Professional chess expertise modulates whole brain functional connectivity pattern homogeneity and couplings

Previous studies have revealed changed functional connectivity patterns between brain areas in chess players using resting-state functional magnetic resonance imaging (rs-fMRI). However, how to exactly characterize the voxel-wise whole brain functional connectivity pattern changes in chess players remains unclear. It could provide more convincing evidence for establishing the relationship between long-term chess practice and brain function changes. In this study, we employed newly developed whole brain functional connectivity pattern homogeneity (FcHo) method to identify the voxel-wise changes of functional connectivity patterns in 28 chess master players and 27 healthy novices. Seed-based functional connectivity analysis was used to identify the alteration of corresponding functional couplings. FcHo analysis revealed significantly increased whole brain functional connectivity pattern similarity in anterior cingulate cortex (ACC), anterior middle temporal gyrus (aMTG), primary visual cortex (V1), and decreased FcHo in thalamus and precentral gyrus in chess players. Resting-state functional connectivity analyses identified chess players showing decreased functional connections between V1 and precentral gyrus. Besides, a linear support vector machine (SVM) based classification achieved an accuracy of 85.45%, a sensitivity of 85.71% and a specificity of 85.19% to differentiate chess players from novices by leave-one-out cross-validation. Finally, correlation analyses revealed that the mean FcHo values of thalamus were significantly negatively correlated with the training time. Our findings provide new evidences for the important roles of ACC, aMTG, V1, thalamus and precentral gyrus in chess players. The findings also indicate that long-term professional chess training may enhance the semantic and episodic processing, efficiency of visual-motor transformation, and cognitive ability.


Introduction
Long-term skill acquisition and the repetitive rehearsal of skills can result in the representative neural structural and functional changes (Draganski et al., 2004;Gaser & Schlaug, 2003a;Gaser & Schlaug, 2003b, Maguire et al., 2000, Song et al., 2020a, Zou et al., 2012. Chess is seen as a typical example for an expertise task requiring domainspecific experience to study brain structural and functional plasticity. Resting-state functional magnetic resonance imaging (rs-fMRI) has been applied to investigate the changed intrinsically functional activities and interactions in chess players (Duan et al., 2012a(Duan et al., , b, 2014Greicius et al., 2003, Song et al., 2020aWang et al., 2020b). These findings suggested that chess experts are different from novice players in brain functional organization of local connections and global topologies.
Mounting studies have documented that brain functions are determined by its external functional connectivity patterns, i.e., connectivity fingerprints (Passingham et al., 2002;Toro et al., 2008;Zhang et al. 2016a). To characterize the functional similarity of whole brain functional connectivity pattern, several data-driven methods have been proposed. Zang and colleagues (Zang et al., 2004) developed a regional homogeneity (ReHo) method using Kendall's coefficient concordance (KCC) (Kendall and Gibbons) to measure the similarity of the time series of a given voxel and its nearest 26 neighbors. Integrated local correlation (ILC) used for assessing the brain local coherence was introduced (Deshpande et al., 2009). Tomasi and Volkow proposed functional connectivity density (FCD) to further characterize the regionally functional homogeneity in human brains (Tomasi & Volkow, 2010). Buckner and colleagues proposed the functional connectivity strength (FCS) as the average of the correlations between this voxel and all other voxels in the brain as the "degree centrality" (Buckner et al., 2009). Based on the ReHo and FCS methods, our previous studies explored the similarity of time series and identified the changed functional activities in chess players compared with novices (Song et al., 2020a, b). Although several different measurements have been proposed to characterize the similarities of time series and regionally functional activities in chess players, it remains largely unclear whether longterm and intensive practice of chess expertise can lead to the changed voxel-wise whole brain functional connectivity pattern similarity. Exploring the whole brain functional connectivity pattern similarity changes will provide more convincing evidences for localizing the exact functional modulation in chess players. Whole brain functional connectivity homogeneity (FcHo) method is defined by measuring the similarity of whole brain functional connectivity pattern of a specific voxel with that of its nearest 26 neighborhood voxels (Wang et al., 2018b). Compared with FCD and FCS method, FcHo directly measures the similarity of the whole brain functional connectivity map of a given voxel with the nearest neighborhood voxels instead of counting the number of connectivities above a specific threshold. Compared with ReHo calculating the time-series similarity, FcHo can better identify the association cortex areas with high FcHo values. In addition, this approach has been used to reveal the whole brain functional connectivity pattern abnormalities in depression and the neural mechanism of depression patients after electroconvulsive therapy (Wang et al., 2019b(Wang et al., , 2020a. Thus, FcHo is developed to better delineate the voxel-wise similarity of whole brain functional connectivity pattern (Wang et al., 2018b). In this study, our goal is to reveal whether/ how a large amount of chess practice alters the voxel-wise whole brain functional connectivity pattern similarity and functional couplings using FcHo and seed-based functional connectivity analysis methods. The rs-fMRI data of 28 chess players and 27 gender-, age-and education-matched novice players was acquired, and a voxel-wise FcHo map for each subject was calculated. The resting-state functional connections of the brain areas with changed FcHo were mapped to further reveal corresponding changed functional networks in chess players. Moreover, multivariate pattern analyses using linear support vector machine (SVM) was employed to test whether these altered neural indices can be served as markers in distinguishing the chess playing levels.

