Large-scale brain networks and intra-axial tumor surgery: a narrative review of functional mapping techniques, critical needs, and scientific opportunities

2.1. Somatomotor network

Topographically, the somatomotor network is defined largely by the pre- and post-central gyri (i.e., peri-rolandic area) that execute movement and process sensation (Yeo et al., 2011). This network is commonly considered “eloquent” (Penfield et al., 1937; Burchiel et al., 1989) due to the potential for permanent and debilitating neurological deficits if injured (Kandel et al., 2012). Classically, its somatotopy has been defined as a stereotyped medial-to-lateral modular progression of body parts from distal lower extremity to face (Roux et al., 2018, 2020). However, newer connectomic-based discoveries suggest that previously unknown somatomotor association areas may exist between and modulate classical effector regions to facilitate whole-body motor planning (Gordon et al., 2023). The clinical significance of this finding is unknown, but the discovery of these interdigitated areas and its associated redefinition of the classical homunculus may reveal more neuroplastic potential for this network than previously recognized.

Clinically, the somatomotor region has demonstrated only minimal plastic potential in adults (Picart et al., 2019). Because of this, portions of tumors invading the precentral gyrus are sometimes inoperable, depending on the tumor type (Seidel et al., 2013). Unsurprisingly, reduced connectivity between homotopic pairs of cortical parcels within the somatomotor network is associated with reduced overall survival (Daniel et al., 2021), as is reduced structural connectivity between neighboring regions (Wei et al., 2022). Likewise, tumors invading the corticospinal tract in otherwise healthy individuals correlates with worse overall survival (Mickevicius et al., 2015). Therefore, an adaptation of lesion connectome mapping (i.e., “survival connectome mapping”) may provide valuable prognostic information.

Additionally, somatomotor structural and behavioral alterations associated with chemo and radiation therapies is being investigated (Seibert et al., 2017). While these therapies are known to result in decreased cortical thickness (i.e., gray matter loss), associated connectomic changes are newer targets of investigation (Seibert et al., 2017; Mitchell et al., 2020; Daniel et al., 2022; Zhu et al., 2022).

2.2. Language network

Unlike many of the canonical resting-state networks (i.e., somatomotor, default mode, dorsal attention, etc.), the language network is best described functionally based on various language tasks (Aabedi et al., 2022). Topographically, it contains the classic Broca’s and Wernicke’s areas with the arcuate fasciculus joining them, as well as their dorsal and ventral processing streams (Chang et al., 2015). The language network is also considered “eloquent” in that damage to its key components can result in permanent debilitating deficits. Unlike the somatomotor network, language areas appear more widely distributed and can be highly plastic under the right circumstances (Picart et al., 2019; Aabedi et al., 2022). This plasticity can involve increases in activation of the right hemisphere and cerebellum; however, language functions can also relocate to “atypical” regions of the frontal and temporal lobes ipsilesionally (Thiel et al., 2001). Interestingly, recent evidence suggests that language network plasticity may be facilitated to extend surgical resections by intracranial electrical stimulation and prehabilitation (Rivera-Rivera et al., 2017). For those interested in a more detailed review of language network plasticity, we refer the readers to a recent review (Pasquini et al., 2022).

2.3. Cognitive networks

While certain cortical processes (i.e., language and motor output) can be easily mapped intraoperatively with direct electrical stimulation (see section “3.1.1. Direct electrical stimulation”), more complex and abstract cognitive processes [i.e., working memory, attention, decision making, etc., (Buckner et al., 2008)] are more difficult to pinpoint. Consideration of such networks is critical as these networks may be directly affected in as much as 80% of patients with brain tumors (Morell et al., 2022). However, the difficulty in assessing them intraoperatively, difficulty in predicting long-term outcomes, and the unpredictable natural history of many postoperative cognitive deficits makes it uncommon to consider the networks in standard of care surgery (beyond avoiding operating on bilateral lesions). Table 1 represents a brief summary of the most well-known cognitive networks, their associated anatomical contributions, functions, and pathologies. Figure 1 provides a visual representation of these networks. For a more detailed review of these cognitive networks, we refer readers to the reviews and articles listed in the table.

