Cortical activation with deep brain stimulation of the anterior thalamus for epilepsy

Abstract

Objective

We studied the relation between thalamic stimulation parameters and the morphology, topographic distribution and cortical sources of the cerebral responses in patients with intractable epilepsy undergoing deep brain stimulation (DBS) of the thalamus.

Methods

Bipolar and monopolar stimuli were delivered at a rate of 2 Hz to the anterior (AN, four patients), the dorsomedian (DM, four patients), and the centromedian nucleus (CM, one patient) using the programmable stimulation device (Medtronic ITREL II). Source modeling was carried out by using statistical non-parametric mapping of low-resolution electromagnetic tomography (LORETA) values.

Results

All patients demonstrated reproducible time-locked cortical responses (CRs) consisting of a sequence of components with latencies between 20 and 320 ms. The morphology of these CRs, however, was very heterogeneous, depending primarily on the site of stimulation. Following AN stimulation, cortical activation was most prominent in ipsilateral cingulate gyrus, insular cortex and lateral neocortical temporal structures. Stimulation of the DM mainly showed activation of the ipsilateral orbitofrontal and mesial and lateral frontal areas, but also involvement of mesial temporal structures. Stimulation of the CM showed a rather diffuse (though still mainly ipsilateral) increase of cortical activity. The magnitude of cortical activation was positively related to the strength of the stimulus and inversely related to the impedance of the electrode.

Conclusions

The pattern of cortical activation corresponded with the hodology of the involved structures and may underscore the importance of optimal localization of DBS electrodes in patients with epilepsy.

Significance

The method of analyzing sources of CRs could potentially be a useful tool for titration of DBS parameters in patients with electrode contacts in clinically silent areas. Furthermore, the inverse relation of the cortical activation and the impedance of the electrode contacts might suggest that these impedance measurements should be taken into consideration when adjusting DBS parameters in patients with epilepsy.

Introduction

Approximately, 25% of patients with epilepsy remain poorly controlled despite antiepileptic treatment and are not eligible for resective surgery because the seizure focus cannot be identified, or the epilepsy is multifocal, or involves eloquent brain areas (Sillanpaa et al., 1998). Novel therapeutic strategies are required for this group. Electrical stimulation of discrete brain structures has been considered as a possible means of preventing the initiation or early propagation of seizure activity. Since the 1970s, several targets, including the cerebellum (Cooper, 1973), the centromedian thalamic nucleus (CM; Fisher et al., 1992, Velasco et al., 1987, Sussman et al., 1988), the anterior thalamic nucleus (AN; Cooper et al., 1984, Hodaie et al., 2002, Kerrigan et al., 2004, Upton et al., 1985), the caudate nucleus (Chkhenkeli and Chkhenkeli, 1997), the subthalamic nucleus (Benabid et al., 2002, Dinner et al., 2002), and the hippocampus (Velasco et al., 2000, Velasco et al., 2001, Yamamoto et al., 2002) have been stimulated in attempts to suppress seizures. Moreover, epileptiform afterdischarges could be terminated by direct stimulation of the epileptogenic focus (Kayyali and Durand, 1991, Lesser et al., 1999) and results of metabolic imaging studies indicated the involvement of the basal ganglia in the propagation and spread of seizures (Franceschi et al., 1995, Rodrigues et al., 1996).

The CM was considered as a target for therapeutic stimulation based on the results from early experimental stimulation studies (Hunter and Jasper, 1949, Starzl et al., 1951). Unblinded clinical studies of CM stimulation in patients with various intractable epilepsies reported significant benefits, particularly in patients with generalized tonic-clonic seizures and atypical absences (Velasco et al., 1987, Velasco et al., 1993), although a small controlled trial showed no significant benefit (Fisher et al., 1992). Several observations have indicated that the anterior thalamic region plays an important role in the maintenance and propagation of seizures. For example, the anterior thalamus has been reported to show increased metabolic activity during seizures (Mirski and Ferrendelli, 1986), and lesioning (Mirski and Ferrendelli, 1984) or high-frequency stimulation (Hamani et al., 2004, Mirski and Fisher, 1994) in AN or its afferent pathways can reduce seizure activity in models of epileptogenesis in experimental animals. The dorsomedial nucleus of the thalamus (DM), situated posterior and inferior to AN, has also been shown to be intimately involved in the maintenance and propagation of seizures, specifically those involving limbic brain structures (Bertram et al., 2001). The AN as a target for stimulation to treat epilepsy in man was pioneered by Cooper and colleagues (Cooper et al., 1984, Upton et al., 1985). Subsequent unblinded pilot trials have shown that chronic anterior thalamus stimulation resulted in a statistically significant decrease in seizure frequency (Hodaie et al., 2002, Kerrigan et al., 2004, Sussman et al., 1988). The results of these small series have rekindled interest in the treatment of epilepsy by electrical stimulation and have led to the initiation of a prospective trial across multiple centers in the United States (Graves et al., 2004).

