Parcellation‐based anatomic modeling of the default mode network

Abstract Background The default mode network (DMN) is an important mediator of passive states of mind. Multiple cortical areas, such as the anterior cingulate cortex, posterior cingulate cortex, and lateral parietal lobe, have been linked in this processing, though knowledge of network connectivity had limited tractographic specificity. Methods Using resting‐state fMRI studies related to the DMN, we generated an activation likelihood estimation (ALE). We built a tractographical model of this network based on the cortical parcellation scheme previously published under the Human Connectome Project. DSI‐based fiber tractography was performed to determine the structural connections between cortical parcellations comprising the network. Results Seventeen cortical regions were found to be part of the DMN: 10r, 31a, 31pd, 31pv, a24, d23ab, IP1, p32, POS1, POS2, RSC, PFm, PGi, PGs, s32, TPOJ3, and v23ab. These regions showed consistent interconnections between adjacent parcellations, and the cingulum was found to connect the anterior and posterior cingulate clusters within the network. Conclusions We present a preliminary anatomic model of the default mode network. Further studies may refine this model with the ultimate goal of clinical application.

and anatomy of higher-order cerebral networks are therefore likely to manifest in advances in brain tumor surgery.
Several studies have characterized the anatomy of the default mode network (DMN) since its discovery in 2001 (Andrews-Hanna et al., 2010;Buckner et al., 2008;Horn et al., 2014;Raichle et al., 2001). It serves a primary role in passive states of mind, though it is also active during some goal-oriented tasks (Bressler & Menon, 2010;Chand et al., 2017;Fox, Snyder, et al., 2005;Greicius et al., 2003). The network is typically described as consisting of the anterior and posterior cingulate cortices, and the lateral parietal lobe bilaterally (Alves et al., 2019;Andrews-Hanna et al., 2010;Buckner et al., 2008). While important, existing descriptions of the DMN offer limited anatomical specificity, making it difficult to compare findings among different papers. This study instead relies on newly published parcellated brain maps to study the network anatomy of the DMN using a standard cortical atlas and nomenclature .
In this study, a new cortical model of the DMN was constructed based on the parcellation scheme previously published under the Human Connectome Project (HCP) . The HCP atlas is among the most detailed in vivo parcellation scheme constructed by combining automated machine learning approaches with extant neuroanatomical literature. It allows consistent and detailed delineation of cortical areas which we employed for its potential for reproducibility across studies and in clinical contexts. After identifying the cortical regions of interest involved in the network, we performed DSI-based fiber tractography to demonstrate the structural connections between parcellations within the network. Our goal is to move toward a more precise anatomic model of the DMN for use in future studies.

| Literature search
We initially searched for relevant fMRI studies related to the DMN in BrainMap Sleuth 2.4 Fox & Lancaster, 2002;Laird et al., 2005). No research articles were identified using this software. Other literature software's were not queried as PubMed was subsequently queried to cover the literature gap on October 1, 2020, for fMRI studies relevant to the default mode network. We used the following search algorithm: (default mode OR default mode network OR DMN) AND "resting-state fMRI" AND controls." Studies relevant network were reviewed and included in our analysis if they fulfilled the following search criteria: (a) peer-reviewed publication, (b) resting-state fMRI study examining the DMN, (c) based on whole-brain, voxelwise imaging, (d) including standardized coordinate-based results in the Talairach or Montreal Neuroimaging Institute (MNI) coordinate space, and (e) including at least one healthy human control cohort. Only coordinates from healthy subjects were utilized in our analysis. Twenty-eight papers met criteria for inclusion in this study (Anderson et al., 2011;Che et al., 2014;Chen et al., 2017;Chiong et al., 2013;Clemens et al., 2017;Crittenden et al., 2015;De Luca et al., 2006;Doll et al., 2015;Fransson, 2006;Greicius et al., 2003;Horn et al., 2014;Kennedy & Courchesne, 2008;Konishi et al., 2015;Laird et al., 2009;Lin et al., 2017;Luo et al., 2016;Maresh et al., 2014;Mason et al., 2007;Piccoli et al., 2015;Pletzer et al., 2015;Poerio et al., 2017;Spreng & Schacter, 2012;Stawarczyk et al., 2011;Taruffi et al., 2017;Utevsky et al., 2014;Vatansever et al., 2015;J. Xu et al., 2014;Yang et al., 2012). The details of these studies are summarized in Table 1.

