Temporal Lobe Surgery and Memory: Lessons, Risks, and Opportunities

Abstract

Careful study of the clinical outcomes of temporal lobe epilepsy surgery has greatly advanced our knowledge of the neuroanatomy of human memory. After early cases resulted in profound amnesia, the critical role of the hippocampus and associated medial temporal lobe structures to declarative memory became evident. Surgical approaches quickly changed to become unilateral and later, to be more precise, potentially reducing cognitive morbidity. Neuropsychological studies following unilateral temporal lobe resection have challenged early models which simplified the lateralization of verbal and visual memory function. Diagnostic tests, including intracarotid sodium amobarbital procedure (WADA), structural MRI, and functional neuroimaging (fMRI, PET, and SPECT), can more accurately lateralize and localize epileptogenic cortex, and predict memory outcomes from surgery. Longitudinal studies have shown that memory may even improve in seizure-free patients. From seventy years of experience with epilepsy surgery, we now have a richer understanding of the clinical, neuroimaging, and surgical predictors of memory decline---and improvement---after temporal lobe resection.

I. Introduction

Memory dysfunction is the chief cognitive complaint in temporal lobe epilepsy (TLE) (Bell, Lin, Seidenberg, & Hermann, 2011) with some degree of impairment in most patients (Fisher et al., 2000; B. P. Hermann & Seidenberg, 2008). Refractory TLE is associated with progressive memory impairment (Dodrill, 2004; C. Helmstaedter & Elger, 1999; Hendriks et al., 2004; B. Hermann et al., 2008; Hoppe, Elger, & Helmstaedter, 2007; Jokeit & Ebner, 1999; Pitkanen & Sutula, 2002; Seidenberg, Pulsipher, & Hermann, 2007). These patients may be outstanding candidates for epilepsy surgery, which offers a potential cure for seizures and further memory decline. Careful longitudinal studies of TLE patients before and after resective surgery has accelerated our understanding of the neuroanatomy of human memory (Bell et al., 2011). Neuropsychology has been established as a fundamental tool for monitoring outcome and quality measures after epilepsy surgery (Christoph Helmstaedter & Witt, 2012).

What kinds of memory decline do TLE patients suffer? Patients with TLE typically report problems with declarative memory (See Box 1), which is memory that can consciously or explicitly communicated to others (Cohen & Squire, 1980) and includes episodic and semantic memory (Squire, 1992). Episodic memory, a term coined by Endel Tulving, involves personal events embedded in a spatio-temporal context (Tulving, 1972). Active retrieval of an episodic memory includes using the content of the event to retrieve specific details such as when and where the event occurred, its emotional valence, and other individuals present. Episodic memories are autobiographical in nature, and consitute a form of mental time travel in which we can recover the associated context of the event (Kahana, 2012). Episodic memories can include the vivid recollection of the family members, food, and atmosphere of a holiday celebration; what we remember about yesterday’s lecture and speaker; and where we placed our keys this morning. Semantic memory concerns factual knowledge about the world which is accumulated over time but informs present understanding. Examples include knowledge of public, historical events; object concepts including sensory properties, names, and functional uses; and scientific facts, numbers, and mathematical equations. Semantic memory has been studied through probing recall of famous public events or people to determine if a temporal gradient in retrospective memory decline exists (Barr, Goldberg, Wasserstein, & Novelly, 1990; Markowitsch, 2000).

Memory and Epilepsy Surgery: Lessons, Risks and Opportunities.

• Some early patients who underwent mesial temporal lobe surgery and suffered severe amnesia revealed the critical function of the hippocampus and related structures to declarative memory.

• Further study of post-operative cognitive outcomes of standard anterior temporal lobe resections, as well as the outcomes of newer more selective surgical procedures has yielded additional insights into the neuroanatomy of memory.

• Better understanding of the neuroanatomy of memory has motivated improved localization of seizure onset zone through diagnostic tools, discovery of patient risk factors for postoperative memory decline, and development of more selective surgical approaches.

• The concept of “nociferous cortex,” that epileptogenic cortex exerts harmful effects on cognition, suggests that cognitive outcomes should routinely be measured over years after surgery. Positive memory outcomes illustrate the possibility of functional recovery in epilepsy patients once seizures are treated.

Among patients with unilateral mesial TLE, patients with high seizure burden demonstrated greater anterograde episodic memory impairment. Both high and low seizure burden patients had poorer retrograde memory for autobiographical episodes and public events memories compared to healthy controls (Voltzenlogel, Vignal, Hirsch, & Manning, 2014). The dissociation between these cognitive phenotypes suggest differing neuroanatomical substrates for these memory categories. The structure-function relationships can be further tested by examining the cognitive outcomes from this unique set of well-circumscribed surgical “lesions” (Liu et al., 2017).