3.1. Brain stimulation

3.2. Electromagnetic recording

In this section, we review several common electromagnetic recording techniques followed by an overall comparison. EEG, MEG, and electrocorticography (ECoG) capture signals that derive from the same underlying electrical activity of the cerebral cortex: ionic currents generated by summated excitatory and inhibitory post-synaptic potentials in the most superficial layers of the cerebral cortex.

3.3. Magnetic resonance imaging

3.3.2. Functional MRI

Functional MRI (fMRI) images the blood oxygen level dependent (BOLD) hemodynamic response to neural activity (Biswal et al., 1995). fMRI has become ubiquitous in tertiary care centers as a non-invasive marker of large-scale network topography and, more recently, connectivity. The utility of fMRI rests in the physiological coupling between neuronal electrical activity, cellular metabolism, and blood flow (Agarwal et al., 2021). That is, neuronal depolarizations necessitate oxygen delivery which increases local blood flow. Therefore, fMRI indirectly measures brain activity via fluctuations in deoxygenated blood (Agarwal et al., 2021). Conventionally, sampling the whole brain requires ∼2–3 s (0.33–0.5 Hz) to complete; however, some technological advances have allowed whole-brain sampling as quickly as ∼0.72 s (1.39 Hz) or lower as in the Human Connectome Project (Constable and Spencer, 2001; Van Essen et al., 2013). This limitation in sampling rate is obviated by the slowness of the hemodynamic response, as the peak power in the hemodynamic signal is below 0.1 Hz (Biswal et al., 1995).

Studies examining large-scale networks using fMRI typically employ a resting-state or task-based design or combine these approaches (Van Essen et al., 2013; Cole et al., 2021; Vinehout et al., 2022). In the resting-state, several canonical networks (i.e., default mode, dorsal attention, sensorimotor, etc.) are readily identifiable (Smith et al., 2009; Schaefer et al., 2018). During tasks, connectivity within networks necessary for completing the task (i.e., “task-positive” networks) demonstrate increased connectivity relative to rest, whereas networks unrelated to the task demonstrate decreased connectivity (Kalinosky et al., 2019; Cole et al., 2021; Vinehout et al., 2022).

4.1. Electrophysiological techniques

Since Dr. Penfield, DES has been used in standard clinical practice to map positive and negative responses across cortical and subcortical areas (Aaronson et al., 2021). One group with extensive experience in this area has generated a probabilistic atlas of numerous functional perturbations across sensorimotor and language domains throughout the brain (Sarubbo et al., 2020). Importantly, while this work has often confirmed the conventional motor and sensory distributions in the pre- and post-central gyri, respectively, it did also identify some reciprocity more recently (thought to be indicative of U-shaped fibers) (Sarubbo et al., 2020), and newer techniques suggest that our knowledge of the somatotopy of the motor homunculus may be incomplete (Jensen et al., 2022; Gordon et al., 2023).

Focal slow rhythms in EEG have been recognized with brain tumors and other structural lesions from the earliest days of EEG. In 1936 using an array of just 3 scalp electrodes, Walter (1936) showed the presence of focal slow waves that he named “delta” in the vicinity of brain tumors. Subsequent studies determined that slow activity in the delta (0.5 to 4.0 Hz) and theta (4 to 8 Hz) frequency ranges can be seen in association with a variety of conditions including tumors, traumatic brain injury, degenerative disorders, hypoxic states, and metabolic disturbances (Joynt et al., 1965; Fischer-Williams and Dike, 2005). Power-ratio analyses indicate that the increase of low-frequency power around tumors is typically also associated with a loss of high-frequency power (Nagata et al., 1985).

