Figures
Abstract
Eating disorder is highly associated with obesity and it is related to brain dysfunction as well. Still, the functional substrates of the brain associated with behavioral traits of eating disorder are underexplored. Existing neuroimaging studies have explored the association between eating disorder and brain function without using all the information provided by the eating disorder related questionnaire but by adopting summary factors. Here, we aimed to investigate the multivariate association between brain function and eating disorder at fine-grained question-level information. Our study is a retrospective secondary analysis that re-analyzed resting-state functional magnetic resonance imaging of 284 participants from the enhanced Nathan Kline Institute-Rockland Sample database. Leveraging sparse canonical correlation analysis, we associated the functional connectivity of all brain regions and all questions in the eating disorder questionnaires. We found that executive- and inhibitory control-related frontoparietal networks showed positive associations with questions of restraint eating, while brain regions involved in the reward system showed negative associations. Notably, inhibitory control-related brain regions showed a positive association with the degree of obesity. Findings were well replicated in the independent validation dataset (n = 34). The results of this study might contribute to a better understanding of brain function with respect to eating disorder.
Citation: Lee H, Park B-y, Byeon K, Won JH, Kim M, Kim S-H, et al. (2020) Multivariate association between brain function and eating disorders using sparse canonical correlation analysis. PLoS ONE 15(8): e0237511. https://doi.org/10.1371/journal.pone.0237511
Editor: Tyler Davis, Texas Tech University, UNITED STATES
Received: November 25, 2019; Accepted: July 28, 2020; Published: August 12, 2020
Copyright: © 2020 Lee et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: Data of the discovery set are available from the enhanced Nathan Kline Institute-Rockland Sample (eNKI-RS) database website (http://fcon_1000.projects.nitrc.org/indi/enhanced/access.html). The eNKI-RS Institutional Data Access Committee grants access to researchers who meet the criteria for access to confidential data upon completion of the Data Usage Agreement. Researchers should contact the database administrator to gain access to data. Data of the validation set are not available due to patients’ privacy issues directed by the Institutional Review Board (IRB) of Catholic University of Korea. Data are available for researchers who meet the criteria of the IRB of Catholic University of Korea. Researchers can contact either one of the co-authors (Dr. Se-Hong Kim, iron1600@catholic.ac.kr) or the IRB of Catholic University of Korea (cmcohrp@catholic.ac.kr).
Funding: This work was supported by the Institute for Basic Science (http://www.ibs.re.kr/, grant number IBS-R015-D1, recipient:NA), the National Research Foundation of Korea (http://www.nrf.re.kr/, grant number NRF-2020M3E5D2A01084892, recipient: HP), the MIST (Ministry of Science and ICT) of Korea under the ITRC (Information Technology Research Center) support program (http://iitp.kr, grant number IITP-2020-2018-0-01798, recipient:NA), and the IITP grant funded by the Korean government under the AI Graduate School Support Program (http://iitp.kr, grant number 2019-0-00421, recipient:NA). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors declare no competing interests.
Introduction
Eating disorder is one of the important factors underlying obesity [1–3], which is related to type 2 diabetes, cardiovascular diseases, stroke, and cancers [3–5]. Previous neuroimaging studies have shown that eating disorders were highly associated with the executive function of people with obesity [3,6–9]. Siep et al. found that an individual’s appetite was regulated by the inhibitory control centers of the brain [9]. Val-Laillet et al. reported a significant relationship between eating disorder and altered cognitive- and reward-related brain systems [3]. Our previous studies also demonstrated that executive- and inhibitory-related brain regions had strong associations with eating disorders in people with obesity [6–8]. These studies collectively suggested that neurologic dysfunctions were highly related to the eating disorders of people with obesity.
The association between brain alterations and eating disorders can be effectively explored using neuroimaging modalities such as functional magnetic resonance imaging (fMRI), positron emission tomography (PET), and single-photon emission computed tomography (SPECT) [3,9–11]. However, most existing neuroimaging studies did not make full use of the available information regarding participants’ eating disorders [6–8]. They used summary scores instead of the responses to the full list of questions. For example, the Three-Factor Eating Questionnaire (TFEQ) quantifies behavioral traits of eating disorders and has a total of 51 questions condensed into three factors of dietary restraint, disinhibition, and hunger scores [12,13]. The Eating Disorder Examination Questionnaire (EDE-Q) evaluates the psychopathology of eating disorders and has a total of 34 questions condensed into four factors of dietary restraint, eating concern, shape concern, and weight concern [14,15]. This approach allows researchers to quantify an individual’s eating disorders succinctly, but it does not utilize all the available information, and thus could lead to suboptimal findings regarding neurological alterations in the brain. Thus, considering all the questions in the questionnaire might allow us to better explore the associations between brain function and eating disorder.