| Creation of 3D Regions of Interest
In the original HCP study, parcellation data were analyzed using the CIFTI file format. CIFTI files use a surface-based coordinate system termed greyordinates, which localizes regions of interest (ROIs) on inflated brains . This is in contrast to traditional file formats, such as NIFTI, which denote regions based on volumetric dimensions (Larobina & Murino, 2014).
As a result, it was difficult to perform deterministic tractography using ROIs in CIFTI file format. To convert the parcellations files to volumetric coordinates, the greyordinate label parcellation fields were standardized to the three-dimensional volumetric working spaces of DSI Studio (Carnegie Mellon, http://dsi-studio. labso lver.org) using the structural imaging data provided by the HCP. This operation was performed using Workbench Command within Connectome Workbench . This allowed us to convert all 180 parcellations from surface-based coordinates to volumetric coordinates and perform deterministic fiber tractography.

| Activation likelihood generation and identification of relevant cortical regions
We used BrainMap Ginger ALE 2.3.6 to extract the relevant fMRI data for creation of an activation likelihood estimation (ALE; Eickhoff et al., 2009Eickhoff et al., , 2012Turkeltaub et al., 2012 Visuospatial planning was investigated by presenting participants with 2 configurations on a single screen: the "goal" position and the "initial" position.

| Network tractography
Publicly available imaging data from the Human Connectome Project were obtained for this study from the HCP database (http://human conne ctome.org, release Q3). Diffusion imaging with corresponding T1-weighted images from 25 healthy, unrelated subjects were analyzed during fiber tracking analysis (Subjects IDs: 100307, 103414, 105115, 110411, 111312, 113619, 115320, 117112, 118730, 118932, 100408, 115320, 116524, 118730, 123925, 148335, 148840, 151526, 160123, 178950, 188347, 192540, 212318, 366446, 756055). We used 25 brains as it is comparable to the number of subjects used in studies of a similar aim. We have previously tested the variability of tractography results above utilizing 25 subjects, and however, it was too small to justify using additional subjects as it is unlikely to alter the findings of the study. Often, beyond 10 subjects, the results do not change significantly. The demographics of the patients used in this study are detailed in Table 2. A multi-shell diffusion scheme was used, and the b-values were 990, 1985, and 1980 s/mm 2 . Each b-value was sampled in 90 directions. The in-plane resolution was 1.25 mm. The diffusion data were reconstructed using generalized q-sampling imaging with a diffusion sampling length ratio of 1.25 (Yeh et al., 2010).
All brains were registered to the Montreal Neurologic Institute (MNI) coordinate space (Evans et al., 1992), wherein imaging is warped to fit a standardized brain model comparison between subjects (Evans et al., 1992). Tractography was performed in DSI Studio (Carnegie Mellon, http://dsi-studio.labso lver.org) using a region of interest approach to initiate fiber tracking from a user-defined seed region (Martino et al., 2013). A two-ROI-approach was used to isolate tracts (Kamali et al., 2014).
Voxels within each ROI were automatically traced with a maximum angular threshold of 45 degrees. When a voxel was approached with no tract direction or a direction change of greater than 45 degrees, the tract was halted. Tractography was terminated after reaching a maximum length of 800 mm. In some instances, exclusion ROIs were placed to exclude obvious spurious tracts that were not involved in the white matter pathway of interest.

| Measuring connection strength
To quantify the strength of the connections identified within the DMN across all subjects, the tracking parameters used within DSI Studio were modified such that the program would count the total number of tracts between any two ROIs based on a random seed count of 2.5 million. Working sequentially through ROI pairs in the network, the number of tracts between regions was recorded for each subject after fiber tractography was terminated under these new conditions. The connection strength between ROI pairs within the DMN was calculated by averaging the number of tracts between each ROI pair across all subjects.