A few recent studies have applied functional-connectivity estimation methods to study peritumoral networks with EEG. Higher pre-operative local and global functional connectivity in patients with low-grade gliomas or meningiomas in the language-dominant hemisphere appears to predict worse language outcomes on 1-year follow-up (Wolthuis et al., 2021). Although changes in EEG power across frequency bands tend to be most salient in the vicinity of brain tumors, a recent study found that cognition in patients with left frontal gliomas was negatively correlated with beta frequency band connectivity between the contralateral (right) anterior prefrontal and parietal lobules (Ueda et al., 2022). Remote changes in functional connectivity may seem puzzling at first glance, but there is evidence supporting the notion that increased responses and enhanced functional connectivity in regions far from the tumor location may be related to functional compensatory mechanisms (Yang et al., 2022).

In the context of tumor surgery, ECoG has historically been used to identify epileptogenic peritumoral areas or monitor for after-discharges during during electrical stimulation mapping around the tumor (Yang et al., 2014). Intraoperative ECoG recordings can identify epileptic peritumoral networks based on interictal biomarkers of epilepsy such as epileptic spikes and high frequency oscillations (Berger et al., 1993; Mikuni et al., 2006; Yao et al., 2017; Feyissa et al., 2018). Interestingly, in patients with brain tumor-related epilepsy, (Feyissa et al., 2018) high frequency oscillations (i.e., > 80 Hz epileptic associated oscillations) are associated with neuroinflammation (Sun et al., 2022) and a variety of tumor molecular pathways (Hills et al., 2022). Additionally, increased high frequency oscillations have been associated with IDH-wildtype status, and overall survival is associated with greater surface EEG excitability (Feyissa et al., 2018; Tobochnik et al., 2022). While not fully understood, tumor-related epilepsy (and thus high frequency oscillations) is thought to be related to IDH-related increases in glutamate and/or D-2-hydroxyglutarate and reduced GABAergic inhibition (Armstrong et al., 2016; Chen et al., 2017). Thus, ECoG assessments of high frequency oscillations may be a critical marker of length of survival in patients with brain tumors.

As with scalp EEG, the association between abnormal low frequency neuromagnetic activity and brain tumors is well recognized (Baayen et al., 2003; de Jongh et al., 2003). Tumor-related focal delta activity in the peritumoral cortex has also been noted to be greater in the case of intra-axial tumors that involved subcortical fibers than for extra-axial tumors, and patients with an increased delta activity exhibited poor recovery of function, at least within the early postoperative period (Oshino et al., 2007). As observed in scalp EEG studies, a decrease in power in the high-alpha and beta frequencies has also been reported in peritumoral regions (Shekoohishooli et al., 2021).

There is vast literature on the use of MEG for localizing the sources of epileptic spike activity in patients with drug-resistant epilepsy, but few have focused specifically on peritumoral epilepsy (Patt et al., 2000; Birbilis et al., 2014). In a study that correlated tumor type with the location of source dipoles, sources were noted to be significantly closer to the tumor border in glial tumors compared to mixed glial-neuronal neoplasms or metastatic tumors, although tumor cells did not appear to be directly involved in generating spikes (Patt et al., 2000). Similarly, although MEG is an approved tool for the presurgical mapping of motor and primary sensory cortex and for determination of hemispheric language dominance in patients with epilepsy, brain tumors, and cerebral vascular lesions, the existing literature specific to functional mapping in brain tumors is sparse. One of these studies found that at a 10-mm distance-threshold, MEG showed a greater specificity than fMRI for both motor and language mapping but a lower sensitivity for motor mapping (Ellis et al., 2020). Several studies of language mapping have claimed good accuracy with MEG in localizing Broca’s and Wernicke’s areas in brain tumor patients using a variety of source modeling approaches and tasks with significant overlap with fMRI activations in some of these studies (Grummich et al., 2006; Pang et al., 2011; Huang et al., 2016).

