestive of onset in the contralateral mesial temporal region specifically.³³

Version is the forced, sustained, and unnatural movement of the head and eyes contralateral to the epileptic focus. This sign typically occurs near the end of a CPS (Video 9-1) and usually precedes a secondary generalized tonic-clonic seizure.³⁴ This lateralizing sign has been noted in 26 to 35% of patients^29–31 and 24% of seizures³² in patients with TLE undergoing VEM. The positive predictive value is 100% if this sign immediately precedes a secondary generalization. The positive predictive value decreases to 60 to 100% if one includes any versive movement at any time during the seizure.^30,31,35 The variation in the positive predictive value may be explained by the difficulty in recognizing true version (sustained, forced, and unnatural positioning) from nonversive head movements, as this may be more difficult when not followed by generalization.

Another common lateralizing sign noted immediately prior to secondary generalization is unilateral tonic facial contraction. In one series of patients with mesial temporal lobe epilepsy (MTLE), this occurred in ~40% of patients during the early phase of the secondary generalization, and the tonic contraction was 100% contralateral to the epileptic focus.³⁶

Ictal speech may be described as the production of meaningful language during a seizure with an alteration of awareness. Single-word utterances or common phrases, such as “Oh my God,” should not be considered ictal speech, as this is more likely to represent a verbal automatism. Occasionally, ictal speech will also occur along with automatisms (oral or manual) in the context of preserved responsiveness. This lateralizing sign has been noted in 16 to 19% of patients^30,35 and 13 to 20% of seizures^32,37 in patients with TLE undergoing VEM. Ictal speech lateralizes the ictal onset to the nondominant hemisphere in 80% of patients.^30,35

During the postictal phase, the patient should be examined specifically for language function, as postictal dysphasia is also a useful sign. It is often very difficult to differentiate aphasia from generic postictal confusion immediately following a seizure, but with careful language testing or a standard reading protocol,^l this sign can be reliably identified.³⁸ Postictal dysphasia has been noted in 21 to 51% of patients^30,35 and 29% of seizures³⁷ in patients with temporal lobe epilepsy undergoing VEM. This sign lateralizes the ictal onset to the dominant hemisphere in 67 to 100% of patients.^30,35

Unilateral blinkingcan occur during focal seizures with an approximate frequency of 0.8 to 1.5%.^39,40 The majority of patients reported in the literature have unilateral blinking that is ipsilateral to the epileptic focus.^39–41 Benbadis et al.³⁹ calculated a positive predictive value of 83% when compared with EEG localization; however, this included temporal and extratemporal seizures.

Postictal nose wiping is a common sign that occurs during the postictal phase of temporal lobe seizures. The reported frequency of postictal nose wiping in patients with temporal lobe epilepsy ranges from 41 to 56%.^42–45 The hand used to wipe the nose is typically ipsilateral to the epileptic focus. It has a positive predictive value ranging from 90 to 97% and has a high interrater reliability.^42–44

EEG

The literature on the EEG patterns of temporal lobe CPSs is quite consistent in its descriptions, with a few predominating motifs. The most common by far is the initial focal theta-alpha discharge of 5 to 9 Hz waves (Figure 9-1), localized to the anterior or inferior temporal electrodes unilaterally and found to occur in 65 to 80% of temporal lobe CPSs.^46,47 This pattern was first thoroughly elaborated by Risinger et al.⁴⁸ The theta-alpha frequency discharge may occur as the very first EEG change, or it may follow a period of diffuse or focal EEG change at a different (usually slower) frequency. It is nonetheless highly localizing and reliable as long as it occurs within the first 30 seconds of ictal onset.⁴⁸ In fact, at least two studies have correlated this pattern with intracranial electrode onset, with one investigation showing that this pattern correlated with onset in the mesial contacts of temporal lobe depth electrodes in 82% of cases.⁴⁸ Another study refined this further, noting that this type of ictal onset is highly correlated with hippocampal seizure generation but does not appear until seizures spread into the basolateral cortical regions.⁴⁹ Yet another study directly correlated this pattern of onset with pathology, finding that theta frequency onset was strongly correlated with moderate or severe HS.⁵⁰ Thus, the unilateral temporal 5 to 9 Hz discharge within 30 seconds of onset has been established as the veritable EEG “signature” of the hippocampal seizure. Given the occasional false localization, however, it appears that this signature can sometimes be forged.

