Intraoperative mapping of executive function using electrocorticography for patients with low-grade gliomas

Surgical resection for diffuse low-grade gliomas (dLGG) involves a fine balance between maximal extent of resection and preserving cognitive function [1]. A prominent effort in the field in recent years has been dedicated to extending the range of cognitive domains that are functionally mapped intraoperatively to optimize this onco-functional balance [2, 3]. In many centers, intraoperative functional mapping of some aspects of language is already done as part of standard care in an awake procedure using direct electrical stimulation (DES), in addition to mapping of motor and sensory pathways. Several other tasks have been proposed for the mapping of a range of cognitive domains, such as visuospatial functions, verbal memory, calculation, dual tasking, and executive function (EF) [4‐10]. Among these, the mapping of EF is particularly challenging, and only a handful of studies are reported in the literature. These studies used DES with the Stroop task that tests for one aspect of EF, namely, conflict monitoring [8‐10]. Wager et al. [10] reported effects on performance in the Stroop task when DES was applied on the anterior cingulate cortex. Puglisi et al. [8, 9] found that DES led to impairments in the Stroop task intraoperatively only when subcortical sites were stimulated, but not cortical ones. These DES-positive sites were confined to white matter tracts located under the inferior and middle frontal gyri. Furthermore, they reported that patients who underwent intraoperative mapping with the Stroop task had less EF deficits on average compared with a control group following surgery, most in verbal fluency. Nevertheless, beyond this sparse evidence, no established tests for mapping of EF have been proposed.

Several considerations may limit the ability to map EF intraoperatively. First, EFs are inherently complex, referring to a wide range of functions such as attention, reasoning, dual-tasking, conflict monitoring, hierarchical planning, and allocation of resources. Given the time constraints during surgical procedures and the complexity of intraoperative testing, it is impossible to test for all these aspects of EF. Second, it is widely accepted that EF is supported by a distributed network of regions across the frontal and parietal cortices. However, the detailed role of each of the regions or sub-networks within this network is not yet known. Third, unlike the primary motor cortex, for example, which is anatomically well-defined, the neuroanatomical markers for the frontoparietal network are only loosely defined and cover large cortical areas that may be hard to identify in full during surgery.

Robust evidence from studies that use functional magnetic resonance imaging (fMRI) data in healthy participants reveal consistent recruitment of a frontoparietal cortical network involved in control processes. This network is recruited when task difficulty level increases across a wide range of cognitive domains, such as spatial working memory, verbal working memory, maths, and multisource response interference, and has been termed the “multiple demand” (MD) network [11‐14]. This network shows similar activation patterns across the two hemispheres and includes an anterior-posterior axis along the middle frontal gyrus (MFG), the posterior dorso-lateral frontal cortex (pdLFC), the anterior insula and adjacent frontal operculum (AI/FO), the pre-supplementary motor area (pre-SMA) and dorsal anterior cingulate cortex (preSMA/ACC), and the intraparietal sulcus (IPS) in the parietal lobe. Similar regions have been reported to be involved in multiple aspects of cognitive control across many other neuroimaging studies [15‐21].

Task-based fMRI can be used to functionally identify the MD network in individual participants. In this procedure, neural activity evoked by a given task (e.g., spatial working memory) is contrasted with an easier version of the same task. Such contrast allows for manipulating task difficulty while controlling for other task factors such as stimuli, responses, and duration of trials. While the group-level activation map is highly consistent, substantial differences in activity patterns of individuals are observed [14, 22], demonstrating that neuroanatomical markers are insufficient to reliably identify the control network in individuals. Furthermore, at the individual participant level, MD regions in the frontal cortex can be dissociated from adjacent areas involved in language processing [23, 24].