Over the last couple of decades there have been many studies of functional connectivity based on resting state MEG source dynamics in patients with brain tumors. The studies examine both the pathophysiology of cognitive impairment in this population and clinical applications such as identifying high functional connectivity (HFC) nodes around tumors. An early study of broad band (0.5–60 Hz) and gamma band (30–60 Hz) MEG functional connectivity performed at the sensor level, suggested loss of functional connectivity in multiple regions with left hemispheric tumors registering a greater impact (Bartolomei et al., 2006). Subsequent studies that have looked at multiple frequency bands consistently show evidence of increased theta-band connectivity both at short and long distances from tumor locations (Bosma et al., 2008, 2009; Derks et al., 2014, 2021). These studies also show significant correlations between increased theta-band connectivity and degree of cognitive dysfunction associated with low-grade gliomas. This suggests that cognitive dysfunction in these patients may be related to widespread changes in the strength and spatial organization of functional brain networks. Increased theta-band connectivity also appears to correlate with increased seizure frequency in patients with epilepsy related to low-grade gliomas (Douw et al., 2010; van Dellen et al., 2012). Furthermore, while gliomas tend to occur more often in brain regions that have high network-clustering, they have greater detrimental effect on whole-brain networks and cognition when located in regions with low-clustering (Derks et al., 2021).

From the perspective of immediate clinical applications, several studies have now demonstrated that preservation of peritumoral areas with HFC in the alpha band (8–12 Hz) based on resting state MEG can predict post-operative neurological outcomes (Guggisberg et al., 2008; Martino et al., 2011; Tarapore et al., 2012; Lee et al., 2020). Such HFC nodes also have a high probability of yielding positive responses during intraoperative electrical stimulation mapping of sensory, motor, and language functions, and, quite remarkably, a 100% negative predictive value for the absence of eloquent cortex during stimulation mapping (Martino et al., 2011).

Magnetoencephalography (MEG) has also been used by a few groups to study the reorganization of cortical areas that support motor and language function around brain tumors before and after surgical resections, especially in patients with recurrent tumors. Detectable shifts in the location of motor representations that correlated with a frontoparietal location of the tumor, presence of motor deficits, and interval between the scans have been reported (Bulubas et al., 2020). Likewise, laterality indices of hemispheric language dominance were noted to change in a majority of patients with recurrent gliomas, with the scale and direction of change depending on initial degree of lateralization and tumor location (Traut et al., 2019). Enhancement of post-operative alpha-band connectivity has been observed in peritumoral regions in all patients regardless of the low grade glioma location and has been suggested as a marker for post-operative plasticity (Lizarazu et al., 2020). After glioma resections, an increase in alpha-band connectivity compared to the pre-operative baseline between the default-mode network areas was found to correlate with improvement in post-operative verbal memory, while an increase in alpha-band connectivity across the frontoparietal areas correlated with improvement in attention (Van Dellen et al., 2013).

4.2. MRI-based techniques

Tractography is now used routinely in many centers to identify the relationship of brain tumors to critical white matter tracts (Figure 6). More recently, researchers have used tractography to examine structural “disconnectomes,” or the probability that a lesion will interrupt the structural connections between paired regions. Similarly, recent work has demonstrated that, in patients with glioblastoma, both structural connectivity is decreased relative to controls and reductions of long range connections is associated with poorer overall survival (Wei et al., 2022).

Similarly, resting-state fMRI has been frequently used to define the topography of networks associated with language and motor behavior (Hacker et al., 2019; Kumar et al., 2020; Mitchell et al., 2020; Daniel et al., 2021; Yang P. et al., 2021). Commonly, these studies employ seed-based correlation (Figure 7) analyses to define the boundaries of these networks for the purpose of surgical planning (Hacker et al., 2019). The major benefit of resting-state fMRI for defining network boundaries is its independence from task performance in the scanner. This is important in participants with poor ability to perform the desired task or for networks poorly defined by a task paradigm (such as the salience network). While clinically useful in determining the hemispheric laterality of language dominance and general localization of motor, speech, and language areas, fMRI does not provide as assessment of the critical or hierarchical nature of BOLD-positive sites. In other words, areas that light up with BOLD may not result in a neurological deficit if injured, meaning that, while they are active during a task, they are not necessary for its performance (i.e., redundant). Additionally, pure motor information can be difficult to isolate from sensation during a motor task, as sensory information is almost constantly flowing. fMRI is also limited by its large temporal scale and the indirect nature of the recording; that is, fMRI measures the hemodynamic response to neural activity rather than neural activity directly. The latter can confound results in brain tumor cases when neurovascular decoupling can occur (Agarwal et al., 2021).