Yet, the mechanisms by which thalamic stimulation through implanted DBS electrodes may reduce seizure frequency in patients with epilepsy are not clear. Therapeutic stimulation has typically used pulses of intermittent high-frequency stimulation conceptualized to ‘desynchronize’ the same thalamocortical networks that are prone to synchronization in epilepsy and in response to low-frequency stimulation. However, no stimulation parameters or intrathalamic localizations of specific clinical benefit have been identified to date to guide treatment (Hodaie et al., 2002, Kerrigan et al., 2004). Though clinically unproven, it can be speculated that the positive effect of anterior thalamus stimulation might be due to the wide connections of these nuclei with limbic structures. With low-frequency stimulation, it has been shown that stimulation of the AN synchronizes the pattern of electroencephalographic (EEG) activity and leads to the generation of cortical responses (CRs) (Steriade, 1997). CRs are evoked by direct stimulation of various subcortical structures including specific and unspecific thalamic nuclei, or by direct or antidromic excitation of pathways connecting these structures with the cortex. According to the dichotomous concept of the specific and unspecific thalamic projection system of Dempsey and Morison (1942), stimulation of the specific thalamic relay nuclei, such as the ventrolateral or ventrobasal complex, evokes a so-called augmenting rhythm, whereas stimulation of the unspecific thalamic relay nuclei, such as the intralaminar nuclei or the centromedian nucleus, leads to a so-called recruiting rhythm (Morison and Dempsey, 1942, Morison and Dempsey, 1943).

The augmenting response upon low threshold stimulation of the specific nuclei typically has a short latency, is surface-positive, and is followed by one or, with increasing stimulus strength, several slow waves (the so-called secondary response). The augmenting response is generally well detectable to the first stimulus, and the amplitude may decrease on prolonged repetitive stimulation (Eccles, 1951, Ganglberger et al., 1969). This response is mediated through monosynaptic pathways of thick fibers with a specific point-to-point arrangement, and is, thus, topographically limited to related cortical projection areas. Using laminar field study methods, the generator(s) of the augmenting response could be attributed to deep or middle cortical layers (the phase reversal to negative occurs at a cortical depth of 0.25 mm or more). Consequently, these responses have been named deep thalamocortical responses (Sasaki et al., 1970). Anatomically, these precise, area-specific projections to deep or middle cortical layers can be best attributed to a core of parvalbumin-immunoreactive neurons. Core neurons are superimposed upon the matrix in some thalamic nuclei (sensory relay nuclei such as the ventrolateral nucleus are prototypical), and project in a topographically ordered fashion to deep or middle layers of the cortex in an area-specific manner (Jones and Hendry, 1989).

The recruiting rhythm following stimulation of the unspecific nuclei, on the other hand, is surface-negative and shows a latency between stimulus and onset of the initial deflection of 20–40 ms. The response typically builds up or recruits on repetitive stimulation (hence the name recruiting response), with usually little or no response to the initial stimulus. The response is mediated through poorly defined, mostly polysynaptic pathways of fine fibers with diffuse cortical connections, thus causing a widespread topographical distribution (Jasper et al., 1953). The generator(s) of the recruiting response could be attributed to superficial cortical layers (the phase reversal to positive occurs at a cortical depth of 0.1–0.25 mm) using laminar field study methods. These responses have thus been named superficial thalamocortical responses (Sasaki et al., 1970). Anatomically, this widespread superficial thalamocortical network can be best attributed to a matrix of calbindin-immunoreactive neurons. This matrix extends throughout the (dorsal) thalamus, and its neurons project to superficial layers of cortex over wide areas, unconstrained by nuclear borders or by distinctions between intralaminar, relay or other nuclei (Jones and Hendry, 1989).

In patients with anterior thalamic DBS for the treatment of epilepsy, the generation of CRs following stimulation has been suggested to predict clinical efficacy of stimulation (Hodaie et al., 2002). The exact pattern of cortical involvement following AN stimulation in humans, however, is largely unknown. Where does stimulation spread? What is the relationship between the CRs and stimulation parameters, and what parameters are optimal? In order to try to clarify these questions we investigated the relation between thalamic stimulation parameters and the cortical sources of the CR by using LORETA, which is one of several distributed linear source algorithms so far available (Pascual-Marqui, 1999, Pascual-Marqui et al., 1994). Given the supposedly complex nature of cortical activity following thalamic stimulation, LORETA might be an ideal tool to study such activity. Unlike dipole model strategies, LORETA takes into account that cerebral sources are extended areas of active cortex rather than point-like dipoles. The localization accuracy of LORETA has been repeatedly validated in previous studies (Anderer et al., 1998, Pizzagalli et al., 2000). In patients with limbic epilepsy, mesial temporal interictal epileptiform discharges could be reliably detected by using the non-parametric mapping of LORETA values (Zumsteg et al., in press). Our goals were to define the characteristics of the CR following electrical stimulation of various sites of the thalamus, and to the reveal the cortical sources of these CRs.