This paper reviews TLE surgical cases and post-operative cognitive outcomes, followed by presurgical diagnostic assessments to improve lateralization and localization of the seizure onset zone and predictors of memory decline after surgery. We review how surgical techniques have become more precise, to potentially reduce post-surgical cognitive deficits. Finally, we survey post-operative cognitive outcomes, with an emphasis on what temporal lobe resection (TLR) has revealed about the neuroanatomy of memory.

II. Early Lessons on Memory Loss after Medial Temporal Lobe Resection

Insights into the neuropsychology of memory have been informed by cognitive changes after surgery. The first epilepsy surgery was performed in 1886 in the United Kingdom, by Horsely and MacEwen in collaboration with Hughlings Jackson (Taylor, 1987). These pioneering surgeries involved the identification and removal of lesions in 3 epilepsy patients. The earliest surgeries for TLE, performed by Penfield and Jasper in Montreal and Bailey and Gibbs in Chicago, avoided the medial temporal lobe. Kluver and Bucy’s monkey experiments showed significant behavioral decline with bitemporal lobe resection (Bruce P. Hermann & Stone, 1989). After the role of medial temporal lobe (MTL) in seizure networks was identified in the 1950s, surgeries often included the MTL. At the time, the function of the hippocampus and associated structures was poorly understood.

The cognitive catastrophes suffered by several patients who underwent MTL resection in the late 1950s revealed the essential role of the hippocampus and neighboring cortical areas in memory (Xia, 2006). These early disasters fostered the development of diagnostic tests to better localize seizure focus and cognitive function, to improve surgical outcomes and reduce memory impairment. In parallel, cognitive neuroscience and experimental animal studies more precisely defined the neuroanatomy of different memory systems.

Case H.M. revolutionized our understanding of human memory. In 1953, the neurosurgeon Scoville performed a bilateral MTL resection on Henry Molaison (HM), a man with normal intelligence but medication-refractory seizures. The resection included “the anterior two-thirds of the hippocampus and hippocampal gyrus bilaterally, as well as the uncus and the amygdala.” Previously, Scoville performed similar surgeries on schizophrenic patients to reduce their psychosis, but did not adequately assess their post-operative memory. Careful study of HM’s cognitive function revealed that a significant reduction in seizures cost him the ability to form new, stable memories (Dossani, Missios, & Nanda, 2015; Scoville & Milner, 1957).

While Scoville and neuropsychologist Milner initially thought that this was a pure anterograde memory deficit (Scoville & Milner, 1957), further study revealed episodic autobiographical memory was also impaired for events occuring during the prior year to surgery (Steinvorth, Levine, & Corkin, 2005). H.M. had relatively preserved semantic memory for vocabulary, object, and factual knowledge acquired before surgery. Episodic memory for public events before surgery was also intact, as was recognition memory (Freed & Corkin, 1988). His personality, social skills, and intelligence appeared unchanged (Scoville & Milner, 1957). Implicit learning, including priming and acquiring new motor skills, was also preserved (Suzanne Corkin, 1968).

Even after death and over 60 years after his surgery, H.M.’s case still generates new insights in the neuroanatomy of memory (Dossani et al., 2015). Immediately post mortem, several 3T and 7T MRIs revealed that H.M.’s lesion was not purely hippocampal as once thought, but included medial temporal cortex, piriform cortex, entorhinal cortex, anterior parahippocampal gyrus, most of the amygdala, perirhinal cortex, and subiculum; and only the anterior half of the hippocampus (Augustinack et al., 2014). Autopsy confirmed that H.M. retained a significant amount of hippocampal tissue. However, most of the entorhinal cortex was removed bilaterally, thereby deafferenting the remaining hippocampus. Thus, extensive bilateral medial temporal lobe resection, not selective hippocampal damage, resulted in his significant episodic memory deficit (Annese et al., 2014).

Penfield’s surgical cases in the 1950s demonstrated that even unilateral temporal lobe resection could severely harm memory. One patient who underwent left anterior temporal lobe resection, including the anterior half of the hippocampus, experienced severe anterograde memory deficits similar to H.M. (S. Corkin, Amaral, Gonzalez, Johnson, & Hyman, 1997; Milner & Penfield, 1955; W. Penfield & Mathieson, 1974). The patient had semiology, scalp electroencephalogram (EEG), electrocorticography, and intraoperative cortical stimulation that supported seizure onset from the left temporal lobe. However, autopsy later demonstrated right sided hippocampal sclerosis. Re-evaluation of the preoperative EEGs showed one seizure with ictal spread from the right temporal lobe to the left (W. Penfield & Mathieson, 1974). Thus, a unilateral temporal lobectomy caused a near-global amnesia due to inadequate cognitive reserve of his remaining temporal lobe (S. Corkin et al., 1997; Dossani et al., 2015).