Functional magnetic resonance imaging (fMRI) and diffusion tractography mapping prior to surgical resection have been reviewed previously (Lee et al., 2016; Hu and Hoch, 2021; Ille and Krieg, 2021; Sparacia et al., 2021). The review by Sparacia et al. (2021) includes a clinical workflow for their institution based on readily available software. Newer research has extended these fMRI and tractography mapping techniques by using tractography to identify an individual patient specific parcellation for the brain (Doyen et al., 2022; Dadario et al., 2023). Similarly, this same group has proposed that highly connected areas (i.e., “hubs”) may identify non-speech/motor areas that should be left during surgical resection (Tanglay et al., 2023). This concept will be explored further in the “Future directions” section below.

4.3. Applications to pediatric brain surgery

Pediatric neurosurgery patients pose unique challenges for many reasons, some of which include their inability to participate in complex tasks and an actively developing brain. Unfortunately, the pediatric-specific literature on brain mapping techniques is limited and largely focused on epilepsy (Coppola et al., 2016; Chou et al., 2018). Presurgical planning with fMRI can pose several challenges, including inability to stay still in the scanner and lack of cooperation with task performance (Maheshwari, 2019). In younger children and non-cooperative patients, passive sensorimotor mapping under sedation may be an option (Maheshwari, 2019). High concurrence of the cortical activation between active and passive motor tasks has been demonstrated in adults (Kocak et al., 2009). Comparison of passive motor mapping under sedation with active motor mapping and direct cortical stimulation has been reported in children with focal brain lesions, masses, or cortical dysplasia (Ogg et al., 2009).

Rosazza et al. (2014) assessed the degree of correlation between sensorimotor mapping with task-based and rs-fMRI techniques. While larger areas of activation were seen with rs-fMRI, they concluded there was enough agreement between the techniques to warrant its use in patients who were unable to perform the tasks. Similarly, several studies comparing language localization between task-based fMRI and rs-fMRI in adult patients found good concordance between task-related language activation and resting-state language networks (Branco et al., 2016; Sair et al., 2016; Maheshwari, 2019). Given these results, rs-fMRI is particularly appealing for the evaluation of whole brain networks in children.

The pediatric brain tumor population is also unique in that nearly two-thirds of pediatric brain tumor patients are diagnosed with an infratentorial brain tumor. The most frequent infratentorial or posterior fossa tumors (PFTs) are medulloblastomas (40%), astrocytomas (30%), and ependymomas (10%) (Baudou et al., 2021; Sleurs et al., 2021). Their treatment depends largely on tumor histology and the patient’s age (Baudou et al., 2021). Many of these patients suffer long-term effects secondary to direct damage from the tumor, damage from surgery, and/or toxicity from adjuvant therapy such as chemotherapy and radiotherapy. Young children with posterior fossa brain tumors are particularly at risk for the development of cerebellar mutism syndrome (CMS), which may present with key symptoms including transient muteness/reduced speech, ataxia, hypotonia, dysphagia, cranial nerve impairment, and emotional lability. Its onset can be delayed and recovery is prolonged and not always complete. Previously thought to be responsible only for balance and coordination, a newer body of evidence has revealed the cerebellum also plays a central role in working memory and procedural memory, both of which are involved in motor and cognitive learning (Baudou et al., 2021). The constellation of symptoms in CMS, including its similarities to other syndromes arising from topographically distinct regions such as the supplementary motor area (SMA) involved in SMA syndrome, have raised theories on the existence of interrelated neural loops supporting a range of higher functions (Grønbæk et al., 2020).

