{"id":19252270,"url":"https://github.com/sccn/groupsift","last_synced_at":"2025-04-21T13:30:49.652Z","repository":{"id":54180665,"uuid":"327717657","full_name":"sccn/groupSIFT","owner":"sccn","description":null,"archived":false,"fork":false,"pushed_at":"2024-08-02T19:40:59.000Z","size":2716,"stargazers_count":11,"open_issues_count":1,"forks_count":4,"subscribers_count":3,"default_branch":"master","last_synced_at":"2025-04-01T13:38:03.435Z","etag":null,"topics":[],"latest_commit_sha":null,"homepage":null,"language":"MATLAB","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":null,"status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/sccn.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":null,"code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null}},"created_at":"2021-01-07T20:27:27.000Z","updated_at":"2025-02-08T21:13:13.000Z","dependencies_parsed_at":"2022-08-13T08:31:11.104Z","dependency_job_id":"9d93d92e-a09d-4f3c-baeb-dfea248ec970","html_url":"https://github.com/sccn/groupSIFT","commit_stats":null,"previous_names":[],"tags_count":0,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/sccn%2FgroupSIFT","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/sccn%2FgroupSIFT/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/sccn%2FgroupSIFT/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/sccn%2FgroupSIFT/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/sccn","download_url":"https://codeload.github.com/sccn/groupSIFT/tar.gz/refs/heads/master","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":250064541,"owners_count":21368922,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":[],"created_at":"2024-11-09T18:26:11.733Z","updated_at":"2025-04-21T13:30:46.138Z","avatar_url":"https://github.com/sccn.png","language":"MATLAB","funding_links":[],"categories":[],"sub_categories":[],"readme":"[![IMAGE ALT TEXT HERE](http://img.youtube.com/vi/jHngHEIsg7Q/0.jpg)](http://www.youtube.com/watch?v=jHngHEIsg7Q)\nClick to play the movie on Youtube. This movie shows the first application of *groupSIFT* in Loo et al. (2019) *NeuroImage: Clinical*\n\n# The GROUPSIFT EEGLAB Plugin\n\nThis page is for those who wants to cooperate with me to test groupSIFT\ntoolbox and EEGLAB plugin. Unfortunately, at this point we cannot provide much user\nsupport. We know this page does not cover full detail either (including\nhow to determine SIFT parameters). So this plugin is for advanced user\nwho have good experience with Matlab and SIFT. When you use it, please\ndo so at your own risk!\n\nIf you are interested in learning basic SIFT functions using its own built-in simulator, see [this page](https://sccn.ucsd.edu/wiki/How_to_run_SIFT_simulation)\n\n# Environment\n\n-   I originally developed it with Matlab R2013a runninig on Fedora 22\n    (64bit) with dual monitors with 1600 x 1200 resolution. In a recent\n    project, I updated it for Matlab 2017b, EEGLAB 14.1.2, and SIFT 1.52\n    (with customized movie function). I have not fully investigated\n    dependency on different environments other than Linux. Matlab Image\n    Processing Toolbox is necessary to use bwlabel().\n\n# Required preprocessing\n\n-   You need EEGLAB .set files that are processed with ICA and\n    subsequent IC selection: ALL the ICs in your data will go to SIFT\n    preprocess. You don't want to use 100 ICs that will create 100 x 100\n    x frequency x time tensor for n subjects in the end! There is also\n    datapont-to-parameter ratio you want to consider. Also, DIPFIT must\n    be done. Use ICLabel() plugin to select ICs with 'brain' label\n    probability \\\u003e 0.7, for example.\n-   Downsample the data to nearly 100-120 Hz using the additional two\n    options in this way: pop_resample(EEG, 100, 0.8, 0.4) The extra\n    optional parameters are to use mild low-pass filter slope in\n    anti-aliasing to suppress AR model order. See [this\n    page](https://sccn.ucsd.edu/wiki/Firfilt_FAQ#Q._For_Granger_Causality_analysis.2C_what_filter_should_be_used.3F_.2804.2F26.2F2018_Updated.29)\n    for detail.\n-   **Separate conditions so that one .set file, one condition.** All\n    the SIFT-related additional preprocessing should be applied to each\n    of subdivided single-condition data sets separately.\n-   Note that **groupSIFT is independent of EEGLAB STUDY**.\n-   Follow this rule in using the groupSIFT all the time: Locate the\n    preprocessed .set files into a dedicated folder which does not have\n    anything other than the processed files. For example, if you have\n    condition A, condition B, as well as you plan to test A-B, create\n    folders for A, B, and A-B separately.