| Structural connections within the default mode network
Deterministic tractography was utilized to show the basic structural connectivity of the DMN. These results are shown in Figure 3.
Individual connections within the network are presented in Table 3 which tabulates the strengths of individual connections and lists the type-specific white matter connections identified between regions.
The cortical areas identified as part of the DMN can be grouped into three distinct clusters: an anterior cingulate cluster (10r, a24, p32, s32), a posterior cingulate cluster (31a, 31pd, 31pv, d23ab, POS1, POS2, RSC, v23ab), and a lateral parietal cluster (IP1, PFm, PGi, PGs, TPOJ3). U-shaped fibers form a majority of the connections between ROI pairs of the network. These fibers generally have the same morphology, arising within one part of the cortex before curving 180 degrees to terminate in a part of the brain immediately adjacent to its origin. These U-shaped fibers represent the local connections between anterior cingulate, posterior cingulate, and lateral parietal areas in close proximity.
The cingulum was also identified during fiber tracking analysis.
This white matter bundle was found to connect the anterior and posterior cingulate clusters within the DMN. These fibers arise from the anterior cingulate cortex, and curve posteriorly to run within the deep white matter adjacent to the cingulate gyrus. The fibers course along the length of the corpus callosum, until they terminate in the posterior cingulate cortex and parieto-occipital sulcus ( Figure 3).
All four parcellations of the anterior cingulate cluster (10r, a24, p32, and s32) contribute to the cingulum, though with variable frequency. Areas a24 and p32 demonstrated consistent connections across all 25 subjects to all parcellations of the posterior cingulate cluster (31a, 31pd, 31pv, d23ab, POS1, POS2, RSC, and v23ab). In contrast, the connections from areas 10r and s32 were occurred infrequently, and the parcellations were found to connect to fewer regions of the posterior cingulate cortex (Table 3).
No long-association fiber bundle was found to connect the lateral parietal regions to either the anterior cingulate or posterior cingulate cortices. This was expected, as no such connection has been described previously. However, IP1, PFm, PGi, PGs, and TPOJ3 all There was no relationship between demographic data and network anatomy within our cohort. Our cohort was not diverse enough to observe significant differences as this was not a primary aim of this study and we wanted to produce a model from healthy controls to avoid confounding factors. It may, however, be interesting to study changes in the DMN under different demographic characteristics in the future.

| D ISCUSS I ON
In this study, we utilized meta-analytic software and deterministic gions of the DMN are pivotal to understanding its function which may then offer insights for clinicians into the mechanism and potential therapies for these disorders. In addition, a precise anatomic and connectomic description of the network will allow surgeons to make better judgments during brain surgery. The anatomic constituents of this network are discussed below.

| The anterior cingulate cluster
Cortical areas 10r, a24, p32, and s32 overlap with the ALE in the region of the anterior cingulate cortex, which has been identified as a com- Area 10r is one of several newly described divisions of the original Brodmann Area 10 , which expanded the entire frontal polar cortex from the medial superior frontal gyrus to the dorsolateral prefrontal cortex (Burgess & Wu, 2013;Peng et al., 2018).
Little is known about this region; however, it is located in the anterior inferior portion of the medial superior frontal gyrus. Just posterior to area 10r is area s32 which lies in the subcallosal cortex. This region is interconnected to other areas of the limbic system and is known to play a role in emotional response regulation and reward expectation (Beckmann et al., 2009;Palomero-Gallagher et al., 2008).
Superior to areas 10r and s32 are areas p32 and a24, respectively. Area p32 is located in the antero-medial superior frontal gyrus, bordering the inferior bend of the callosal sulcus. This region plays a role in the integration of emotional and cognitive information during social interaction tasks to assist in error monitoring (Beckmann et al., 2009;Palomero-Gallagher et al., 2008). Area a24 is located in the anterior cingulate gyrus proper, lying anterior to the genu of the corpus callosum. This region has been implicated as part of the "affect division" of the anterior cingulate cortex and has been linked to the analysis of internal and external states of mind to assist in emotional expression and motivation (Devinsky et al., 1995;Drevets et al., 2008).
In contrast to areas 31a and d23ab, areas 31pd, 31pv, and   This region is primarily responsible for transitioning between allocentric or view-independent spatial perspectives and egocentric or view-dependent spatial perspectives (Aggleton et al., 2014;Glasser et al., 2016;Leech & Sharp, 2014;Vann et al., 2009). The RSC is implicated in spatial navigation, episodic memory, future planning, and imagination. In addition, the RSC has been suspected of being involved in the retrieval of recent autobiographical information from memory (Aggleton et al., 2014;Glasser et al., 2016;Leech & Sharp, 2014;Vann et al., 2009).
Regions POS1 and POS2 occupy the inferior and superior halves of the anterior bank of the parieto-occipital sulcus, respectively. Task fMRI studies demonstrate that areas POS1 and POS2 are activated during the working memory processes of place images , and it has been suggested that both regions have a strong, coupled functional correlation with area RSC related to scene comprehension .
All eight regions are interconnected by U-shaped fibers to one other and demonstrate fiber projections consistent with the