Machine learning is widely used in neuroimaging and medical field, and we applied one of machine learning technique, canonical correlation analysis (CCA), for data analysis [16,17]. CCA is a useful method to identify the multivariate association between two types of high-dimensional features (e.g., neuroimaging and questionnaire) [18]. Sparse CCA (SCCA) is an extension of CCA, which aims to identify a sparse set of loading vectors in two types of features using a regularization technique [19–30]. SCCA could be more suitable than the regular CCA since it helps to reduce the overfitting problem, which is becoming more problematic with increased dimensionality of modern neuroimaging data. Associations between brain regions over the whole brain and all the questions of the eating disorder questionnaire could be effectively assessed using the SCCA approach. Especially, incorporating the information on eating disorders in a question-level might allow us to better model the complex associations between brain and behavioral traits of eating.
We adopted a functional connectivity analysis based on graph theory to estimate brain function [31,32]. Graph theory-based connectivity analysis is a representative method to identify the connection strengths of different brain regions, which requires graph nodes and edges [31,32]. Graph nodes are the brain regions defined by pre-defined structural or functional atlases and graph edges are the connections between two different nodes [31,32]. Graph theory-based functional connectivity analysis has been useful for explaining brain alterations in people with obesity, as well as identifying associations between the brain and eating disorders [6–8].
In the current study, we aimed to investigate the multivariate association between the functional connectivity of all brain regions and eating disorders at question-level using the SCCA approach. Our study is not limited to assess obesity-related eating disorders in the obese group, rather we aimed to associate functional connectivity of brain regions and behavioral patterns of eating disorders in a general population. The results of this study might contribute to a better understanding of brain function with respect to eating disorder.
Materials and methods
Imaging data and participants
This study was a retrospective analysis of anonymized data and institutional review board (IRB) approval was obtained at Sungkyunwkan University. All data were obtained with informed written consent in accordance with established human subject research procedures expressed in the Declaration of Helsinki. Our study was performed in full accordance with the local IRB guidelines. For the discovery set, the T1-weighted structural MRI and resting-state fMRI (rs-fMRI) data were obtained from the enhanced Nathan Kline Institute-Rockland Sample database [33]. All imaging data were obtained using a 3-T Siemens Magnetom Trio Tim scanner. The scanning parameters of the T1-weighted structural data were as follows: repetition time (TR) = 1900 ms, echo time (TE) = 2.52 ms, flip angle = 9°, field-of-view (FOV) = 250 mm × 250 mm, 1 mm3 voxel resolution, and 176 slices. Those of the rs-fMRI data were as follows: TR = 645 ms, TE = 30 ms, flip angle = 60°, FOV = 222 mm × 222 mm, 3 mm3 voxel resolution, 40 slices, and 900 volumes. Among the total of 650 participants, 320 participants have full demographic information, obesity-related clinical scores (body mass index [BMI] and waist-to-hip ratio [WHR]), and TFEQ scores. BMI and WHR are calculated from the vitals (i.e., height, waist measurement, hip measurement, and weight) that were recorded by staff with a standardized procedure. Thirty-six subjects with medical conditions (e.g., attention deficit hyperactivity disorder, balance dysfunction with severe motion sensitivity, depression, diabetes, high blood pressure, high cholesterol, and hypoglycemia) were excluded and total 284 participants were considered in this study. Detailed demographic information is reported in Table 1. The study participants had a large variation in BMI as stated in Table 1 and the S1 Fig in S1 File.
Data obtained from Saint Vincent’s hospital in South Korea was used for validation. All data were obtained with informed written consent in accordance with established human subject research procedures expressed in the Declaration of Helsinki. Our study was performed in full accordance with the local IRB guidelines. The dataset includes T1-weighted data, rs-fMRI, full demographic information, BMI, WHR, and Eating Attitudes Test (EAT-26) scores of 34 participants [34]. The T1-weighted structural MRI and rs-fMRI data were scanned using a Siemens Magnetom 3T scanner housed at St. Vincent’s Hospital. The image acquisition parameters of the T1-weighted structural MRI were as follows: TR = 1900 ms; TE = 2.46 ms; flip angle = 9°; FOV = 250 mm × 250 mm; 1 mm3 voxel resolution; and 160 slices. Those of the rs-fMRI data were as follows: TR = 2490 ms; TE = 30 ms; flip angle = 90°; FOV = 220 mm × 220 mm; 3.4 × 3.4 × 3.0 mm3 voxel resolution; 36 slices; 150 volumes.
Eating disorder scores
The eating disorder scores were measured using the TFEQ, which contains a total of 51 questions [12]. All the questions were assigned to one of the three factors of dietary restraint, disinhibition, or hunger. The question list of the TFEQ is stated in the Supplementary Tables in S1 File. Each question had a binary response and the score of each factor was calculated by summing the responses of the assigned questions. In this study, the response to each question was considered rather than the score for each factor.