A second type of common ictal onset is the unilateral temporal delta frequency discharge of 2 to 5 Hz (Figure 9-2). This type of onset has been found to correlate with extrahippocampal seizure generation, usually from the temporal neocortex.⁴⁹ Using the “surgical standard” for verification, this pattern is associated with ictal generators in the temporal tip or basal temporal cortex, although it is likely that anterolateral or posterolateral ictal onset can also produce this pattern.⁵¹ Delta frequency temporal activity at ictal onset has been found to occur specifically in patients with mild or no HS, validating that it is typically seen with onset outside the hippocampus.⁵⁰ Nonetheless, delta frequency onset can occasionally occur with hippocampal epilepsy.⁴⁹

Other onset patterns, though not entirely atypical, occur less frequently, and their localizing value is less well established. Sometimes the initial activity is within the delta frequency range but takes the form of sharp waves rather than slow waves. These sharp waves, often resembling the patient’s typical interictal sharp waves, occur in a 1 to 4 Hz rhythmic pattern at onset before evolving into other rhythms⁵² (Figure 9-3). Paradoxically, in patients with very frequent interictal spikes, onset may take the form of abrupt cessation of spikes.⁵³ This is generally accompanied by focal attenuation of activity, then followed by the typical theta or delta frequency rhythmic activity described above. This is probably a reliable localizing pattern, as it denotes a highly irritative region that further betrays its role in seizure generation by becoming “otherwise occupied” at ictal onset—hence the cessation of interictal spikes.

One other pattern described is that of diffuse, irregular slowing without any clear evolving rhythmic discharge.⁴⁹ On intracranial EEG, seizures of this type produce an evolving rhythmic discharge that remains very restricted in extent, thus producing a vague scalp correlate. This pattern is nonspecific with regard to localization, being seen in seizures from the mesial temporal structures or the neocortex. In the authors’ experience, this type of discharge is not commonly seen with temporal lobe CPSs.

It is worth reviewing temporal lobe seizures to the end, as post-ictal EEG changes can be highly valuable for localization. Focal delta slowing or focal attenuation in one temporal lobe is seen post-ictally in about two thirds of temporal lobe CPS (Figure 9-4), and has been found to have a localizing value of 95 to 100%,^46,47 a correlation stronger than that seen for any ictal onset pattern.

When correlating EEG with semiology, the onset of behavioral change prior to EEG change is not a cause for concern in TLE. In fact, it is uncommon for the EEG to precede the behavioral event in mesial TLE,⁴⁶ presumably due to the time required for a mesial temporal discharge to spread and involve enough neocortex to be visible at the scalp. Of course, EEG can precede behavioral change in CPS of temporal neocortical origin.

Because the onset of temporal lobe CPSs is so often in structures that are at some distance from the surface, there is a long history of use of alternative electrode types in an effort to get closer to the source of discharges in a less than fully invasive manner. The two major types currently in use are sphenoidal electrodes and foramen ovale electrodes. Both types are used as supplements to the traditional array of scalp electrodes with the notion that because of their physical proximity to the mesial temporal structures, they should be more likely to detect interictal epileptiform activity or the earliest ictal changes. Evidence to support this proposition is mixed at best, particularly with regard to ictal onsets. Sphenoidal electrodes are often compared to inferior temporal electrodes, called T1 and T2 in the traditional 10/20 system or FT9 and FT10 in the 10/10 system. Using this comparison, two studies found no difference in ictal onsets between sphenoidal and anterior temporal electrodes.^54,55 Some investigators have reported that a small fraction of ictal recordings are detected first in the sphenoidals prior to the scalp electrodes.^56,57 In addition, placement of the sphenoidals under fluoroscopic guidance, so that they are ideally situated in close proximity to the foramen ovale, has been reported to notably improve ictal onset detection.⁵⁸ One of the authors has suggested that paper versus digital review of ictal recordings might be responsible for this discrepancy.⁵⁹

Foramen ovale electrodes may be considered almost an extension of fluoroscopically-placed sphenoidals, with the electrode crossing through the foramen ovale instead of simply remaining near it.⁵⁷ Consequently, they offer something like a single “virtual depth electrode” in the mesial temporal regions without the need for general anesthesia. In one study of patients with HS, foramen ovale electrodes revealed seizure onset within 0.3 to 1.7 seconds after depth electrode ictal onset and allowed for localization in 87.5% of patients who were not lateralized by scalp and sphenoidal electrodes.⁶⁰ They were also useful in verifying ictal onsets that were either bilateral or contralateral to the side of HS. Their utility is nonetheless limited to a very narrow group of patients, which has undoubtedly impeded the widespread adoption of this technique.

Ictal Behavior

The interesting phenomenon of automatisms with preserved responsiveness (Video 9-2) typically occurs in patients with nondominant TLE. Strictly speaking, this would not qualify as a CPS, but it is discussed here because its features mimic those of typical temporal lobe CPSs. Ebner and colleagues⁶¹ reported that ~6% of their patients with TLE (all from the nondominant right temporal lobe) demonstrated this finding, in which the patient displays oral and/or manual automatisms without losing responsiveness and with full memory of the event. Typically, the patient can communicate during the seizure, following commands and even talking while repetitively lip-smacking and the like. This is an important sign to recognize since it strongly suggests a non-dominant temporal lobe onset.^61–63 Only one patient has been reported in the literature with onset from the dominant temporal lobe.⁶⁴