Participants
Twenty-eight chess players (female/male = 10/18, mean and standard deviation of age = 27.64 ± 9.15 years; mean and standard deviation of education = 13.43 ± 2.71 years) and 27 novice players (female/male = 15/12, mean and standard deviation of age = 26.37 ± 6.68 years; mean and standard deviation of education = 14.24 ± 3.06 years) were used in this study. This data-set was accessed from the "1000 Functional Connectomes Project" https://fcon_1000.projects. nitrc.org/indi/pro/wchsu_li_index.html. These professional chess players had been seriously training regularly (training time: 4.17 ± 1.72 h/day). The 27 gender-, age-, and education-matched novice players understand the rules and simple strategies of Chinese chess playing (gender: p = 0.14; age: p = 0.57; education: p = 0.31). No differences on observation skills or clear-thinking ability were found between these two groups (Li et al., 2015). All participants were right handed and had no history of psychiatric or neurological disorders. Written informed consent was obtained from each subject and approval was obtained through the local Institutional Review Board of the West China Hospital of Sichuan University. The detailed information of this dataset can be found in a previous study (Li et al., 2015).

Rs-fMRI data acquisition
The rs-fMRI data were acquired on a 3.0 T Siemens Trio system at the MR Research Center of West China Hospital of Sichuan University, Chengdu, China. All MR scans were performed when subjects were relaxed with their eyes open and fixated on a cross-hair centered on the screen. A T2-weighted gradient echo planar imaging (EPI) sequence was used to collect the fMRI images. A total of 205 whole brain volumes were acquired using the following parameters: Repetition time (TR) = 2000 ms, Echo time (TE) = 30 ms, flip angle = 90°, axial slice thickness = 5 mm, with no gap, slice number = 30, voxel size = 3.75 × 3.75 × 5 mm 3 .