Scoville and Penfield’s surgical cases offered early lessons into the importance of the hippocampus and related medial temporal lobe structures for declarative memory function. Ability to form new declarative memories (both episodic and semantic) was affected, while sparing memory for prior semantic knowledge. Ability to learn new motor sequences remained intact. Penfield’s case emphasized the need to assess the functional reserve of the contralateral MTL (Chelune, Naugle, Luders, & Awad, 1991). Further, effective surgical therapy depends on accurate lateralization and localization of the epileptogenic cortex. Presurgical testing has evolved in response to these early instructive cases to improve seizure focus localization and reduce the risk of postoperative cognitive deficit.

III. Dominant temporal lobe resection produces a decline in verbal memory but there is significant variability in cognitive outcomes.

Unilateral temporal lobe resection is an effective therapy, resulting in seizure remission in up to 80% of patients with refractory TLE (de Tisi et al., 2011). However, surgery risks further memory impairment. While not resulting in the profound amnesia which bilateral MTL resection produced, memory often declines after surgery.

The material-specific model of memory, developed by Milner and colleagues at the Montreal Neurological Institute (Milner, 1970), proposed that left (or language-dominant) and right (or non-dominant) temporal lobes process verbal and non-verbal material differently (Milner, 1970). Verbal memory impairment is observed in left TLE patients, which can further decline with resection. Conversely, non-verbal memory declines with non-dominant temporal lobe resection, although this finding has been inconsistent and less robust (Barr 1997). Nevertheless, this simplistic model continues to influence presurgical decision making and interpretation of post-operative outcomes today (Baxendale, 2008; Bell et al., 2011; Saling, 2009).

The material-specific model’s predictions for verbal memory decline is generally supported by large, observational studies. In a meta-analysis of neuropsychological outcomes after temporal lobe surgery, Sherman et al. (2011) reported that 44% of patients with a left TLR had verbal memory decline, at twice the rates for right TLR patients (20%). Rates of verbal memory decline have varied from 30–60% for left (speech dominant) ATL (Milner, 1975; Rausch et al., 2003).

However, the material-specific model’s predictions for non-verbal memory decline have been variably supported by the evidence. In a meta-analysis, visual memory declines after left and right sided surgeries at equal rates (21% and 23% respectively) (Sherman et al., 2011). Visual memory outcomes after right TLR depend on the cognitive task, demonstrate small effect sizes, and are inconsistent (Vaz, 2004). Barr has proposed that visual memory must be conceptualized – and tested – as a dorsal (where) stream and ventral (what) stream (Barr 1997) to more precisely capture the memory deficits in non-dominant TLE.

These heterogeneous findings led to a critical reappraisal of the material specific model of memory (Saling, 2009). Saling proposed that verbal and visual memory should not be considered to be unitary constructs and strictly lateralized to dominant and non-dominant lobe. Instead, performance must be considered in light of the specific task demands. For example, different tasks of verbal (episodic) memory (e.g. list learning, prose recall) place differing demands on prior semantic knowledge. Confrontation naming (i.e. Boston Naming Test) is a verbal naming task, but objects are presented visually to the subject. Ability to name the object likely depends on familiarity with its sensory and functional properties. Thus, neuropsychological tests are not purely verbal or visual tasks, but entail complex demands on episodic memory, semantic knowledge, and visual or auditory processing. These considerations may explain why decline after unilateral TLR is highly variable across patients, and why non-dominant TLR produces inconsistent cognitive outcomes. Saling argued that we must disambiguate the medial versus lateral contributions to memory tasks (Saling, 2009).

Patient characteristics also contribute to variable cognitive outcomes. The degree of existing neuronal loss in the left hippocampus predicts worse performance in on a verbal episodic memory task of unrelated word-pair associates (Rausch, 1987; Saling et al., 1993; Sass et al., 1992). A structurally intact hippocampus predicted greater functional decline after surgery (B. P. Hermann, Wyler, Somes, Berry, & Dohan, 1992; Rausch & Babb, 1993). In other words, removal of atrophic, sclerotic hippocampus is less likely to result in significant memory decline, compared to a removal of a healthy, functional hippocampus (Bell et al., 2011). Patients with little or no hippocampal sclerosis undergoing left ATLR demonstrated approximately 35% decrease in long delay memory (as measured by the California Verbal Learning Test, or CVLT) after surgery. In comparison, patients with sclerotic hippocampal tissue experienced little decline after surgery (Bell et al., 2011). Consistent with these imaging findings, patients with better preoperative memory and language performance experienced greater memory decline after left TLR compared to those with worse preoperative performance (Chelune et al., 1991).

Finally, some variability in the magnitude of decline likely results in difference in surgical technique, which can differ by surgeon and center. Overall, larger left temporal lobe resections result in worse verbal memory (C. Helmstaedter, Petzold, & Bien, 2011; C. Helmstaedter, Roeske, Kaaden, Elger, & Schramm, 2011). For example, after accounting for baseline performance, the extent of left parahippocampal resection accounted for 27% of the variance in short delay free recall on a word list task; while the extent of left entorhinal resection accounted for 37% of the variance in performance (Liu et al., 2017).

