Topographic preservation of functional connectivity among the association areas in the human cerebral cortex

In the human sensorimotor cortex, some long-range corticocortical connections appear to preserve a fine-scale topology, in which physically close locations in the cortical region are functionally connected to physically close locations in other cortical regions. However, little is known about whether such topography preservation is unique to the sensorimotor areas or is general across other cortical areas. To investigate this question, we measured voxel-level functional connectivity using functional magnetic resonance imaging (fMRI) and visualized the fine-scale spatial organization of the connectivity patterns across the cortical surface. We found topographical preservation across regions, including the default mode network. Our results suggest that the topographical preservation of functional connectivity is not restricted to the sensorimotor cortex but also occurs in the association cortex.

, and category representation in the 36 association areas [Huth et al., 2012]. Huth et al., 2016[Huth et al., 2016 reported that the representation of speech-37 content changes smoothly from sensory to more abstract representation across the temporoparietal junction (TPJ), 38 which is an association area within the default mode network (DMN) [Greicius et al., 2003, Fox et al., 2005. 39 Some of these smooth representational gradients have been linked to the topographic organization of 40 connectivity patterns between distinct brain regions. These patterns are known as connectopic patterns, in which 41 neighboring locations in a cortical region are connected with neighboring locations in a distant region [Thivierge 42 and Marcus, 2007]. This type of organization has been observed in the human sensorimotor cortex: blood oxygen

47
However, such topographical preservation of connectivity has not been widely observed in the association areas, 48 even though smooth gradients of representation have previously been observed in one association area-the TPJ 49 [Huth, 2016]. is independent of the resolution of cortical parcellation or the duration of the data collection used to calculate 56 functional connectivity [Vanderwal et al., 2017]. Less head motion was observed in a movie-viewing task than in a 57 resting-state task [Vanderwal et al., 2015], possibly leading to higher-quality-data acquisition in a movie-viewing 58 task. Therefore, a naturalistic viewing paradigm considered to be an efficient approach for the measurement of fine-59 scale functional connectopic patterns.

60
In this study, we addressed the question whether a topographic organization of functional connectivity 61 (fcTO) is present in the association areas. We specifically hypothesized that fcTO exists between the TPJ and the 62 association areas in the DMN. We used functional magnetic resonance imaging (fMRI) to measure BOLD signals 63 while eight subjects watched movie clips. To quantify the intensity of connectivity at a fine scale, we calculated the 64 correlation between BOLD signals in each voxel in the TPJ, and each voxel in the external regions. To visualize the 65 organization of the connectivity, we labeled each voxel in the external regions using the spatial coordinates of the 66 strongly-connected voxels in the TPJ (see Supplemental Information). By visualizing the cortical map of the spatial 67 coordinates, we demonstrated fcTO preservation within the DMN. We also performed a comprehensive automated 68 search for fcTO for 156 anatomically-defined target regions across the cortex. We detected fcTO across several 69 networks, including the DMN-related areas.

72
To capture the topographic organization of functional connectivity (fcTO), we developed a method for 73 visualization of the structure of fine-scale connectivity structure between a seed region and external regions in the 74 cerebral cortex (Figure 1). We focused on one seed region (e.g., the right transverse occipital sulcus), and labeled 75 each voxel in the seed region with its spatial coordinate. For each voxel in the external region we calculated the 76 strength of the connection to each voxel in the seed region using Pearson's correlation coefficient between the 77 BOLD-signal time courses [Kahnt et al., 2012, Cauda et al., 2011. The voxel in the external region was labeled 78 with the representative spatial coordinates of strongly-connected seed voxels (see Supplemental Information). If an 79 external region showed a spatial gradient of labels with a similar topological order of the seed spatial coordinates, 80 fcTO is considered to be preserved between the external region and the seed region. To visualize the spatial 81 coordinates of the seed region and the external region, we assigned different colors to the spatial coordinates on the 82 cortical connectopic map. An example of a cortical map was shown in Figure 1B. In this map connections are 83 indicated by color: voxels in the seed region with a specific color (e.g., turquoise) had been shown to have a strong 84 connection with voxels in the external regions represented by the same color.

142
Effect of stimulus-induced BOLD signals on functional connectivity measurements

143
The connectopic map for the TPJ was created using BOLD signals measured during a natural-movie viewing task.

144
It is possible that stimulus-induced co-activation of BOLD signals across voxels might be interpreted as functional

165
To confirm that the results of the analysis using TS values were consistent across subjects, we 166 arranged the 156 seed regions in the ascending order of the TSs computed from a single subject (S1) and arranged 167 the TSs computed from each of the seven remaining subjects (S2-S8) in the same order ( Figure 4B). The TSs 168 averaged across the seven subjects gradually decreased ( < 10 −47 , F test on the linear regression, Figure 4B).

169
To further examine which regions show high TS across subjects, we selected the seed regions with 170 the top 30 TSs for each of the eight subjects. The seed regions selected in more than six subjects are listed in Table   171 1. We found that more than half of the selected seed regions are related to the TPJ, suggesting that the fcTO 172 preservation was found more frequently in the TPJ than in the other regions.

325
To confirm whether the fcTO was preserved between the TPJ and the external regions, we 326 manually defined seven representative lines with strong connections, and examined whether the spatial coordinates 327 in the seed were significantly similar to those on each line ( Figure 3A). Lines were defined on the flattened cortical 328 map. The lines were defined firstly for a single subject S1. For each of the seven remaining subjects (S2-S8), the 329 lines corresponding anatomically to S1's lines were defined. We summarized the anatomical labels obtained from 330 the regions upon which each line was laid on ( Figure S2). As a result, at least one or two labels for each line were 331 shared across all subjects, and the remaining three to six labels were neighbors to the shared labels (the total  Figure S4A-1). Then, to avoid detecting overwrapping regions multiple times, we masked out a 359 circular area with a radius of 140 mm at the first peak location ( Figure S4A-2). Subsequently, we searched for the 360 location with the highest connectivity in the remaining region as the second peak location. We again masked out a 361 circle with a radius of 140 mm at the second peak, before searching for the third peak location. This detection step 362 was repeated 15 times, resulting in 15 peak locations were identified as the candidate regions ( Figure S4A-3).

363
To quantify the degree of topography preservation for each of the candidate regions, we set a 364 representative curve line within each candidate region ( Figure S4B). The curve line was optimized so that the 365 spatial coordinates along the line would be as topographically linearly ordered as possible ( Figure S4C) and were 366 obtained using the following three steps. First we obtained the candidate endpoints of a representative curve line.

367
The candidate endpoints were pairs of the three minimum-peak locations and the three maximum-peak locations of 368 the connectivity-strength map within the region. The candidate peak locations were detected iteratively using the 369 detect-and-mask procedure described above to find the global-peak locations, with a circular area of radius 24 mm greater than 0.2. We then calculated the difference between the obtained spatial coordinates (modes) and a linearly-377 ordered coordinates using MSE (as in Figure 3B).  Figure S4D). the TSs in the first subject (S1, Figure 4B) and confirmed that TSs decreased as the arranged-region order 396 increased. To find seed regions that had high TSs consistently across subjects, for each subject, we selected the 397 seed regions having the 30 highest TSs across the 156 seed regions. We then counted the number of subjects in 398 which a given region was selected. The number ranged from one to eight, with the number eight denoting that the 399 seed region was shown in the top 30 seed regions across all subjects. We listed seed regions with more than six 400 occurences in Table 1