Four-Way Wada: SEEG-based mapping with electrical stimulation, high frequency activity, and phase amplitude coupling to complement traditional Wada and functional MRI prior to epilepsy surgery

Abstract

Presurgical evaluation of refractory epilepsy involves functional investigations to minimize postoperative deficit. Assessing language and memory is conventionally undertaken using Wada and fMRI, and occasionally supplemented by data from invasive intracranial electroencephalography, such as electrical stimulation, corticortical evoked potentials, mapping of high frequency activity and phase amplitude coupling. We describe the comparative and complementary role of these methods to inform surgical decision-making and functional prognostication. We used Wada paradigm to standardize testing across all modalities. Postoperative neuropsychological testing confirmed deficit predicted based on these methods.

1. Introduction

Epilepsy surgical candidacy is based on findings from clinical examination and a mixture of well-established techniques (Vakharia, 2018). The use of stereoencephalography (SEEG) in North America is increasing with a parallel rise in minimally invasive epilepsy surgical procedures. SEEG is used to identify seizure onset zone and is complemented by electrocortical stimulation mapping (ESM) for functional localization and to measure connectivity and changes in brainwave activity during cognitive tasks (i.e., passive mapping) (Babajani-Feremi et al., 2018, Ervin, 2020, George et al., 2020).

ESM can either excite or inhibit neural activity and provides information on which brain regions support a cognitive function (Babajani-Feremi et al., 2018). Results highlight patient-specific functional-anatomical correlations and inform potential postsurgical functional deficits (Trébuchon Chauvel, 2016, Young, 2018). However, ESM is laborious, can provoke seizures, and may result in equivocal clinical findings. In contrast, passive mapping via SEEG depth electrodes does not use electrocortical stimulation and instead examines changes in intracranial local oscillatory activity during cognitive, behavioral, and sensory tasks (Arya et al., 2020; Cuisenier et al., 2020). When compared to ESM, high frequency activity (HFA) mapping is less laborious and does not increase the risk of seizure (Drane, 2021). The downside of HFA is that it is an activation rather than inactivation paradigm and does not confirm a given region is essential for a specific function. The combination of ESM and HFA mapping is more practical for surgical planning than when either is used independently. Drane et al. (2021) described ESM as more useful than HFA when mapping language functions involving temporal lobe structures, whereas HFA mapping can be used to identify and exclude sites with higher activation from the ESM procedure, thereby reducing its duration.

In addition to HFA, identifying tissue showing phase amplitude coupling (PAC) during cognitive tasks using SEEG may be a useful predictor of cognitive decline associated with epilepsy surgery. Specific patterns of PAC have been observed during language processing. Theta-gamma bands support bihemispheric processing of syllabic information in children and adults, delta-theta PAC is thought to subserve grouping units of spoken language with unique intonation patterns, and beta-gamma PAC is involved in language comprehension (Giraud, 2012, Meyer, 2018, Pulvermüller et al., 1997). Despite research on PAC associated with language skill, research on its utility for intracranial language mapping remains limited.

We present data on a patient with drug-resistant temporal lobe epilepsy affecting the dominant hemisphere. We uniquely compare the use of four different modalities for functional mapping in addition to the study of functional connectivity.

2. Case report

2.1. History

Our patient was a 19-years-old, right-handed male at the time of presentation. His first seizure was at age 17 years old, which began with a strange smile and cyanosis, followed by bilateral tonic clonic activity. Weekly auras included an “awkward feeling” and right-lip twitching. He was trialed unsuccessfully on multiple antiseizure medications. He was a university senior with no history of academic challenges. Early medical and family history was noncontributory. Over the next several months, he underwent neuropsychological testing, long-term video EEG monitoring, brain MRI, CT/PET scan, functional MRI (fMRI), Wada testing, and SEEG.

2.2. Surgical workup, surgery, and outcome

Presurgical data included MRI, which was negative, and a PET scan that demonstrated left anterior temporal hypometabolism. EEG showed ictal onset in the left anterior temporal region. Functional MRI and Wada indicated left-hemisphere language dominance and bihemispheric verbal and visual memory. In addition to conducting a standard Wada procedure with the patient, we used Wada paradigm for passive mapping and ESM procedures. Given the likely overlap of dominant language and memory networks with the patient’s epileptogenic network, we took additional lengths to characterize the functional-anatomical organization of language and memory.