V. Modern presurgical assessments to localize seizure onset zone and assess risk of memory decline

Since scalp EEG, electrocorticography, and even intraoperative stimulation could not always localize seizure foci or identify risk of postoperative cognitive impairments (W. Penfield & Mathieson, 1974), other assessments were needed. These assessments aimed to lateralize memory and language, and also test the cognitive reserve of the cortex contralateral to the planned resection.

Intracarotid Sodium Amobarbital (WADA) Test

In 1948, Wada performed an intracarotid artery (ICA) injection of sodium amytal (amobarbital) to study the epileptic discharges across hemispheres. He serendipitously founded a test to lateralize speech and memory as injecting via the dominant hemisphere’s carotid artery transiently impaired ipsilateral cerebral hemispheric function. In 1960, he demonstrated that amobarbital injections accurately lateralized speech and language function, by correlating with post-surgical outcomes (Wada & Rasmussen, 1960).

Milner, Rasmussen, and Branch (1962) first used the amobarbital test to assess hippocampal function contralateral to the probable temporal resection. Patients were presented drawings of objects before unilateral ICA amobarbital injection. A few minutes later, if the patient failed to spontaneously recall the objects, recognition memory was probed (Milner et al., 1962). Since the amobarbital procedure anesthetizes the brain regions supplied by the middle and anterior cerebral artery ipsilateral to the injection, failure to recall or recognize the presented items suggested inadequate functional reserve of the contralateral medial temporal lobe. Temporal lobe resection ipsilateral to injection would likely impair postoperative memory (Loring, Meador, & Lee, 1992).

Routine WADA testing has declined since the 1990s (Baxendale, 2008). While the WADA obtained gold-standard status in assessing lateralized material-specific memory outcomes, recent studies demonstrated that baseline neuropsychological evaluation, structural imaging, and neuropathology effectively predict quantitative post-operative memory status (Baxendale, Thompson, Harkness, & Duncan, 2006), with the WADA making little or no independent contribution (Binder, 2011; Binder et al., 2010; Chelune et al., 1991; Lineweaver et al., 2006; Stroup et al., 2003). Despite these criticisms, the amobarbital test remains the only functional test to assess each hemisphere’s individual contribution to memory (Loring, Barr, Hamberger, & Helmstaedter, 2008).

Structural Neuroimaging

Brain magnetic Resonance Imaging (MRI), developed in the 1980s, can identify structural lesions causing epilepsy^. Concordance of MRI lesion with ictal EEG onset predicts seizure freedom in most cases (Fish and Spencer 1995). Early low resolution MRI scans were more sensitive than CT. Among 48 temporal lobe epilepsy (TLE) patients, 71% had abnormal 0.5 Tesla (T) MRI scans while only 17% had abnormal CT scans. The MRI at 0.5 T correctly identified all patients with large structural lesions, including arteriovenous malformations, gliomas, hamartomas, and meningioangiomatosis. (Fish & Spencer, 1995). High field 3 T and T MRI has further increased our ability to identify epileptogenic lesions. 3T MRI is more than twice as likely to identify epileptogenic lesions than 1.5 T MRI (Phal et al., 2008), and provides greater resolution of the grey-white junction. 7 T MRI identified focal cortical dysplasias and malformations of cortical development in 23% of epilepsy patients who were MRI “negative” on 1.5 T or 3 T scans or in those with suspected dual pathology (i.e., a structural lesion in addition to mesial temporal sclerosis) (Veersema et al., 2017).

Further, 0.5 T MRI can identify more than 75% of patients with severe neuronal loss and gliosis of the mesial or lateral temporal lobe; and half of all patients with mild to moderate mesial or lateral temporal lobe neuronal loss and gliosis (Kuzniecky et al., 1987). Hippocampal atrophy and T2 hyperintensity on 0.5 T MRI correlates with hippocampal sclerosis verified with pathology (Berkovic et al., 1991). Identification of hippocampal sclerosis remains an important biomarker for lateralization and localization of epileptogenic networks, as well as predicting memory outcomes.

Functional Neuroimaging

Since the 1990s, measurement of brain activity during cognitive and motor tasks has supported presurgical planning. Functional MRI (fMRI) maps cortical function by identifying regions of increased neuronal activity and coupled blood flow during cognitive tasks, measured as the blood-oxygen-level dependent (BOLD) contrast between the task and rest condition. Compared to the amobarbital test, fMRI is noninvasive, less expensive, and possesses finer spatial resolution. However, because regional blood flow changes happen over seconds, fMRI lags behind the temporal resolution of EEG or magnetoencephalography (MEG), which can detect changes occuring over milliseconds (Binder & Carlson, 2011).