EAT-26 is a reliable screening tool for eating disorders [34–38] and it was used to access the degree of eating disorders in the validation set. This has 26 questions and the answers are scored among 0, 1, 2, 3, 4, or 5 (correspond to never, rarely, sometimes, often, usually, or always respectively). Those scores are rescored into 0 to 3 and summed up into three subscales (diet, bulimia and food preoccupation, and oral control).
Data preprocessing
The T1-weighted structural data and rs-fMRI data were preprocessed using a Fusion of Neuroimaging Preprocessing (FuNP) pipeline that integrates AFNI, FSL, and ANTs software [39–42]. The T1-weighted structural data were preprocessed as follows: the distortion arising from the magnetic field inhomogeneity was corrected and non-brain tissues were removed. The rs-fMRI data were preprocessed as follows: the volumes from the first 10 s of the scan were removed and distortions caused by head motions were corrected. Intensity normalization of the 4D volumes was applied and nuisance variables of the cerebrospinal fluid, white matter, head motion, and cardiac- and large-vein-related artifacts were removed using FIX software [43]. The artifact-removed rs-fMRI data were registered onto preprocessed T1-weighted structural data and then subsequently registered onto 3 mm isotropic Montreal Neurological Institute (MNI) standard space. Spatial smoothing with a full width at half maximum of 5 mm was applied. The preprocessing pipeline is summarized in Fig 1A.
(a) Data preprocessing pipeline of T1-weighted structural data and rs-fMRI data. (b) Functional connectivity analysis steps applied to preprocessed rs-fMRI data. The mean BOLD signal of each region as defined by the BNA atlas was calculated and the connectivity matrix was computed using Pearson’s correlation coefficients from the mean BOLD signal of two different regions. This connectivity matrix was soft-thresholded and then z-transformed. The between centrality (BC) values were estimated from this matrix. (c) Sparse canonical correlation analysis (SCCA) was adopted to find an association between brain function and eating disorders. The BC values of all 246 brain regions and 51 responses of the TFEQ assessment were considered. (d) The post-hoc correlation analysis was conducted between obesity-related scores and brain regions showing associations with eating disorders as result of (c). Pearson’s correlation was computed from the BC values of the corresponding brain regions and BMI/WHR scores.
Functional connectivity analysis
Graph theory-based functional connectivity analysis was adopted in this study. Graph nodes were the brain regions defined by the Brainnetome atlas [44] and the graph edges were defined by Pearson’s correlation coefficients of time series between two different nodes. Recent resting-state fMRI studies suggested that inter-regional functional connectivity shows a small-world property linked with scale-free topology [45–47]. The soft-thresholding approach was applied to the correlation coefficients to satisfy the scale-free topology using the following formula: {(r+1)/2}β, where r is the correlation coefficient and β is the scale-free index that was set to six [48,49]. The soft-thresholded correlation coefficients were transformed to z-values using Fisher’s r-to-z transformation. The betweenness centrality, a graph measure estimating the importance of a given node, was calculated by counting the number of shortest paths between any two nodes that run through that node [31,32]. The steps for functional connectivity analysis are represented in Fig 1B.
Association between functional connectivity of the brain and eating disorders
The multivariate association between the functional connectivity of the whole brain and eating disorders was explored using the SCCA approach. The betweenness centrality values of 246 brain regions and responses of 51 questions in TFEQ assessment were considered (Fig 1C). The goal of SCCA is to compute loading vectors (u and v) of two feature matrices, X and Y, by maximizing the correlation between the linear combinations of the features from two matrices (Xu and Yv), where X is the betweenness centrality values of all brain regions (n × 246 regions), Y is the responses to all questions in the TFEQ assessment (n × 51 questions), and n is the number of participants. L1 regularization was used to control the sparse relationship between the two different feature matrices [50]. The objective function of SCCA can be defined as (1) We can rewrite the objective function as (2) Solving Eq (1) under the sparsity constraints lead to solving Eq (2). We initially tuned the regularization parameters (i.e., βu and βv) using five-fold cross-validation where the built SCCA model from the training fold was tested using various beta values in the left-out fold to maximize the canonical correlation. The final beta values were determined by averaging them across the cross-validation. The SCCA is a multivariate approach, which aims to optimize the association between one set of variables (i.e., functional connectivity of various brain regions) and another set of variables (i.e., questions in the TFEQ) by maximizing the canonical correlation between the linear combination of a subset of each variable and relevant loading vector. This approach does not involve multiple tests of univariate statistics and therefore, we evaluated the model just once. Thus, there are no corrected p-values in that sense. Bootstrapping was performed in order to assess the reliability of the results, randomly selecting 80% of subjects for 1000 times. The loading vectors were computed for each bootstrapped subset and the frequently selected features were compared with those obtained from the full data.
