In this study, we evaluated a cohort of patients who presented to our center with insular and peri-Sylvian gliomas and utilized a machine learning based analytic tool to determine the frequency of involvement of large-scale brain networks and/or associated en passage white matter fiber bundles in these tumors. We found that in these patients, greater than 98% of these patients had tumors which had sub-centimeter distance to the cortical or subcortical components of a large-scale brain network or major white matter pathway (or it invaded these structures). Most of these patients had multiple structures at risk from their tumor. All of this highlights the challenges posed by these complex tumors, and suggests that these cases really seldom are no risk, even when on in right hemisphere.
There are substantial advantages to reconsidering the deep white matter anatomy in terms of large-scale brain networks. Notably, many large white matter bundles are in fact numerous bundles connecting different areas, not all of which are necessary for human cognitive function. This is especially true for the superior longitudinal fasciculus (SLF)-arcuate fasciculus (AF) complex, which is a highly complex network of connections, of varying importance. Given that some parts of these bundles and their branches are more important than others, and we need to cut something in order to remove a glioma, network-based approaches provide the potential benefit for better defining which white fibers are more important than others. Ultimately, all of brain surgery involves accurately calculating risks vs rewards of our actions.
The Non-Traditional, Large Scale Brain Networks
1. Default Mode Network
The default mode network (DMN) serves a primary role in passive states of mind (29), like internal thought or contemplation, however, it is also active during some goal-oriented tasks. The DMN network is typically described as consisting of the anterior and posterior cingulate cortices, and the lateral parietal lobe bilaterally. DMN also plays an integral role in coordinating with other networks for passive sensory processing (30). These connections often include: visual system (31), language subnetwork (32), and limbic system (33). In Figure 1, we demonstrate a tumor invading the DMN.
2. Salience Network
The Salience Network (SN) is integral for sensorimotor processing, general cognition, and coordinating between emotion, pain, and physical action (34). As the mind’s moderator, the SN constantly monitors the external environment and decides how other brain networks react to new information and stimuli, it also plays an essential role in switching between the internal and external processing (35) of the brain’s two main control networks (36): the default mode network (DMN) and central executive network (CEN). The main functional areas of the salience network are located in the anterior cingulate, the anterior insula (37, 38) and the presupplementary motor areas (39). The SN also includes nodes in the amygdala, hypothalamus, ventral striatum, thalamus, and specific brainstem nuclei (37), anterior cingulate cortex (ACC), medial temporal network, parahippocampal gyrus, olfactory lobe, and the ventral tegmental area (VTA). In Figure 2, we demonstrate a tumor invading the SN.
3. Central Executive Network
The central executive network (CEN) exists as a superordinate control network (40). It uses input from other networks for task selection and executive function. By integrating with the other brain networks, the CEN processes a varied set of information, such as flexibility, working memory, initiation, and inhibition, all of which had previously been thought to be separate processes. Since its initial discovery in the anterior frontal lobe (41),the central executive network has been found to be functionally connected to regions in the anterior cingulate cortex, the inferior parietal lobe (42), and the posterior most portions of the middle and inferior temporal gyri (43, 44). In Figure 3, we demonstrate a tumor invading the CEN.
4. Dorsal Attention Network
The dorsal attention network (DAN) is an important mediator of goal-directed attentional processing (24), and it has many ways of contributing to intellectual capabilities. As a bilateral network, DAN demonstrates strong connectivity between areas in the lateral occipital lobe, the pre-central sulcus, the dorsal-most portion of the superior frontal sulcus considered to be the frontal eye fields (FEF), the ventral premotor cortex, superior parietal lobule, intraparietal sulcus, and motion-sensitive middle temporal area (45). In Figure 4, we demonstrate a tumor invading the DAN.
5. Ventral Attention Network
The ventral attention network (VAN) is one of two-network model of cortical attention (46, 47), which is involved in reorienting attention when a new, unexpected stimulus, like shock, frightening events, or “oddball” occurrences, is detected within the environment (47, 48). Multiple cortical areas, such as the middle and inferior frontal gyri, anterior insula, inferior parietal lobule, and temporo-parietal junction have been linked in this processing (25). In Figure 5, we demonstrate a tumor invading the VAN.
Limitations and Future Directions
The specific goal of this study was to estimate how often a surgeon performing surgery for infiltrating gliomas around the Sylvian fissure could reasonably expect that a network not normally addressed at surgery would be inside of the tumor, and this ultimately suggested that this was the expected nature of almost all of these tumors. This study is a survey of the insula-Sylvian region gliomas encountered in at a single center over a 3-year period. Ultimately, this study was not intended to be an exhaustive epidemiologic study of all patterns of glioma spread in cases around the Sylvian fissure, and it is possible that a different center may find different frequencies of network involvement with a similar cohort. We studied a consecutive series and generally aim to perform resectable surgery of some type for every case reasonable, even if not completely resected. Thus, there is not an obvious selection bias which skewed the results which is readily apparent to us other than the inherent referral patterns which route tumors around the Sylvian fissure frequently to university-based neuro-oncology programs. While it is possible that a cohort of gliomas seen over a large community cohort, with different complexity and disease severity from ours, might have different specific rates of involvement of the brain networks than our study, the fact remains that the rate of these networks being involved in these tumors is likely high. Therefore, they will continue to pose a risk of cognitive and emotional morbidity in the majority of patients if we continue to not address the situations where they cannot or should not be transgressed in the name of increasing the extent of resection a small amount.