Compared to the amobarbital test, fMRI has concordance rates of 86–91% for language lateralization (Arora et al., 2009; Janecek et al., 2013), with better sensitivity for right hemispheric language function (Janecek et al., 2013). Discordant findings between fMRI with WADA is a reflection of language lateralization by fMRI as a continuous variable rather as a binary function (left versus right hemisphere) (Binder et al., 1996). fMRI allows calculation of relative language dominance expressed as the laterality index from −1 for pure right-sided dominance to +1 for pure left-sided dominance (Seghier, 2008).

fMRI can also assess brain areas involved in memory tasks, to predict memory outcomes following surgery (Binder, 2011; Bonelli et al., 2010; Sidhu et al., 2015). Patients with greater left frontal and anterior hippocampus activation, during a word encoding task had greater verbal memory decline after left anterior temporal lobectomy. Conversely, patients with left greater than right posterior hippocampal activation had less verbal memory decline (Bonelli et al., 2010; Sidhu et al., 2015). Patients with right greater than left anterior hippocampal fMRI activation during a face encoding task had greater visual decline after right-sided surgery, while posterior hippocampal activation predicted better memory outcome (Bonelli 2010). Left-sided memory lateralization index (LI) was also associated with significant postoperative verbal memory decline (Sidhu et al., 2015). On the other hand, Binder et al. (2010) did not find hippocampal LI in a word list learning and delayed recall task correlated with verbal memory outcome; however fMRI language LI was predictive of decline in patients who received a left ATL. In a series of stepwise multiple regression analysis, Binder et al found that clinical traits such as preoperative memory score and age of epilepsy onset accounted for approximately 50% of the variance in list learning memory, while the fMRI LI accounted for an additional 10% in list learning outcome. In their model, WADA results did not improve the predictive power of the model (Binder et al., 2010).

Positron emission tomography (PET) and single photon emission computed tomography (SPECT) are imaging techniques which can improved localization of seizure onset zones. PET and SPECT use radiolabeled probes whose emissions are detected by the scanner. FDG-PET uses radiolabeled glucose or oxygen to assess areas of altered metabolism or blood flow in the brain (Gaillard, 2011). The PET glucose ligand has a half-life of 110 minutes, and assesses interictal blood flow patterns (la Fougere, Rominger, Forster, Geisler, & Bartenstein, 2009). Gaillard et al. (1995) used interictal PET to show that TLE patients have decreased interictal glucose metabolism in the ictal onset temporal lobe. PET was more sensitive in patients with MRI abnormalities, correctly lateralizing seizure onset zone in 87% of cases with MRI lesions and only 60% of cases without lesions (Ho et al., 1995).

The SPECT ligand technetium 99m hexamethylpropylene amine oxime (99m Tc-HMPAO) is rapidly fixed in the brain allowing the study of cerebral blood flow at the time of injection (Newton et al., 1994; Rowe, Berkovic, Austin, McKay, & Bladin, 1991). The majority of SPECT ligand crosses the blood-brain-barrier quickly and becomes trapped within the cell compartment (Kim & Mountz, 2011). Peak brain levels of the SPECT ligand occur within 2 minutes after injection, with little redistribution for at least two hours, which makes it useful to study ictal blood flow (la Fougere et al., 2009). The SPECT subtraction method compares the ictal to interictal SPECT blood flow patterns to determine the most likely ictal onset zone. Among 35 patients with well localized TLE, correct seizure onset was lateralized correctly in 89% of patients with ictal SPECT versus 63% with interictal FDG-PET. However, using less strict criteria (i.e., lower level of confidence) for lateralization, there was no significant difference between ictal SPECT (94%) and interictal PET (83%) (Ho et al., 1995).

In summary, these diagnostic tools have been used to more accurately pinpoint seizure onset zone, and predict risk of memory decline. Regarding the latter, Stroup et al have found that clinical and imaging data including (1) resection of the dominant hemisphere, (2) MRI findings besides unilateral mesial temporal sclerosis, (3) intact preoperative verbal memory performance, and (4) good intracarotid sodium amobarbital (WADA) test performance after injection of the hemisphere contralateral to the seizure focus, predicted memory decline after surgery (Stroup et al., 2003). Together, these risk factors suggest that the functional and structural integrity of the to-be-resected temporal lobe anticipates post-operative memory impairment (Chelune et al., 1991).

VI. Surgical methods have become more selective and less invasive, reducing memory morbidity

Multimodal techniques to improve seizure localization and assess medial temporal lobe memory function has led to more restricted surgical resections. W. Penfield and Baldwin (1952) performed their anterior temporal lobectomy including a sucker to extract the hippocampus and amygdala. They described excisions extending beyond the anterior 5.5 cm causing contralateral superior visual field defects. Falconer (1953) modified this procedure using en bloc resection, which enabled better pathological characterization. In 1956, Morris proposed the standard temporal lobectomy including the anterior 6.5 cm of the temporal lobe, including the uncus, amygdala, anterior 2–4 cm of the hippocampus, and lateral temporal cortex (Morris, 1956). The lateral temporal cortex would later be spared by modified temporal lobectomy and selective amygdalohippocampectomy.