Here, we propose a novel approach complementary to DES for intraoperative mapping EF using recordings of activity directly from the surface of the brain using electrocorticography (ECOG). DES and non-stimulatory ECOG data provide different types of information. To ascertain that a stimulated brain area/locus is subservient to EF is time consuming, which is a significant constraint in a time-limited surgery. Functional mapping using ECOG data, either from the surface of the brain or using depth electrodes, has the potential to narrow down the areas to subsequently focus on using DES. We build on neuroimaging evidence and use a similar task difficulty manipulation to identify brain areas that are associated with EF using ECOG. As demonstrated using fMRI, increased task difficulty leads to increased recruitment of attentional resources and therefore increased activity in the control network. By using such task difficulty manipulation, this approach probes the recruitment of networks involved in multiple core aspects of EF rather than tapping on a particular aspect of EF. Adapting tasks to the intraoperative setting, we used two counting tasks with varying difficulty levels: counting from 1 to 20 (simple counting, easy task) and counting while switching between letters and numbers 1-a-2-b-3-c… (switch counting, hard task) [25]. As a first step to investigating the utility and feasibility of this approach, functional measures of the recorded ECOG signals were analyzed offline and contrasted between the two tasks to identify a neural signature associated with increased task difficulty. Based on previously demonstrated links between the fMRI BOLD response and power in the broadband high-gamma frequency range (70–250 Hz) [26‐28], we predicted that increased task difficulty will be associated with increases in high-gamma power in the recorded ECOG signals. In this paper, we describe this approach in detail and how it can be implemented in the operating theater. We then demonstrate the potential of its use in three patients. Finally, we discuss the steps required to further establish the validity of this approach and its real-time clinical use, intraoperatively.

We demonstrate the feasibility of using ECOG for the intraoperative mapping of EF in three patients with diverse clinical profiles. The clinical care and surgery of the patients was not affected by the electrophysiological data and EF mapping presented here. Data were recorded from five electrode strips with four contacts each placed on the frontal cortex, resulting in data from 15 channels overall following bipolar re-referencing. Eleven out of the 15 channels (73.3%) showed significant increases in high-gamma power with increased task difficulty, 4 channels (26.6%) showed no change in power, and none of the channels showed a power decrease. The increases in power were consistent with our prediction and suggest that the recorded areas are associated with EF more generally. Overall, changes in high-gamma power with increased task difficulty were in partial overlap with the FTPN template, with increases being more likely in areas that were within the FTPN template and less likely outside it. Out of the 11 channels that showed power increases, 9 (81.8%) were located within the FTPN template, and only 2 (18.2%) were located outside the template. Of the 4 channels that showed no change in power, 3 (75%) were located outside the FTPN template and only 1 (25%) within the FTPN. These results provide support for the link between the observed changes in high-gamma power and EF-related regions. Below, we describe in detail the clinical profile and electrophysiology results for each of the three patients.

We developed a novel approach for intraoperative functional mapping of EF using ECOG and demonstrated its feasibility and applicability in three patients undergoing awake surgery for the resection of low-grade glioma. This approach is based on the rationale of recruiting areas that are associated with EF by manipulating task demands. We described in detail the procedure and implementation of this approach and showed that increases in broadband high-gamma power with increased task difficulty in two counting tasks may be related to areas that are associated with EF. We emphasize the importance of intraoperative testing for EF and propose that electrophysiological signals may be used intraoperatively complementarily to DES to guide and make the functional procedure more effective, efficient, and fast. Our results serve as “proof of concept,” and future research will be required to establish its validity and benefit in clinical practice.