Recent studies have used lesion-symptom mapping to create a cerebellar white matter atlas that shows cerebello-thalamo-cerebral outflow tracts in the superior cerebellar peduncle (SCP) next to the dentate nucleus (DN) and inferior vermis are associated with impaired motor, cognitive, executive, and behavioral function (Grosse et al., 2021).

Miller et al. (2010) demonstrated decreased cortical perfusion in association with decreased cerebellar perfusion in patients with CMS. Similarly, decreased metabolism in the right cerebellar and left frontal lobes has been detected on fluorodeoxyglucose positron emission tomography/computer tomography in a patient with CMS (Gedik et al., 2017). Furthermore, functional connectivity studies have demonstrated connectivity of the dentate nuclei with SMA and Broca’s areas, further strengthening the hypothesis of dysfunctional overlapping pathways or networks in PFT survivors (Grønbæk et al., 2020). Language processing is complex and may differ from classical language areas in the setting of anatomic anomalies (e.g., cortical dysplasia, tumor invasion, etc.), abnormal electrical circuitry (e.g., epilepsy), or as a response to brain injury, particularly in the neuroplastic pediatric brain (Vadivelu et al., 2013; Chou et al., 2018).

Sleurs et al. (2021) reported widespread microstructural changes in supratentorial brain areas in addition to macrostructural alterations in infratentorial brain areas in survivors of PFTs. When comparing all brain regions (nodes) in these patients compared to healthy controls, significant differences in nodal strength was seen in “hubs” or regions that were most densely connected. Furthermore, these highly connected nodes were most affected in CNS tumor survivors and highly correlated with intelligence scores (Sleurs et al., 2021).

While gross total resection of posterior fossa tumors contributes to relapse-free survival (Grosse et al., 2021), preventing injury to the SCP and deep cerebellar nuclei is critical for preserving optimal motor and cognitive function in pediatric cerebellar tumor survivors. Multimodal perioperative imaging aimed at evaluating structural and functional alterations with high spatial and temporal resolution (e.g., tractography, MEG, task-based fMRI and/or rs-fMRI, with regional network and cross-network connectome analysis) may help identify imaging biomarkers in high-risk patients. These imaging biomarkers may then allow physicians to individualize treatments, avoid complications, and ultimately improve the recovery and long-term outcomes of these patients.

4.4. Brain tumor-related cognitive dysfunction and depression

A decline in cognitive function and the emergence of depression can have similar and compounding effects on brain tumor patients. Either symptom can diminish independence, hinder social relationships, and/or negatively impact overall perceived quality of life (Noll et al., 2017; Lange et al., 2019). Coping and adjustment related syndromes may also occur after diagnosis, during treatments, or during survivorship (Noll et al., 2019). Some estimate depression rates in glioma patients to be around 15–20% (Rooney et al., 2014), and up to about 90% show select features of depression (Noll et al., 2019). Not only is an individual’s depressive status important to know for quality of life purposes, but it may also help to inform prognosis, as depression has been associated with poorer overall survival in gliomas (Noll et al., 2019).

In addition to the psychological manifestation of these symptoms, brain tumors themselves may directly impact emotional regulation, personality, and behavior when dysregulating their associated networks. For example, lesions in the dorsolateral prefrontal cortex typically result in a primary executive dysfunction syndrome (e.g., deficits in problem solving and reasoning), while lesions in orbital frontal regions can result in increased disinhibition, impulsivity, and inappropriateness. Similarly, lesions in the medial frontal region can result in apathy and abulia (Filley and Kleinschmidt-DeMasters, 1995; Cummings and Mega, 2003; Boele et al., 2015). The third syndrome, medial frontal syndrome, is most likely to be mis-interpreted as “depression” given the overlap in some of the key behavioral changes, such as social withdraw, emotional blunting, and disengagement.