In 1984, Spencer, Spencer, Mattson, Williamson, and Novelly (1984) found that 20% of TLE patients had a seizure focus including posterior hippocampus, beyond the limits of the standard anterior 6.5 cm lobectomy, but were hesitant to extend the posterior resection margin further due to speech function typically residing in the lateral temporal lobe. Instead, only the anterior 4.5 cm of lateral temporal lobe was resected, which allowed better exposure to resect the amygdala, hippocampus, parahipppocampus, uncus, and fusiform gyrus (Spencer et al., 1984).

Selective amygdalohippocampectomy offered a strategy to limit lateral temporal cortex resection. Niemeyer’s transventricular amygdalohippocampectomy involved an incision in the second temporal gyrus to access the lateral ventricle and remove the hippocampus and amygdala (Niemeyer, 1958). Limited data regarding seizure and cognitive outcomes limited widespread adoption. Later, Wieser and Yarsagil developed the transylvian amygadalohippocampectomy aimed to preserve more neocortex than Niemeyer’s original transventricular amygdalohippocampectomy (Wieser & Yasargil, 1982). Their small study observed that amygdalohippocampectomy caused less verbal memory deficits than anterior two thirds temporal lobectomy.

Seizure outcomes in anterior temporal lobectomy and selective amygdalohippocampectomy remain an area of active research. In anterior temporal lobectomy, seizure outcomes depend on the extent of resection and preoperative pathology. Randomized and retrospective studies reveal that anterior temporal lobe resections with more extensive hippocampal removal result in twice the likelihood of achieving seizure freedom (Stavem, Bjornaes, & Langmoen, 2004; Wyler, Hermann, & Somes, 1995). Anterior temporal lobectomy is more likely to result in seizure freedom if the MRI shows a concordant lesion, such as temporal lobe atrophy, tumor, or mesial temporal lobe sclerosis (Stavem et al., 2004). Recent studies favor anterior temporal lobectomy over selective amygdalohippocampectomy. While initial studies demonstrated similar rates of seizure freedom after anterior temporal lobectomy and selective amygdalohippocampectomy (Clusmann et al., 2002; Paglioli et al., 2006), meta-analyses revealed that anterior temporal lobectomies were more likely to achieve seizure freedom (Hu et al., 2013; Josephson et al., 2013).

The Responsive Neurostimulation System (RNS, NeuroPace) is an FDA-approved device to detect and treat refractory focal-onset epilepsy using closed-loop electrical stimulation. Patients with refractory focal epilepsy, who are poor candidates for resection due to overlap between epileptogenic and eloquent cortex, and have one or two seizure foci are ideal RNS candidates (Bergey et al., 2015; Geller et al., 2017; Morrell, 2011). The system includes two four-contact leads placed directly on the seizure focus, which records and store changes in local field potentials. Clinicians customize the RNS System to detect patient-specific epileptiform activity. Median seizure reduction with was 53% 2 years after implantation (Heck et al., 2014) and 62% after 5 years (Bergey et al., 2015).

Neuropsychological testing of RNS patients reveals no significant cognitive decline at two years after implantation (Loring, Kapur, Meador, & Morrell, 2015). Patients with neocortical seizure onsets were more likely to experience modest improvements in naming, while those who had medial temporal lobe onsets were more likely to have improvements in verbal learning. The reason for these improvements may reflect reduced seizures or interictal discharges or neuromodulatory effects of electrical stimulation (Loring et al., 2015).

Laser interstitial thermal ablation (LITT) is a minimally invasive surgery that can treat epilepsy caused by small lesions such as mesial temporal sclerosis, cavernomas, or cortical dysplasias. The procedure utilizes a stereotactically inserted catheter which is then heated with a laser to thermally ablate the surrounding area. LITT is less invasive, requires a shorter hospital stay, and permits a faster return to normal activities compared to open surgery (Kang & Sperling, 2018).

Seizure and cognitive outcomes after LITT remain limited by small cohort studies. Among 23 TLE patients who underwent LITT, 65% remained free of disabling seizures at 1 year, with 73% of patients with mesial temporal sclerosis attaining seizure freedom (Jermakowicz et al., 2017). LITT may result in better cognitive outcomes compared to anterior temporal lobectomy, due to smaller volume of tissue ablated. Among 19 TLE patients who underwent LITT, there was no decline in recognizing or naming famous faces or in naming common nouns; in contrast to 39 who underwent open resection. Left ATLR resulted in impaired naming of famous faces and common objects, while right ATLR impaired face recognition (Drane et al., 2015). However, two TLE patients who underwent LITT showed significant postoperative verbal and visual memory decline with intact naming, visuospatial ability, and attention (Dredla, Lucas, Wharen, & Tatum, 2016). Another report of five left TLE patients who underwent LITT had intact contextual (narrative) verbal memory postoperatively, but three experienced significant non-contextual verbal memory decline as measured by list learning (Kang et al., 2016). Available data show that LITT spares semantically-loaded memory tasks and naming compared to standard temporal lobe resections, but long-term seizure outcomes remain poorly defined and may depend on identification of a single small lesion.