Interictal SEEG showed near continuous slowing and epileptiform discharges from the left temporal pole (mesial, lateral, and basal regions), and to a lesser extent from the left hippocampus. The ictal network showed an onset in the left temporal pole with instantaneous propagation to the left hippocampus. ESM using Wada paradigm (50 Hz, 3 mA, 300 microseconds, 7 s train) caused naming deficit and world list recall deficit in the pole and hippocampus respectively. CCEPs (corticocortical-evoked potentials) showed robust bidirectional connectivity between the temporal pole, entorhinal cortex, amygdala and hippocampus. Bidirectional connectivity was also seen between the temporal pole and pars triangularis (1 Hz, 5 mA, 300 microseconds, 15 s train).

This HFA/PAC analysis was based on foundational studies (Supplemental file). HFA/PAC using Wada Paradigm showed that word-list learning and word-list recall tasks were associated with Delta-HFA PAC, Beta-HFA PAC, and Theta-HFA and Alpha-HFA across the left pars triangularis, hippocampus, and anterior temporal pole (Figs. 1,2).

Fig. 1.

HFA/PAC comparisons of three tasks: 1) confrontation naming, 2) word-list recall, and 3) word-list encoding with MaxPAC values projected to the cortex including a) left lateral view and b) left inferior oblique view. Analyses are displayed using MaxPAC as Delta-HFA, Theta-HFA, Alpha-HFA, and Beta-HFA.

Fig. 2.

HFA/PAC comparisons of three tasks: 1) confrontation naming, 2) word-list recall, and 3) word-list encoding with fLow of MaxPAC values projected to the cortex including a) left lateral view and b) left inferior oblique view. Analyses are displayed using fLow of MaxPAC as Delta-HFA, Theta-HFA, Alpha-HFA, and Beta-HFA.

The hippocampus could not be spared given its early ictal involvement and strong connectivity to the pole. Based on these findings, left temporal polectomy and hippocampal transections offered the highest probability of improving seizures while preserving language function when compared to a standard temporal lobectomy.

The patient underwent uncomplicated surgery. The only functional change was word-finding problems first reported at a six-month neurosurgical follow-up visit. He has been seizure free since surgery (22 months, Engel Class Ia). Postoperative MRI was consistent with surgical procedure.

Results from a six-month post-operative neuropsychological evaluation (age 20) (Table 1) indicated a statistically notable decline from baseline in confrontation (EVT-3) and receptive naming (PPVT-5) and verbal learning and memory (CVLT-3). When compared to results at six months, findings from a one-year post-operative evaluation showed most skills improved marginally. An exception included a statistically notable improvement in verbal memory following a short delay (CVLT-3). We also compared performances between baseline and one-year post-operative time points. Results showed statistically notable declines in immediate verbal memory (WMS-IV Logical Memory I) and semantic fluency and flexibility (DKEFS Category Fluency, Category Switching). The patient reported good quality of life one year after surgery and had begun driving, completed college, and was pursuing a career.

Table 1.

Comparison of language and auditory-verbal learning and memory findings pre-operatively and post-operatively (Carrow-Woolfolk, 1995, Delis, 2017, Dunn, 2019, Wechsler, 2009a, Williams, 2019).

Measure	Standard Score		RCI	SD	Standard Score		RCI	SD	Standard Score		RCI	SD
wPPVT–5	Preop Baseline	6 Months Postop			6 Months Postop	1 Year Postop			Preop Baseline	1 Year Postop
Total Score	111	98	−2.19	< 1.0	98	105 *	ns	< 1.0	111	105*	ns	< 1.0
EVT– 3
Total Score	102	84	−4.45	> 1.0	84	92 *	ns	< 1.0	102	92*	ns	< 1.0
OWLS–2
Listening Comprehension	121	133	ns	< 1.0	133	118	ns	1.0	121	118	ns	< 1.0
CVLT–3
Trials 1–5 Correct	108	93	–	1.0	93	103 *	ns	< 1.0	108	103*	ns	< 1.0
Short Delay Free Recall Correct	105	80	–	> 1.5	80	105 *	2.19	1.0	105	105*	ns	< 1.0
Long Delay Free Recall Correct	105	85	–	> 1.0	85	105 *	ns	< 1.0	105	105*	ns	< 1.0
WMS-IV Logical Memory
Logical Memory I	105	NA	–	–	NA	95	–	–	105	95	−4.78	< 1.0
Logical Memory II	95	NA	–	–	NA	95	–	–	95	95	ns	< 1.0
DKEFS Verbal Fluency
Letter Fluency: Total Correct	80	NA	–	–	NA	80	–	–	80	80	ns	< 1.0
Category Fluency: Total Correct	105	NA	–	–	NA	85	–	–	105	85	−9.83	> 1.0
Category Switching: Correct Responses	105	NA	–	–	NA	90	–	–	105	90	−4.09	1.0
Category Switching: Switching Accuracy	110	NA	–	–	NA	95	–	–	110	95	−4.06	1.0