Rs-fMRI data preprocessing
The rs-fMRI images were preprocessed including the following steps. The first 10 volumes were removed to facilitate magnetization equilibrium effects, and the corrected time series was realigned to the first volume for head motion correction. The data was discarded if the head-movement exceeded 1.5 mm of translation or 1.5 。 of rotation in any direction. All fMRI images were normalized to the Montreal Neurological Institute (MNI) EPI template and resampled to 333 mm 3 . Friston 24-parameter model of head motion, white matter, cerebrospinal fluid, and global mean signals were then regressed out and the functional images were filtered with a temporal band-path of 0.01-0.1 Hz. In order to further exclude the head motion effects on functional connectivity analyses, a scrubbing method was conducted to censor each subject's bad fMRI images finding out the mean frame displacement (FD) which was above 0.5 mm. One volume before and two volumes after the bad volume were discarded (Power et al., 2012). Only the participants with the fMRI images more than half of the total time points were kept for the following analyses. The detailed procedures of fMRI preprocessing could be found in our previous studies (Duan et al., 2012a(Duan et al., , b, 2014Greicius, et al., 2003;Song 2020a, b;Wang et al., 2020a). There is no significant difference in head motion between these two groups (p = 0.33). For resting-state functional connectivity analyses, the fMRI data were smoothed with a Gaussian kernel (full-width at half maximum (FWHM) of 6 mm). The global signal was not regressed to ensure that the obtained results were reliable because the whole brain signal regression will exaggerate anti-correlation (Saad et al. 2013;Wang et al., 2018a).

Whole brain voxel-wise FcHo analyses
The FcHo value measured using KCC was calculated for each voxel in each subject. FcHo of a given voxel was calculated by computing the KCC of the whole brain functional connectivity pattern of this voxel with those of its nearest 26 neighbors (see the following formula). The same procedures were performed for all the voxels of the whole brain, and an FcHo map for each subject was obtained. Then, all the FcHo maps were smoothed with 6 mm FWHM for statistical analyses. After obtaining the FcHo maps, the mean FcHo map for chess and novice players were separately calculated to delineate the distribution patterns. A two-tailed two sample t-test was performed to reveal the changed whole brain functional connectivity homogeneity between chess and novice players. The significance was determined using Alphasim correction method with p < 0.05 and voxel-level p < 0.001.
where R i is the sum rank of the ith voxel of the whole brain; R = ((K + 1)N)/2 is the mean of the R i ; N is the number of a given voxel and its nearest neighbors (N = 26); K is the number of whole brain voxels.

Functional connectivity analyses
A whole brain functional connectivity analysis was performed to identify the changed functional couplings of the identified brain regions which showed significantly different FcHo between chess and novice players. To calculate the functional connectivity, the mean time series of the identified brain regions showing changed FcHo were first extracted. Then the strength of the functional connectivity was measured by means of Pearson's correlations between the averaged time series of the brain areas showing changed FcHo and the voxels of the rest brain. Subsequently, the Fisher's Z transformation was applied to improve the normality of the original correlation maps. A two-sample t-test was performed to determine areas with significantly different functional connectivity between chess and novice players. The significance was determined by using Alphasim correction method with p < 0.05 and voxel-level p < 0.001.

Multivariate pattern analyses using SVM
To explore whether the identified neural indices might serve as markers in distinguishing the chess playing levels, a linear SVM approach was performed (Chang & Lin, 2011). The FcHo and functional connectivity values of the brain regions that showed significant differences between chess and novice players were used as the features for classification. Because of the small number of samples, a leave-one-out cross-validation strategy was used to estimate the generalization ability. The performance of a classifier was assessed using the classification accuracy, sensitivity and specificity based on the results of the cross-validation.

Correlation analyses
To explore the relationship between the rs-fMRI indices and the amount of time chess players spent on professional training, correlation analyses were performed between the mean FcHo and functional connectivity values and the professional training time. The significance was set at p < 0.05.

FcHo pattern in chess and novice players
Spatial distribution of mean FcHo maps for chess and novice players were shown in Fig. 1. High FcHo was mainly observed in association cortical areas, such as parietal cortex, frontal cortex, lateral temporal cortex, dorsal insula, cuneus, posterior cingulate cortex, and dorsomedial prefrontal cortex.

Changed FcHo in chess players
Statistical analyses identified significant differences in FcHo maps between chess and novice players. Significantly increased FcHo in anterior cingulated cortex (ACC), anterior middle temporal gyrus (aMTG) and primary visual cortex (V1) was found in chess players compared with novices ( Fig. 2 and Table 1). A significant decrease of FcHo in chess players was found in thalamus and precentral gyrus ( Fig. 2 and Table 1).