VII. Nociferous cortex and possible functional recovery after surgery.

The dynamic relationship between seizure burden and memory decline and longitudinal clinical outcomes after surgery have yielded important information. Post-operative cognitive outcomes depend on seizure outcomes. For patients with unilateral TLR, ongoing seizure burden appears to worsen memory, causing a “double jeopardy” (C. Helmstaedter, Kurthen, Lux, Reuber, & Elger, 2003). However, patients whose seizures are cured or significantly reduced after surgery may have a long-term cognitive benefit from surgery. These clinical observations illustrate the concept of the “nociferous cortex” and the cognitive impact of ongoing seizures.

Concept of the nociferous cortex

Nociferous cortex refers to epileptogenic tissue that is dysfunctional in three ways, it: 1) is the origin or element in the epileptogenic network, 2) does not perform its normal functions and 3) impairs the function of other brain areas (Nearing, Madhavan, & Devinsky, 2007). Nociferous is derived from Latin nocere, to harm.

The first reference to this concept in epilepsy surgery is from Krause and Schum (1932), who noted that in some cases of infantile hemiplegia, strength improved after resection of the epileptogenic cortex. In 1950, Welch and Penfield reported that after resection of cortical seizure foci in three patients with hemiplegic cerebral palsy, spasticity was reduced and motor function improved (Welch & Penfield, 1950). The first case involved a 22 year-old pathologically left-handed woman who demonstrated right sided hemiplegia at age one month and developed focal epilepsy at age 11 years. After seizure focus resection involving primary sensorimotor cortex,

“There was a remarkable change in the patient’s hemiparesis. Instead of carrying the paretic upper extremity in a flexed and spastic manner as she had done before, she now kept her arm extended by her side …The muscles were plastic. She was beginning to use the hand for eating which had never been possible before. In walking she could swing her leg without the former spastic stiffness so that her hemiplegic limp had actually disappeared. She had spent the year doing satisfactory university work.”

Welch and Penfield concluded that the left postcentral and precentral areas did not support voluntary motor control after the injury and functional reorganization, but could still pathologically influence spinal motor mechanisms. Ablation of these injured regions reduced spasticity.

(Wilder Penfield & Jasper, 1954) first used the term nociferous to describe the dramatic positive transformation of an aggressive boy after hemispherectomy: “Among patients who have large areas of abnormality in one hemisphere, abnormal behavior may appear, together with advancing mental retardation. The behavioral abnormality is often a more important complaint than the seizures themselves. Radical complete excision may correct the abnormal behavior, stop the seizures, and allow improvement in the patient’s mental state.”

The concept of nociferous cortex has quietly persisted in epileptology, with subsequent studies primarily focusing on cognitive and behavioral outcomes. Of the three tenets to establish cortex nociferous, the first is most straightforward: demonstration that a region is the seizure focus or key element in the epileptogenic network. The second tenet, that the region does not function normally, is more difficult to establish. This is supported by 1) neurological deficits concordant with epileptogenic cortex (e.g., left hemiparesis with a right central seizure focus, episodic memory deficit with left mesial temporal sclerosis), or 2) abnormalities in structural (e.g., MRI) or functional testing (e.g. electroencephalogram (EEG), positron emission tomography (PET), functional MRI (fMRI), magnetic resonance spectroscopy (MRS), intracarotid amobarbital (WADA) test). Variable degrees of functional reorganization, most prominent with early-life neurological insults or seizure onset, can further confound localization of sensorimotor, cognitive or behavioral functions. Even with concordant localizing evidence of dysfunction on the neurological examination, structural and functional assessments, residual function in the epileptogenic cortex cannot be excluded. Overall, the greater the preoperative neurological deficit and concordance across structural and functional measures that a candidate region is abnormal and epileptogenic, the more likely that region has little or no function.

The third tenet of establishing nociferous cortex – improved function after resection of epileptogenic tissue – is complex and difficult to quantify. Resection of epileptogenic tissue likely creates both negative and positive functional outcomes. Improvements can occur in motor (i.e., strength, tone, resolution of involuntary movements), cognitive (e.g., attention, verbal memory, executive functions, social language) and behavior (e.g., mood, irritability, anxiety).

Functional studies support that functional recovery can occur removal of nociferous cortex. After successful temporal lobe resections, MRI spectroscopy studies revealed that N-acetyl-aspartate levels increase in the contralateral temporal lobe, consistent with improved neuronal function (Cendes, Andermann, Dubeau, Matthews, & Arnold, 1997; Serles et al., 2001; Vermathen, Laxer, Schuff, Matson, & Weiner, 2003). PET studies reveal normalization of glucose metabolism in the ipsilateral and contralateral temporal lobes (Hajek et al., 1994; Takahashi et al., 2012). Even when functional improvements follow resective surgeries, it can be difficult to disentangle the contributions of altered interictal epileptiform activity, seizures, diaschisis and antiseizure or psychotropic medications.