Correlation with obesity-related clinical scores
Post-hoc correlation analysis between the betweenness centrality values of the identified brain regions, which showed an association with eating disorders, and the obesity-related clinical scores of BMI and WHR was conducted at two levels, one at the large-scale network and the other at the smaller-scale region (Fig 1D). This was to explore whether brain regions were associated with obesity as well as eating disorders. The r- and p-values were calculated and the regions with significant correlation with BMI/WHR were selected (false discover rate [FDR] corrected p < 0.05).
Validation of the association between brain regions and eating disorders
Independent validation was conducted using data of the Saint Vincent’s hospital in Suwon, South Korea. The betweenness centrality was calculated from rs-fMRI data and the same atlas was used. SCCA was applied to the centrality values of 246 brain regions and responses to 26 questions of EAT-26. The optimal regularization parameter was obtained in the same way described before, but three-fold cross-validation was used in parameter tuning due to the small sample size.
Results
Association between functional connectivity of the brain and eating disorders
Twenty-seven brain regions and 19 questions were selected as a result of the SCCA. The canonical correlation coefficient between two linear combinations of two features was 0.4945. The brain regions with a large magnitude of the loading vector (i.e., the strong association with TFEQ questions) were primarily located at the frontoparietal and limbic networks that process cognition- and reward-related functions (Fig 2). The questions in the TFEQ with a large loading vector magnitude were primarily involved in factor 1, dietary restraint, which indicates the ability to control food intake. Detailed information for selected features is stated in the Supplementary Tables in S1 File. Positive associations were observed between the questions of dietary restraint factor and cognition-related brain regions, while negative associations were found between the same factor with reward-related brain regions. The results indicate that the inhibitory function in cognition-related brain regions might regulate eating disorders, while hunger-related responses in reward-related brain regions could lead to a failure in dietary regulation.
The color scale indicates a strong association with blue and red colors. The regions in red (strong positive correlation with dietary restraint) are mostly cognition-related regions. The regions in blue (strong negative correlation with dietary restraint) are mostly reward-related regions. IPL, inferior parietal lobule; OrG, orbital gyrus; INS, insula; FuG, fusiform gyrus; CG, cingulate gyrus.
Bootstrapping was performed for 1000 times and the frequently selected features were compared with those from the main results. Twenty-seven selected brain regions of the main results were compared with the top 27 frequently selected brain regions from the bootstrapping. A total of 21 regions (77.78%) overlapped. Nineteen selected questions of the main results were compared with the top 19 frequently selected questions from the bootstrapping. A total of 18 questions (94.74%) overlapped and overlapped features were mostly assigned in dietary restraint. The list of the most frequently selected features is stated in the Supplementary Tables in S1 File.
Correlation with obesity-related clinical scores
To explore whether the eating disorder-related brain regions were associated with obesity, post-hoc correlation analysis between the betweenness centrality values of the networks in the behavioral domain of the identified regions and BMI/WHR was conducted. Twenty-seven identified brain regions were mapped into five behavioral domains which appeared dominantly in the identified regions (cognition-, reward-, sensorimotor-, visual-, and language-related region) and the averaged centrality values of each domain were computed. The results are shown in Table 2. The cognition-related network showed a significant correlation with WHR with r-value of 0.1588 and FDR corrected p-value of 0.0366. We further explored the association in six regions in the cognitive domain with WHR to give a fine-grained explanation of the results. Table 3 shows the results of the second correlation analysis. Two regions (rostal area 35/36 and caudal area 40) showed a significant correlation with WHR, indicating a significant association among cognition-related brain networks, eating disorders, and obesity.
Validation of the association between brain regions and eating disorders
As a result of SCCA, 16 brain regions and eight questions were selected. The canonical correlation coefficient between two canonical features was 0.8851. The cognition- and reward-related regions showed a large magnitude of loading vectors and the ‘diet’ subscale had the largest magnitude in loading vectors. Thus, high associations between cognition-/reward-related brain regions and diet were observed in the validation. The diet subscale in EAT-26 is not equal to the dietary restraint in TFEQ at the subscale-level but it includes some questions related to dietary restraint and might be partially comparable. In this sense, these results of the validation were largely compatible with those of the discovery set. The list of selected features is stated in the Supplementary Tables in S1 File.
Discussion
Despite extensive research [3,6–10], association between brain connectivity topology and behavioral patterns of eating disorder is underexplored, which makes it difficult to understand the underlying mechanisms of brain function related to eating disorders. Leveraging a multivariate association analysis, we established that reward and frontoparietal networks showed significant associations with eating disorders of dietary restraint. Regions involved in frontoparietal network positively related to restraint eating, while reward regions showed a negative association. These results support that increased inhibitory control of the brain may suppress errant eating behavior, while increased reward signaling induces loss of control for eating. Further association with the degree of obesity revealed a potential link between inhibitory control and obesity. Together, our work provides a new perspective on understanding functional connectivity topology associated with eating disorders.