Changed functional connectivity in chess players
To reveal the changed functional connectivities to the brain areas showing changed FcHo, whole brain functional connectivity analyses were performed. Decreased functional connections between precentral gyrus and V1, and decreased functional connections between V1 and precentral gyrus, parietal opercula/ posterior insula in chess players compared with novices were identified ( Fig. 3 and Table 1).

Classification results
Using the combined features of the mean FcHo and functional connectivity values in the brain regions showing differences between chess and novice players for classification, the linear SVM classifier achieved an accuracy of 85.45% [85.71% for chess players (sensitivity), 85.19% for novice players (specificity)] (Fig. 4).

Correlation analyses
Correlation analyses revealed that the mean FcHo values of thalamus were significantly negatively correlated with the amount of time chess players spent on professional training (r = -0.45, p = 0.02) (Fig. 5), which suggested a direct association between the alterations of FcHo and the chess training frequency.

Discussion
In the present study, FcHo and functional connectivity analysis methods were used to identify the changed whole brain functional connectivity pattern similarity and functional couplings in delineating how the long-term professional training modulates brain functional connectivity patterns. FcHo analyses identified significantly increased whole brain functional connectivity pattern similarity in ACC, aMTG and V1, and decreased whole brain functional connectivity pattern similarity in thalamus and precentral gyrus in chess players. Resting-state functional connectivity analyses further identified chess players showed decreased functional connections between Fig. 1 Distribution patterns of functional connectivity pattern homogeneity (FcHo). One-sample t-tests were used to identify the whole brain FcHo distribution pattern in chess and novice players. The high FcHo is mainly distributed in the association cortex such as parietal, frontal, lateral temporal and occipital lobes V1 and precentral gyrus. Correlation analyses revealed that the mean FcHo values of thalamus were significantly correlated with the amount of time chess players spent on professional training. Our findings provide new evidences for the important roles of ACC, aMTG, V1, thalamus and precentral gyrus in chess players.

aMTG in chess players
In our study, increased FcHo in aMTG was found in professional chess players compared with novices. This finding was supported by a previous study which used the same participants and found significant deactivation in this area Fig. 2 FcHo differences between chess players and novices. A two-sample t-test was used to compare the FcHo maps between chess and novice players. FcHo analyses identified significantly increased whole brain FcHo in anterior cingulated cortex (ACC), anterior middle temporal gyrus (aMTG) and primary visual cortex (V1), and decreased whole brain FcHo of the thalamus and precentral gyrus in chess players compared with novices in professional Chinese chess players (Duan et al., 2012a). Previous resting-state functional connectivity analyses revealed aMTG was dominantly connected with DMN related regions, which indicates that aMTG is an important part of DMN (Buckner et al., 2008;Xu et al., 2015Xu et al., , 2019 and plays an important role in episodic memory (Buckner, et al., 2008). Additionally, the spatial and functional convergence of the DMN and semantic memory system has been explored (Binder et al., 2009;Wirth et al., 2011) Fig. 3 Different functional couplings between chess and novice players. A two-sample t-test was used to identify the significantly different functional interactions to the changed FcHo regions in chess players compared with novices. The functional connectivity analyses identified decreased functional connections between precentral gyrus and V1, and decreased functional connections between V1 and precentral gyrus, parietal opercula and posterior insula in chess players Fig. 4 Classifying chess from novice players. Multivariate pattern analyses using support vector machine (SVM) was applied to determine whether identified neural indices might serve as markers for distinguishing the chess playing level. The regions with changed FcHo and functional connections were used as the features for classification. A leave-one-out cross-validation strategy was used to estimate the generalization ability of our classifier. The classification accuracy, sensitivity and specificity were showed demonstrating aMTG is also important for semantic retrieval (Cho et al., 2012). When encountering the chess playing situations that commonly appear in games, the 'chunks' held in long-time memory are activated and chess players produce ideas for the best following move to advance better performances (Wan et al., 2011). Thus, we speculated that higher FcHo in aMTG found in professional chess players may be related to higher efficiency and less effort for semantic and episodic processing related to chess games.