Note: An

^*indicates an alternate form was used to control for practice effects due to test-retest < 1 year.

RCI = Reliable change index > 1.96 indicates statistically notable change over time and is presented in bold.

SD = Standard deviation > 1.0 SD indicates statistically notable change over time and is presented in bold.

NA = Not administered at six-month postoperative evaluation.

ns = Nonsignificant.

3. Discussion

Our report exemplifies how traditional methods can be complemented by novel techniques using SEEG for the betterment of patients’ postsurgical outcome. We illustrate herein that HFA and PAC can adequately map cortical functions in sampled areas when compared to Wada, fMRI and ESM. To our knowledge, this is the first comparative description of five complementary functional assessment modalities performed using Wada paradigm: fMRI, Wada, passive mapping (HFA/PAC), CCEPs, and ESM.

While left language dominance was noted on fMRI and Wada, ESM localized this more precisely to the temporal pole and hippocampus. Passive HFA/PAC mapping findings (Figs. 1, 2) were concordant with ESM and suggested that resection involving the left anterior temporal pole and left hippocampus could negatively affect expressive language. A resection of these regions will lead to language deficit but this was somewhat mitigated by doing hippocampal transection rather than resection. Moreover, CCEPS and PAC indicated the presence of functional connectivity, more clearly than seen in ESM, between the temporal pole and pars triangularis indicating that a disconnection of one of those areas could further affect language function.

Our PAC results are consistent with previous studies indicating language function correlates to low frequency PAC and HFA (e.g., gamma) (Brennan and Martin, 2020, Karthik, 2021, Lizarazu, 2019). While other studies have focused on a more narrow range of higher frequency activity (gamma, high gamma), our study design intentionally used a wider range of HFA (30–300 hertz). The novel aspect of this analysis is the exploration of a wider range as opposed to restricted range, to better visualize possible HFA that is spread across higher frequencies, versus strict restriction to pre-established narrow band HFA (apriori). Further, our findings corroborate language function as involving neural network interplay that includes spatial (i.e., interregional) and spectral (i.e., phase-amplitude) dynamics, as well as lower order, and more mechanical contributors, versus the higher order HFA components.

One major limitation of passive mapping with SEEG is its inability to sample all tissue. However, passive mapping is less labor intensive and allows more extensive testing of various aspects of language and memory functions. Cortical stimulation is considered the gold standard for language mapping, but it can take a long time to do extensive testing and can lead to false positive or negative results. Hippocampal stimulation may falsely produce seizures, even when it is not critically involved in the early ictal network, and therefore, HFA mapping may offer an excellent alternative to mapping the hippocampus with cortical stimulation. Given strengths and weaknesses of both methods, researchers have recently suggested they be used in a complementary manner to evaluate epilepsy surgery candidates (Drane, 2021). Further studies will be required to rigorously compare methods and determine if the use of multiple methods improves our understanding of functional network localization.

Uniquely, our study highlights how traditional techniques used to determine epilepsy surgery candidacy can be supplemented by SEEG methodologies for optimal surgical planning that limits postsurgical functional deficits thereby improving patients’ long-term outcomes. Additional studies are needed to validate ESM and HFA/PAC mapping with SEEG for functional memory mapping. In addition, future studies should refine the approach to language mapping using SEEG to include more than single-domain language tasks, which may more comprehensively capture brain networks involved in language.

Data Availability

Data are available upon reasonable request.