In this current study, we adopted SCCA to identify highly contributable brain regions associated with eating disorders across a wide range of BMI. We found frontoparietal network to be related to restraint eating, supporting the role of this brain region for regulating inhibitory behavior [51–54]. In addition, reward-related brain regions showed significant associations with eating disorders as well. Unlike inhibition control-related brain regions, these regions are involved in hunger domain, which encodes food-related responses leading to an individual to feel hunger [55–60]. Previous neuroimaging studies proved that dysfunction in frontoparietal network broke the executive and inhibitory control system, yielding aberrant eating behaviors and subsequently weight gain [6,7,53,61–65]. Molecular studies support our findings in a biological context by suggesting atypical signaling from neuroendocrine factors of insulin, leptin, and ghrelin and corresponding gene expressions modulate perturbed inhibitory and reward systems, encouraging eating disorders [66,67]. These studies collectively suggest neural substrates of executive-control and reward-related brain regions that influence the behavioral traits of eating.
Our work adopted eating disorder scores at the question-level rather than the factor-wise scores. We hypothesized that using the full information provided in the question-wise responses (51 features) would be better suited for exploring the link between the brain and eating behaviors than using factor-wise scores (three features) due to the increased information content. Indeed, our previous study investigated the association between functional brain networks and eating behaviors in people with obesity using factor-wise scores and found the association with an effect size (i.e., Pearson’s r) of 0.2422 [6], which was largely improved in our current study (Pearson’s r = 0.4945). These results indicate that our approach could better select the combination of functional brain connectivity to be maximally linked to behavioral traits of eating disorders. Thus, the regions identified in our study might have higher levels of statistical confidence, although the identified regions were similar to those identified in previous studies. Interestingly, we found that the questions related to a similar behavioral trait, especially dietary restraint, showed clustered association with the brain, confirming previous studies that linked cognition- and reward-related brain networks to eating behaviors [3,6–9]. Together, our work supports that the TFEQ at question-level could be a reliable measurement for assessing eating disorders in conjunction with brain function.
Although we found significant associations between the identified brain regions and dietary restraint scores, no significant association was detected for disinhibition or hunger factors. This should be investigated further, but this inability might be due to the demographic characteristics of enrolled participants, in which obesity-related diseases were excluded. Further validations are required by recruiting participants with aberrant eating disorders.
Our study has several limitations. First, we only used betweenness centrality for quantifying complex brain networks. Other graph-related measures such as degree- and eigenvector-centrality, and efficiency might be used to associate the brain and eating behaviors. This is left to future work. Second, obesity is a multifactorial disorder affected by a wide range of factors, including eating behaviors, genetic factors, and other environmental factors [68,69]. Strict control of these factors was difficult since we obtained our data from an open database. Third, a longitudinal study that follows the changes in eating behaviors is needed to assess the stability of our findings. Fourth, patient demographic and eating disorder questionnaires were not exactly matched between discovery and validation sets (p-value < 0.05). Despite these potential confounds, we found largely consistent results between discovery and validations sets. However, a controlled validation set is needed to fully test the generalizability of our approach.
Here, we introduced a framework for associating functional brain connectivity and eating disorders at question-level. Harnessing a multivariate association analysis, SCCA, we identified brain regions in executive control and reward system are functionally associated with restraint eating in people with a wide range of BMI. Association between the functional connectivity and WHR suggested the potential role of functional variations in inhibitory control associated with obesity. Our work provides a new perspective on understanding the relationship between eating disorder and brain function.
Supporting information
S1 File. The list of questions from the TFEQ and the results of SCCA.
The distribution of obesity-related clinical scores in the discovery set. The list of 51 questions of the TFEQ. The list of selected features as a result of SCCA for the discovery and validation set. The most frequently selected features from bootstrap.