Lesions outside of the frontal lobes may also result in changes in mood and emotional processing, such as those that disrupt parietal integration regions and temporal paralimbic structures (Boele et al., 2015). A voxel-based lesion-symptom mapping study examined mood and emotional perception as well as higher level emotional reasoning and abstraction in patients with frontal, parietal, and temporal tumors (Campanella et al., 2014). They found basic emotional interpretation was affected by temporo-insular lesions, emotion recognition was maximally impacted by anterior temporal and amygdala lesions, and higher-level emotional reasoning was maximally affected by prefrontal lesions (Campanella et al., 2014).

4.5. Future directions

An intriguing emerging notion is that “centrality/hubness” (i.e., the graph theory idea that an area with many strong connections is more critical than an area with fewer connections) may provide a marker of resectability beyond the boundaries of historically “eloquent” tissue (Tanglay et al., 2023). For example, an area in the superior frontal lobe (Ahsan et al., 2020) that has a number of important connections may not be suitable for resection in a given person. Conversely, if another person has a portion of “eloquent” cortex with low centrality, that could hypothetically be resectable in that individual (Ahsan et al., 2020; Yeung et al., 2021). Importantly, while tractography has been used to assign “connectivity” and subsequently “centrality” to certain nodes, many of the techniques in this review may expand upon these assessments. For example in patients with epilepsy, eigenvector centrality calculated from spontaneous ECoG was found to correlate directly with spiking at an electrode (Zweiphenning et al., 2016). This has been useful in identifying seizure onset zones (Burns et al., 2014). Likewise, centrality of a given cortical node may prove beneficial in isolating resectable boundaries. It may also be possible to apply this concept to map nodes critical for higher cognitive functions.

One example of a well-studied temporary motor phenomenon is supplementary motor area syndrome (SMA syndrome), or when resection of the SMA in the posterior superior frontal gyrus results in acute impairments in motor control and speech (Palmisciano et al., 2022). Most often, symptoms are temporary with mean duration ∼45 days (∼80% of cases), but it can be clinically difficult to differentiate from true hemiparesis and mutism (Palmisciano et al., 2022). Persistent SMA syndrome occurs in ∼20% of reported cases (Palmisciano et al., 2022). As intraoperatively monitoring motor evoked potentials (MEPs) incompletely predicts persistent motor impairments (Giampiccolo et al., 2021a), a network approach may be able to predict patients likely to have persistent motor deficits for reasons other than primary motor tract integrity. Because it is known that the contralateral SMA is critical to the recovery from the syndrome (Chivukula et al., 2018), higher interhemispheric connectivity between the respective SMAs may be protective from persistent deficits (Young et al., 2021). It is possible that the pre-resection strength of connectivity from the SMA to the motor cortex will be greater in those who have persistent symptoms (Figure 8). Alternatively, it is possible that diffusion tractography will identify a greater number of descending streamlines directly from the SMA to the corticospinal tract in those with persistent motor impairments. Ultimately, such investigations have the potential to guide pre- and post-operative rehabilitation strategies to facilitate more rapid resolution of SMA syndrome, and/or to better predict which individuals will have persistent deficits.

Likewise, connectomic techniques may soon enable the mapping, monitoring, and protection of higher-order cognitive functions. Notably, there is a paucity of studies systematically mapping connections between nodes with high fidelity functional and health related outcome measures for many cognitive processes.

Additionally, there is untapped potential in combined mapping techniques. For example, concurrent TMS-fMRI (Bergmann et al., 2021) or TMS-EEG (Tremblay et al., 2019) have, to our knowledge, never been examined in patients with brain tumors during pre-resection mapping. Concurrent TMS-fMRI and TMS-EEG involve interleaving TMS stimulations into the fMRI and EEG recording paradigms. Techniques such as TMS-fMRI could solve some of the issues with pre-surgical mapping with task-based fMRI (i.e., motivation, functional capacity, understanding directions, etc.). That is, measuring BOLD responses elicited by TMS stimulation would be task-independent, similar to resting state fMRI mapping, but could also alleviate the computational difficulties of resting-state fMRI (i.e., seed-based vs. independent component analysis-based techniques). However, this would require specialist trained staff competent on both techniques (Bergmann et al., 2021).