We demonstrated the use of this approach in three patients with diverse clinical and cognitive characteristics. Patient 1 had a temporal tumor, away from the recording locations in the frontal lobe; therefore, electrophysiology activity likely reflects normal responses to the task difficulty manipulation, within surgical limitations and possible anesthetic effects. Indeed, five out of the six channels showed the expected electrophysiological signature of EF in the form of increases in high-gamma power with increased task difficulty, and these channels were located within the FTPN template. The only channel that did not show such an increase was located outside the FTPN template. None of the recording areas was resected or in proximity to the resected tissue in the temporal lobe, and EF and attention performance following surgery were preserved, or even mildly improved. While this is in line with evidence for the role of frontal and parietal areas in EF, it has been recently shown that the temporal lobe is implicated specifically in set-shifting during a trail making test [34, 35], and further studies will be required to clarify the involvement of the temporal lobe in EF. The case of patient 2 demonstrates the potential benefit of intraoperative EF monitoring using ECOG in areas around the tumor for preventing post-surgery EF deficits while maximizing the extent of resection. The tumor was located in the middle frontal gyrus, an area that is associated with EF and attentional processes. The tissue at the vicinity of the tumor, where resection could extend to, may also be involved in such cognitive functions, either because it was involved in EF before the tumor developed and/or because of plasticity processes and potential migration of functionality due to tumor growth. Therefore, mapping EF in areas adjacent to the tumor could contribute to refining the resected areas to preserve EF following surgery. Indeed, although the patient did not have EF or attention impairments prior to surgery, such deficits were identified in the post-surgery assessment and were self-reported in the 6-month follow-up appointment. Electrophysiology data were recorded from the inferior frontal gyrus in areas largely considered as part of the EF network and in partial overlap with the resting-state FTPN and showed increases in high-gamma power with increased task difficulty in some of the channels. Possibly, areas in close proximity to the tumor and within the EF-related network exhibit similar response profiles that could be identified using ECOG. In patient 3, the recording locations were mostly areas that are classically associated with EF. The insular tumor was further away from the areas where electrophysiology data was recorded, suggesting that anatomo-functional relations in the frontal cortex are unlikely to change because of tumor growth and that the neural activity recorded from the frontal cortex likely reflects activity in normal tissue, in line with the results. However, the insula is considered as part of the EF/control network, and possible impairment of the FTPN during the resection of the insula may have contributed to the mild deficits in processing speed that were self-reported by the patient’s partner following surgery.

The intraoperative mapping of EF is considered to be challenging and difficult and has been so far addressed in only a handful of studies using DES with Stroop or trail making tasks as single aspects of EF [8‐10, 35]. We showed that using counting tasks with increased difficulty while recording activity directly from the surface of the brain using ECOG is a feasible approach that can be conducted intraoperatively. At the behavioral level, patients were able to perform the tasks well. For the functional measures of the recorded brain signals, we build on previous evidence from fMRI in healthy participants that show consistent and robust recruitment of the frontoparietal control network when task demands increase [11, 14, 20, 22]. Rather than a specific aspect of EF, our approach focuses on the recruitment of the control network as has been observed in neuroimaging studies. Supported by ample evidence, this network has been associated with a variety of cognitive functions associated with EF and across multiple cognitive domains. These include working memory, maths, conflict monitoring, complexity, and representation of behaviorally relevant information such as context, stimulus information, and decisions [14, 18‐21, 36, 37]. In accordance with these neuroimaging findings, we used a similar task difficulty manipulation and measured changes in intracranial EEG signals. We focused on power in the broadband high-gamma frequency range (70–250 Hz), which has been previously linked to the fMRI BOLD response [26‐28]. Power increases in the gamma band have been associated with a range of cognitive functions including attention, working memory load, rule abstraction, and stimulus processing [38‐43]. Measures of power in the gamma range have been used to control movement in brain-computer interfaces [44, 45] as well as for mapping in the motor and language domains [46‐49]. Here, we demonstrated that power increases in the gamma range are also observed for increased task difficulty in the frontal cortex, and these can be measured using ECOG during an awake surgery procedure, serving as an approximate measure for the recruitment of the control network.