https://doi.org/10.1371/journal.pone.0237511.s001
(DOCX)
References
- 1. Gerlach G, Herpertz S, Loeber S. Personality traits and obesity: A systematic review. Obes Rev. 2015;16: 32–63. pmid:25470329
- 2. Lee HA, Lee WK, Kong K-A, Chang N, Ha E-H, Hong YS, et al. The effect of eating behavior on being overweight or obese during preadolescence. J Prev Med Public Heal. 2011;44: 226–233. pmid:22020188
- 3. Val-Laillet D, Aarts E, Weber B, Ferrari M, Quaresima V, Stoeckel LE, et al. Neuroimaging and neuromodulation approaches to study eating behavior and prevent and treat eating disorders and obesity. NeuroImage Clin. 2015;8: 1–31. pmid:26110109
- 4. Malik VS, Willett WC, Hu FB. Global obesity: Trends, risk factors and policy implications. Nat Rev Endocrinol. 2013;9: 13–27. pmid:23165161
- 5. Raji CA, Ho AJ, Parikshak NN, Becker JT, Lopez OL, Kuller LH, et al. Brain structure and obesity. Hum Brain Mapp. 2010;31: 353–364. pmid:19662657
- 6. Park B, Seo J, Park H. Functional brain networks associated with eating behaviors in obesity. Sci Rep. 2016;6: 1–8. pmid:28442746
- 7. Park B, Lee MJ, Kim M, Kim S-H, Park H. Structural and functional brain connectivity changes between people with abdominal and non-abdominal obesity and their association with behaviors of eating disorders. Front Neurosci. 2018;12: 1–13. pmid:29403346
- 8. Park B, Moon T, Park H. Dynamic functional connectivity analysis reveals improved association between brain networks and eating behaviors compared to static analysis. Behav Brain Res. 2018;337: 114–121. pmid:28986105
- 9. Siep N, Roefs A, Roebroeck A, Havermans R, Bonte M, Jansen A. Fighting food temptations: The modulating effects of short-term cognitive reappraisal, suppression and up-regulation on mesocorticolimbic activity related to appetitive motivation. Neuroimage. 2012;60: 213–220. pmid:22230946
- 10. Hollmann M, Hellrung L, Pleger B, Schlögl H, Kabisch S, Stumvoll M, et al. Neural correlates of the volitional regulation of the desire for food. Int J Obes. 2012;36: 648–655. pmid:21712804
- 11. Lips MA, Wijngaarden MA, Grond J van der, Buchem MA van, Groot GH de, Rombouts SA, et al. Resting-state functional connectivity of brain regions involved in cognitive control, motivation, and reward is enhanced in obese females. Am J Clin Nutr. 2014;100: 524–531. pmid:24965310
- 12. Stunkard AJ, Messick S. The Three-Factor Eating Questionnaire to measure dietary restraint, disinhibition and hunger. J Psychosom Res. 1985;29: 71–83. pmid:3981480
- 13. Lauzon B De, Romon M, Lafay L, Borys J, Karlsson J, Ducimetie P, et al. The Three-Factor Eating Questionnaire-R18 is able to distinguish among different eating patterns in a general population. J Nutr. 2004; 2372–2380. pmid:15333731
- 14. Fairburn CG, Beglin SJ. Assessment of eating disorders: Interview or self-report questionnaire? Int J Eat Disord. 1994;16: 363–370. pmid:7866415
- 15. Mond JM, Hay PJ, Rodgers B, Owen C, Beumont PJV. Validity of the Eating Disorder Examination Questionnaire (EDE-Q) in screening for eating disorders in community samples. Behav Res Ther. 2004;42: 551–567. pmid:15033501
- 16. Lahmiri S, Dawson DA, Shmuel A. Performance of machine learning methods in diagnosing Parkinson ‘ s disease based on dysphonia measures. Biomed Eng Lett. 2018;8: 29–39. pmid:30603188
- 17. Iqbal S, Ghani MU, Tanzila K. Computer-assisted brain tumor type discrimination using magnetic resonance imaging features. Biomed Eng Lett. 2018;8: 5–28. pmid:30603187
- 18. Hotelling H. Relations between two sets of variates. Biometrika. 1936;28: 321–377.
- 19. Waaijenborg S, Verselewel de Witt Hamerr PC, Zwinderman AH. Quantifying the association between gene expressions and DNA-markers by penalized canonical correlation analysis. Stat Appl Genet Mol Biol. 2008;7. pmid:18241193
- 20. Witten DM, Tibshirani RJ. Extensions of sparse canonical correlation analysis with applications to genomic data. Stat Appl Genet Mol Biol. 2009;8. pmid:19572827
- 21. Kim M, Won JH, Youn J, Park H. Joint-Connectivity-Based Sparse Canonical Correlation Analysis of Imaging Genetics for Detecting Biomarkers of Parkinson’s Disease. IEEE Trans Med Imaging. 2020;39: 23–34. pmid:31144631
- 22. Witten DM, Tibshirani R, Hastie T. A penalized matrix decomposition, with applications to sparse principal components and canonical correlation analysis. Biostatistics. 2009;10: 515–534. pmid:19377034
- 23. Parkhomenko E, Tritchler D, Beyene J. Sparse canonical correlation analysis with application to genomic data integration. Stat Appl Genet Mol Biol. 2009;8. pmid:19222376
- 24. Hardoon DR, Shawe-Taylor J. Sparse canonical correlation analysis. 2011;83: 331–353.