Longitudinal studies demonstrate memory improvement over time in some patients

Cognition after surgery is a dynamic and highly variable between subjects. Longer followup intervals reveal further decline in some patients and improvement in others. The meta-analysis of post-surgical cognitive status (Sherman et al., 2011) reporting declines in verbal and visual memory also found some gains in verbal memory (7% in left, 14% in right) and visual memory (15% for left sided, 10% Right). For example, verbal fluency (generating items in a category) generally improved after surgery, with 27% left TLR patients experiencing gains in verbal fluency compared to 10% experiencing losses (Sherman et al., 2011).

Long-term post-surgical memory outcomes may differ depending on the timepoint of assessment after surgery. Studies investigating the long-term followup followed patients from 2 to 10 years after surgery. Early studies following patients from 2 to 5 years after surgery showed ongoing memory decline in LTL surgical patients (C. Helmstaedter et al., 2000; Milner, 1958), including decreases in verbal memory and visual memory between the 1- year and 9-year assessment (Rausch et al., 2003). Patients who underwent left temporal lobe resection experienced decline in the word-pairs delayed recall task, which is sensitive to left hippocampal integrity (Rausch & Babb, 1993; Wood, Saling, O’Shea, Berkovic, & Jackson, 2000). Right TLR patients and a non-surgical control group had verbal and visual memory declines.

Recent European longitudinal studies demonstrated cognitive stability or even improvement after temporal lobectomy (Alpherts, Vermeulen, van Rijen, da Silva, & van Veelen, 2006; Andersson-Roswall, Engman, Samuelsson, & Malmgren, 2010; C. Helmstaedter, Elger, & Vogt, 2018). Immediate decline in verbal memory after dominant TLR group within 2 years was stable at ten years (Alpherts et al., 2006; Andersson-Roswall et al., 2010). By contrast, non-dominant temporal lobectomy resulted in a positive trend in verbal memory after two years. A large European study found that improvements in memory were more common in younger patients who were seizure free or had reduced drug load (C. Helmstaedter et al., 2018).

These differences in longitudinal outcomes have been attributed to differences in patient populations (including age), differing surgical techniques and variable patient attrition (with patients who are doing poorly more likely to continue to follow at a tertiary care center) (C. Helmstaedter et al., 2018). Further, if cognitive outcomes stabilize 1–2 years after surgery, then measurement at 1 and 10 years may mistakenly suggest ongoing cognitive decline.

While there are few pediatric studies with children, evidence supports favorable cognitive outcomes after surgery. Gains in language performance and attention occur postoperatively regardless of side of surgery. Memory performance improved if surgery resulted in seizure freedom (Lendt, Helmstaedter, & Elger, 1999). After TLR, children may recover from post-operative impairment within the first year of surgery (Gleissner, Sassen, Schramm, Elger, & Helmstaedter, 2005). Conversely, ongoing seizures after TLR is associated with declines over time (C. Helmstaedter et al., 2018).

Together, longitudinal studies suggest that if nociferous cortex is removed, cognitive gains are possible. These improvements in verbal and visual memory are associated with seizure freedom after surgery, and suggest that functional recovery is possible. These positive effects may be mediated by decreased deleterious impact of seizures, interictal discharges, or anti-seizure medications. FInally, functional recovery is more common in younger patients (Andersson-Roswall et al., 2010; C. Helmstaedter et al., 2018).

SUMMARY AND CONCLUSIONS

Careful study of the outcomes of temporal lobe epilepsy surgery has greatly advanced our knowledge of the neuroanatomy of human memory. After the initial devastating outcomes, the critical role of the hippocampus and associated medial temporal lobe structures to declarative memory became evident. Surgical approaches quickly changed to become unilateral and later, to be more precise, potentially reducing cognitive morbidity. Neuropsychological studies following unilateral temporal lobe resection have challenged early models which simplified the lateralization of verbal and visual memory function. Diagnostic tests can more accurately lateralize and localize epileptogenic cortex, and predict memory outcomes from surgery. Longitudinal studies have shown that memory may improve in seizure-free patients. Seventy years after HM, we now have a richer understanding of the clinical, neuroimaging, and surgical predictors of memory decline ---and improvement---after temporal lobe resection.

Despite these advancements, we still lack reliable tools to accurately predict an individual’s long-term potential for functional improvement or decline after resective surgery. Individual traits such as age, focality of seizure onset zone, onset and duration of epilepsy, lifetime generalized tonic-clonic seizures, functional status of the epileptogenic cortex, and resection margins all contribute to seizure and cognitive outcomes, but their relative contributions are not understood. Routine and long-term followup after surgery is needed. A model incorporating the likelihood of seizure freedom, early cognitive deficits and long-term cognitive improvements, for an area of resected tissue would enormously advance epilepsy surgery, as well as our understanding of human memory.