In this study, we recorded the data intraoperatively and analyzed it offline in order to establish the feasibility and utility of this approach. Ultimately, the aim is to provide neurosurgeons with feedback in real time during surgery to identify areas that are most likely to be associated with EF and guide the use of DES toward these areas. Moreover, DES and ECOG data measure different neural responses. DES provides information about the causal role of the stimulated area in the tested function and the potential outcome of resection. In contrast, the electrophysiological signal as measured with ECOG provides evidence for the involvement of the recorded area in a certain function but not if it is essential for this function. While DES serves as the gold standard in awake function mapping, the ability to detect effects of stimulation depends on multiple factors, which may make it challenging for EF in particular. The applied current intensity is usually determined based on motor threshold and may vary between brain regions which may require a higher intensity (and potentially unsafe) to interrupt behavioral responses. For example, effects of DES on performance in the Stroop task have been demonstrated when white matter tracts were stimulated, but not gray matter tissue [8, 9], in line with the higher excitability of white matter [50]. Conversely, a larger area of the brain may be stimulated with higher current settings. Additionally, behavioral performance in some cognitive functions such as EF may depend on a distributed network of regions rather than a focal area that can be stimulated. Therefore, DES of a single area might not result in detectable behavioral impairments intraoperatively, yet post-surgery outcomes may show cognitive deficits with potentially impactful effects on quality of life of patients. Indeed, it has been demonstrated that partial damage to the frontoparietal control network in stroke patients is associated with reduced cognitive ability as measured with IQ [51]. Finally, because of the complex nature of EF, DES that is used in conjunction with a specific task, such as the Stroop that reflects one aspect of EF, may not generalize to other aspects of EF. Altogether, despite the clear benefits of DES in clinical practice for the standard mapping of motor and language-related functions, its use for mapping of EF poses substantial difficulties that may result in some EF-related areas that are missed out. Using ECOG to guide and complement DES may provide useful information to further improve and optimize the functional mapping procedure, and this was the primary clinical motivation for this study. We note, however, that despite the additional information that ECOG may offer, this technique is not suitable for monitoring EF in white matter, in contrast to DES [8, 9, 35]. We emphasize that using ECOG for mapping EF during awake surgery, as we demonstrate here, is not intended to be a substitute for DES mapping. DES remains the gold standard for deciding whether any brain tissue should or should not be removed. ECOG data acquired while patients perform a task may provide important information that is needed to preserve certain functions, such as EF, that cannot be readily assessed in a few seconds following DES. Here, we identified an ECOG signal associated with EF, providing initial evidence that can be further pursued in the field.

The data reported here serve as the first step toward developing ECOG as a tool for intraoperative mapping of EF. We demonstrated the feasibility of this approach, but a careful and rigorous evidence-based investigation at multiple levels is required to further validate and establish this approach and its clinical utility before it can be used routinely. First, a larger sample size and better coverage of the frontal cortex is necessary to provide solid evidence for increases in high-gamma activity with increased demand in the frontal cortex. Linking such increases as measured with ECOG to evidence of frontal activity from other modalities such as neuroimaging will further strengthen such evidence. Specifically, investigating the relationship between ECOG measures and task-based fMRI BOLD response associated with EF may contribute to better understanding of the potential and limitations of using fMRI for pre-operative planning and more generally inform about how the two signals relate to one another. Importantly, conjunctive evidence from DES and ECOG will shed light on the relationship between the two and, in particular, whether areas with increased high-gamma activity when task demands increase as measured by ECOG indeed show higher probability of exerting DES effects. Increases in high-gamma activity with increased task demands may also be related to post-surgery EF deficits, for example, in resected areas adjacent to the tumor. An essential step in bringing our proposed technique into use in clinical practice would involve establishing the importance of preserving areas associated with EF to long-term clinical outcome [52]. While deficits associated with EF following lesions in fronto-parietal areas have been demonstrated in patients with stroke [51], in patients with dLGG, there is remarkable long-term recovery through network reorganization and plasticity following surgery, which is not yet well understood. Ultimately, a controlled trial will be required to demonstrate the potential benefit of using ECOG for intraoperative mapping of EF, for example, in the form of a larger extent of resection and improved cognitive outcome post-surgery and as a result overall better quality of life of patients.

We have shown the feasibility of intraoperative ECOG as a tool for mapping EF, and we believe that it has a potential to augment intraoperative functional brain mapping by guiding and supporting DES. While we used electrodes that were placed on the surface of the brain, similar data can be obtained using depth electrodes, offering access to deeper layers and structures before resection and the possibility of identifying their involvement in the tested cognitive functions. Beyond EF, intraoperative mapping using ECOG and in combination with DES offers opportunities for expansion to other cognitive domains, as part of the growing interest in individually tailored treatments that can be offered to patients.