ACC in chess players
The ACC, a limbic structure, has connections with a set of other limbic and related areas involved in emotion and reward-related processing (Zhang et al., 2014(Zhang et al., , 2016. The ACC performs goal-directed (top-down) cognitive processing which is critical for identifying and responding to stimuli relating to reward in a largely irrelevant sensory world (Chelazzi et al., 1993(Chelazzi et al., , 1998Desimone, 1998;Fuster & Jervey, 1982;Miller et al., 1993). During the chess playing, a chess expert needs to active the 'chunks' held in longtime memory, takes into account the outcomes received after actions, and will not select an action if the goal has been devalued (Kolling et al., 2016;Rushworth et al., 2012) for the best following move. In our study, significantly increased FcHo value in ACC was found in chess experts compared with novice players. Our findings provide critical functional evidence which supports the ACC functions as part of topdown executive networks.

Thalamus and precentral gyrus in chess players.
Many studies on skill acquisition have reported significant changes in the brain areas according to the development of specific skills (Duan et al., 2012a;Gaser & Schlaug, 2003a;Maguire, et al., 2000;Munte et al., 2002;Schlaug, 2001). The human thalamus was a comprehensive hub for functional brain networks and information integrating across cortical networks (Hwang et al., 2017). Spatial visual processing, attention, memory and expression of goal-directed behaviors are thought to be mediated by the thalamus (Browning et al., 2015;Haber & Calzavara, 2009;Sommer & Wurtz, 2006). Previous studies have found an important role of the thalamus in chess playing (Duan, et al., 2014;Wang et al., 2020a;Zhou et al., 2018). In our study, the chess experts showed significantly decreased FcHo value in thalamus in comparison with novice players. Moreover, the mean FcHo values of thalamus were negatively correlated with the professional training time, which is consistent with a previous study finding of a strong relationship between the thalamic radiation and cognitive ability and training frequency (Zhou et al., 2018). All the converging evidences suggest that the thalamus is important in the processes of board-pattern spatial perception, attention, and next-move generation, as well as the long-term input from the motivational and learning processes involved in the problem solving of chess (Weng et al., 2017). In addition, decreased FcHo in precentral gyrus was also found in professional chess expert in this study. The precentral gyrus plays a critical role in motor planning and execution (Liu et al., 2018). Thus, decreased FcHo in both thalamus and precentral gyrus may be related to increased inhibitory control of thalamusprecentral gyrus loop in chess expert.

Functional connectivity between V1 and precentral gyrus
Decreased functional connectivity between V1 and precentral gyrus, parietal opercula and posterior insula was found in chess players compared to novice players. V1 is the main cortical area of the visual system that receives visual information from the external world. The precentral gyrus and parietal opercula/posterior insula are the important parts of the sensorimotor network that is involved in motor planning, execution and control (Eickhoff et al., 2010, Wang et al., 2019a. The decreased functional connections between V1 and precentral gyrus indicate the higher efficiency in vision-motor transition in chess experts than novices.
There are some limitations in our study. First, the current study is a cross-sectional study. The longitudinal study is required to better reveal the changes of whole brain functional connectivity pattern similarity in chess players. Second, the sample size was small, and the conclusion required further validation with data from large samples. Third, our study focused only on resting-state functional connectivities and future studies are needed to integrate multimodal connectivity information.

Conclusion
The present study assessed the changes of voxel-wise whole brain FcHo and resting-state functional connectivities in chess experts. The increased FcHo in ACC, aMTG and V1, decreased whole brain functional connectivity pattern similarity in thalamus and precentral gyrus and decreased functional couplings between V1 and precentral gyrus, parietal opercula and posterior insula were identified in chess experts. These findings suggest enhanced semantic and episodic processing ability, top-down cognitive regulation, and efficiency in vision-motor transition in chess experts.