| The lateral parietal cluster
Cortical areas PFm, PGi, PGs, IP1, and TPOJ3 overlap with the ALE in the region of the lateral parietal lobe, which, as for the cingulate cortices, has been identified consistently as a component of the DMN across multiple studies (Alves et al., 2019;Buckner et al., 2008;Lei et al., 2014;Sestieri et al., 2011;Xu et al., 2016;Zanchi et al., 2017). -Area PFm has been shown to be active during nonspatial attention tasks, during decision making tasks when individuals change choices, during rule changes during visually guided attention tasks, and during attentional reorientation (Caspers et al., 2006). The inferior parietal lobule is also involved in the syntactical components of language processing (Ben Shalom & Poeppel, 2008).
-Area PGs has been shown to be active when individuals change their visuospatial attention from one focus to another (Mars et al., 2011). Specifically, area PGs is involved in the response to biological motion (Mars et al., 2011). The region is also relevant in number processing (Caspers et al., 2011).
-Area PGi has a functional profile similar to that of area PGs. For example, area PGi has been shown to be active when individuals change their visuospatial attention from one focus to another (Mars et al., 2011). Within the original parcellations study, the Human Connectome Project authors discuss both areas PGi and PGs as major nodes in the DMN .
The other two areas identified as part of the DMN were areas IP1 found on the inferior bank of the intraparietal sulcus, and area F I G U R E 4 Simplified schematic of the white matter connections identified between individual parcellations of the default mode network during fiber tracking analysis. Connections are labeled with the average strength measured across all 25 subjects TPOJ3 found on the posterior inferior portion of the inferior parietal lobule. Area IP1 shows significant activation during mental arithmetic activities, and, as part the intraparietal sulcus, supports more complex parts of numeric and mathematical information processing (Uddin et al., 2010;Wu et al., 2009

| The strength of connections within the default mode network
The strength of the connections identified between parcellations of the DMN is reported in Table 3. Two different values for strength are recorded based on the average number of tracts across all subjects versus the average number of tracts across subjects in which the connection was actually identified. It is certainly the case that the structural connectivity of the DMN varies to some degree between individuals, and by presenting both sets of average connectional strengths, one can see how connections can vary in the network. For example, the cingulum projection from area 10r to area 31pv has an average strength of 5.2 across all 25 subjects (meaning one would expect to find 5.2 streamlines using the fiber tracking algorithm discussed in the methods) versus an average strength of 32.2 in the four individuals in which the connection was identified. By reporting both numbers, we can see that, while the connection between 10r and 31pv occurs infrequently in the network, in individuals who have such a connection, it is relatively strong.
It should also be noted that we did not set a threshold for the strength that might limit the connections shown for the DMN. For example, assessing the connection between a24 and d23ab via the cingulum, one sees that the average strength across all 25 subjects used in this study was 4.8 versus 7.9 in the fifteen subjects for whom such a connection was actually identified. If we had set a threshold of an average strength of 10.0 or set a threshold related to the frequency by which we saw the connection, that is, in at least 20/25 subjects, then we would not report this connection at all. In our mind, this is incorrect. Instead, it more appropriate to say that the connection between a24 and d23ab, while relatively weak compared to other connections in the network, occurs relatively frequently within the DMN. This is as opposed to reporting that no such connection exists between the two parcellations.
Despite not setting a threshold for network connectivity, the frequency and strength associated with certain connections identified as part of the DMN (e.g., the connection between 31pd and s32 which was identified in one subject with an overall strength of 0.04) raise an important question of which connections are critical for the functionality of the network. Answering this question is beyond the scope of this study, and further research is needed to understand which connections within the DMN are most critical for the successful functioning of the DMN.

| CON CLUS IONS
We present a preliminary anatomic model of the default mode network. Further studies may refine this model with the ultimate goal of clinical application.

Dr. Sughrue is the Chief Medical Officer of Omniscient
Neurotechnologies. No products directly related to this were discussed in this paper. All other authors have no financial interest or potential conflicts of interest.

E TH I C A L S TATEM ENT
The current study does not have any ethical considerations as it is a meta-analysis on previously completed data.

PE E R R E V I E W
The peer review history for this article is available at https://publo ns.com/publo n/10.1002/brb3.1976.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author (MS), upon reasonable request.