- 25. Lin D, Calhoun VD, Wang YP. Correspondence between fMRI and SNP data by group sparse canonical correlation analysis. Med Image Anal. 2014;18: 891–902. pmid:24247004
- 26. Fang J, Lin D, Schulz SC, Xu Z, Calhoun VD, Wang YP. Joint sparse canonical correlation analysis for detecting differential imaging genetics modules. Bioinformatics. 2016;32: 3480–3488. pmid:27466625
- 27. Du L, Huang H, Yan J, Kim S, Risacher SL, Inlow M, et al. Structured sparse canonical correlation analysis for brain imaging genetics: An improved GraphNet method. Bioinformatics. 2016;32: 1544–1551. pmid:26801960
- 28. Hao X, Li C, Du L, Yao X, Yan J, Risacher SL, et al. Mining outcome-relevant brain imaging genetic associations via three-way sparse canonical correlation analysis in Alzheimer’s disease. Sci Rep. 2017;7: 1–12. pmid:28127051
- 29. Mai Q, Zhang X. An iterative penalized least squares approach to sparse. Biometrics. 2019;75: 734–744. pmid:30714093
- 30. Chen X, Liu H, Carbonell JG. Structured sparse canonical correlation analysis. J Mach Learn Res. 2012;22: 199–207.
- 31. Bullmore ET, Sporns O. Complex brain networks: Graph theoretical analysis of structural and functional systems. Nat Rev Neurosci. 2009;10: 186–198. pmid:19190637
- 32. Rubinov M, Sporns O. Complex network measures of brain connectivity: Uses and interpretations. Neuroimage. 2010;52: 1059–1069. pmid:19819337
- 33. Nooner KB, Colcombe SJ, Tobe RH, Mennes M, Benedict MM, Moreno AL, et al. The NKI-Rockland sample: A model for accelerating the pace of discovery science in psychiatry. Front Neurosci. 2012;6: 152. pmid:23087608
- 34. Garner D et al. The Eating Attitudes Test: psychometric features and clinical correlates. Psychol Med. 1982;12: 871–878. pmid:6961471
- 35. Orbitello B, Ciano R, Corsaro M, Rocco PL, Taboga C, Tonutti L, et al. The EAT-26 as screening instrument for clinical nutrition unit attenders. Int J Obes. 2006;30: 977–981. pmid:16432540
- 36. Siervo M, Boschi V, Papa A, Bellini O, Falconi C. Application of the SCOFF, Eating Attitude Test 26 (EAT 26) and Eating Inventory (TFEQ) questionnaires in young women seeking diet-therapy. Eat Weight Disord. 2005;10: 76–82. pmid:16114220
- 37. Lee S, Kwok K, Liau C, Leung T. Screening Chinese patients with eating disorders using the Eating Attitudes Test in Hong Kong. Int J Eat Disord. 2002;32: 91–97. pmid:12183950
- 38. Dotti A, Lazzari R. Validation and reliability of the Italian EAT-26. Eat Weight Disord. 1998;3: 188–194. pmid:10728170
- 39. Park B, Byeon K, Park H. FuNP (Fusion of Neuroimaging Preprocessing) pipelines: A fully automated preprocessing software for functional magnetic resonance imaging. Front Neuroinform. 2019;13: 1–14. pmid:30792636
- 40. Cox RW. AFNI: Software for analysis and visualization of functional magnetic resonance neuroimages. Comput Biomed Res. 1996;29: 162–173. pmid:8812068
- 41. Jenkinson M, Beckmann CF, Behrens TEJ, Woolrich MW, Smith SM. FSL. Neuroimage. 2012;62: 782–790. pmid:21979382
- 42. Avants BB, Tustison NJ, Song G, Cook PA, Klein A, Gee JC. A reproducible evaluation of ANTs similarity metric performance in brain image registration. Neuroimage. 2011;54: 2033–2044. pmid:20851191
- 43. Salimi-Khorshidi G, Douaud G, Beckmann CF, Glasser MF, Griffanti L, Smith SM. Automatic denoising of functional MRI data: Combining independent component analysis and hierarchical fusion of classifiers. Neuroimage. 2014;90: 449–468. pmid:24389422
- 44. Fan L, Li H, Zhuo J, Zhang Y, Wang J, Chen L, et al. The human brainnetome atlas: A new brain atlas based on connectional architecture. Cereb Cortex. 2016;26: 3508–3526. pmid:27230218
- 45. Lee MJ, Park BY, Cho S, Park H, Kim ST, Chung CS. Dynamic functional connectivity of the migraine brain: A resting-state functional magnetic resonance imaging study. Pain. 2019;160: 2776–2786. pmid:31408050
- 46. Ruiz Vargas E, Mitchell DGV, Greening SG, Wahl LM. Topology of whole-brain functional MRI networks: Improving the truncated scale-free model. Phys A Stat Mech its Appl. 2014;405: 151–158.
- 47. Onoda K, Yamaguchi S. Small-worldness and modularity of the resting-state functional brain network decrease with aging. Neurosci Lett. 2013;556: 104–108. pmid:24157850
- 48. Mumford JA, Horvath S, Oldham MC, Langfelder P, Geschwind DH, Poldrack RA. Detecting network modules in fMRI time series: A weighted network analysis approach. Neuroimage. 2010;52: 1465–1476. pmid:20553896
- 49. Schwarz AJ, McGonigle J. Negative edges and soft thresholding in complex network analysis of resting state functional connectivity data. Neuroimage. 2011;55: 1132–1146. pmid:21194570
- 50. Boutte D, Liu J. Sparse canonical correlation analysis applied to fMRI and genetic data fusion. 2010 IEEE Int Conf Bioinforma Biomed. 2010;1: 422–426. pmid:30713779
- 51. Brooks SJ, Cedernaes J, Schiöth HB. Increased prefrontal and parahippocampal activation with reduced dorsolateral prefrontal and insular cortex activation to food images in obesity: A meta-analysis of fMRI studies. PLoS One. 2013;8: 4. pmid:23593210
- 52. Olivo G, Wiemerslage L, Swenne I, Zhukowsky C, Salonen-Ros H, Larsson EM, et al. Limbic-thalamo-cortical projections and reward-related circuitry integrity affects eating behavior: A longitudinal DTI study in adolescents with restrictive eating disorders. PLoS One. 2017;12: 3. pmid:28248991
- 53. Vainik U, Dagher A, Dubé L, Fellows LK. Neurobehavioural correlates of body mass index and eating behaviours in adults: A systematic review. Neurosci Biobehav Rev. 2013;37: 279–299. pmid:23261403
- 54. Volkow ND, Wang G-J, Telang F, Fowler JS, Thanos PK, Logan J, et al. Low dopamine striatal D2 receptors are associated with prefrontal metabolism in obese subjects: Possible contributing factors. Neuroimage. 2008;42: 1537–1543. pmid:18598772
- 55. Tataranni PA, DelParigi A. Functional neuroimaging: A new generation of human brain studies in obesity research. Obes Rev. 2003;4: 229–238. pmid:14649373
- 56. Tataranni PA, Gautier J-F, Chen K, Uecker A, Bandy D, Salbe AD, et al. Neuroanatomical correlates of hunger and satiation in humans using positron emission tomography. Proc Natl Acad Sci. 1999;96: 4569–4574. pmid:10200303
- 57. Goldstone AP, Prechtl de Hernandez CG, Beaver JD, Muhammed K, Croese C, Bell G, et al. Fasting biases brain reward systems towards high-calorie foods. Eur J Neurosci. 2009;30: 1625–1635. pmid:19811532
- 58. Holland PC, Gallagher M. Amygdala-frontal interactions and reward expectancy. Curr Opin Neurobiol. 2004;14: 148–155. pmid:15082318
- 59. Taste Rolls E., olfactory and food texture reward processing in the brain and obesity. Int J Obes. 2011;35: 550–561. pmid:20680018
- 60. O’Doherty JP, Deichmann R, Critchley HD, Dolan RJ. Neural responses during anticipation of a primary taste reward. Neuron. 2002;33: 815–826. pmid:11879657
- 61. Kim SH, Park BY, Byeon K, Park H, Kim Y, Eun YM, et al. The effects of high-frequency repetitive transcranial magnetic stimulation on resting-state functional connectivity in obese adults. Diabetes, Obes Metab. 2019;21: 1956–1966. pmid:31050167
- 62. Moreno-Lopez L, Contreras-Rodriguez O, Soriano-Mas C, Stamatakis EA, Verdejo-Garcia A. Disrupted functional connectivity in adolescent obesity. NeuroImage Clin. 2016;12: 262–268. pmid:27504261
- 63. Ziauddeen H, Alonso-Alonso M, Hill JO, Kelley M, Khan NA. Obesity and the neurocognitive basis of food reward and the control of intake. Adv Nutr. 2015;6: 474–486. pmid:26178031
- 64. Vainik U, Baker TE, Dadar M, Zeighami Y, Michaud A, Zhang Y, et al. Neurobehavioral correlates of obesity are largely heritable. Proc Natl Acad Sci. 2018;115: 9312–9317. pmid:30154161
- 65. Martin LE, Holsen LM, Chambers RJ, Bruce AS, Brooks WM, Zarcone JR, et al. Neural mechanisms associated with food motivation in obese and healthy weight adults. Obesity. 2010;18: 254–260. pmid:19629052
- 66. Locke AE, Kahali B, Berndt SI, Justice AE, Pers TH, Day FR, et al. Genetic studies of body mass index yield new insights for obesity biology. Nature. 2015;518: 197–206. pmid:25673413
- 67. Steward T, Miranda-Olivos R, Soriano-Mas C, Fernández-Aranda F. Neuroendocrinological mechanisms underlying impulsive and compulsive behaviors in obesity: a narrative review of fMRI studies. Rev Endocr Metab Disord. 2019;20: 263–272. pmid:31654260
- 68. Brownell KD, Wadden TA. The heterogeneity of obesity: Fitting treatments to individuals. Behav Ther. 1991;22: 153–177.
- 69. McLaughlin T. Metabolic heterogeneity of obesity: Role of adipose tissue. Int J Obes Suppl. 2012;2: S8–S10. pmid:25089194