\n-   Also, it is recommended to move to the working folder every time you\n    step forward. groupSIFT interactive file loader in GUI finds the\n    local files with the extension filter.\n-   Do not use '_' in the file name. This character needs to be\n    preserved to identify prefix words.\n\n# groupSIFT GUI menu explained\n\n## 1.Run SIFT Batch\n\nClick this item to perform SIFT on the selected multiple .set files by\nbatch mode. Again, make sure that this is applied for all the\nindividual, condition-separated .set files. If you have two conditions,\nyou have to go through this processes separately for each condition.\nrPDC and dDTF08 are computed because these are the ones you can justify\nthe use in papers. rPDC is theoretically better, but it has known\nbug/problem in the highest frequency result. dDTF08 is widely used.\nrPDC's result has broader spreading in the time-frequency domain, while\ndDTF's result has a sharper resolution in the time-domain. Let GUI\nwindows stay while a process goes on. Closing it in the middle will\ncrash the process.\n\n### SIFT tips (08/15/2020 added)\n\nI found that the equation used in calculating datapoint-to-parameter\nratio is incorrect. According to Schlögl and Supp (2006) as well as\nKorzeniewska et al. (2008), the equation must be\n(num_IC\\*model_order)/window_length\\*num_trials. This change affects how\nyou determine parameters, in a 'better' way I would say as shown below:\n1) more ICs can be included; 2) less number of trials can be tolerated;\n3) shorter sliding window can be used. This change will particularly\nimpacts continuous data analysis, as the current equation would probably\nallow sliding window length of a few minutes! In fact, this\ndatapoint-to-parameter ratio has been a limiting factor for me to apply\nSIFT with confidence.\n\nTo obtain the corrected datapoint-to-parameter ratio based on the\nabove-suggested reason, make the following change on\n\u003cspan style=\"color:#FF0000\"\u003e **est_checkMVARParams() line 85** \u003c/span\u003e\n\n``` matlab\n%winrat = ((M^2)*p)/(wlen*T);\nwinrat = ((M)*p)/(wlen*T);\n```\n\nThat being said, when I discussed this issue with the author, he also\ntold me that if we make a correction, the estimate would be overly lax\nand would even become useless. I kind of see the point from my\nexperience. Somehow, most of SIFT validation measures are always either\ntoo lax (the stability test) or too stringent (the whiteness tests),\nwhich are generally hard to follow. In conclusion, I would recommend the\nabove fix to enjoy more degrees of freedom in the analysis design, while\ntrying to stay as conservative (i.e., lower the number, more\nconservative!) as possible.\n\nNote also that **renormalized partial directed coherence (rPDC) always\nhas noise near the highest frequency**. I confirmed it with the original\nauthor of SIFT, Dr. Tim Mullen. It is advised that one always excludes\nnear-highest freq results in rPDC.\n\n## 2.Varidate AR models\n\nCheck the summary plots and consider to remove outlier subjects\n(horizontal axis represents set file indices). Check the group-mean\nvalue (the rightmost bar) of the datapoint-to-parameter ratio. to\nguarantee validity of the AR modeling stage. By the way, the definition\nof datapoint-to-parameter ratio is calculated differently from the\noriginal paper, so be careful. For detail, see [this\npage](https://sccn.ucsd.edu/wiki/Makoto's_preprocessing_pipeline#SIFT_tips_.2808.2F06.2F2019_updated.29).\n\n## 3.Convert to group anatomical ROIs\n\nFWHM determines the smoothing width for the dipole density. Typically,\nfMRI==8mm, PET==20mm. You also preselect the minimum number of subjects\nso that hereafter you focus on anatomical regions with majority of\nsubjects (e.g., 80%) contributes non-zero dipole density. Press the\n'Compute upper bound for estimation' button to confirm the result. If\nyou are satisfied with the estimation, enter the file prefix name\n(again, DO NOT use '_') and press the button in the bottom 'Select ALL\n.set files and START'.\n\n## 4.Compute t-scores \u0026 p-values\n\nSpecify the folder that has the precomputed results from the process\nabove. To test the difference A-B, specify the folders for A and B, and\nfurther specify the new folder to save the A-B results. Uncorrected\np-value here determines the size of the pixel clusters in the later\ntime-frequency plots. The number of iteration should not be less than\n2000 because the surrogate distribution is used for nonparametric tests\nwhich requires good tail support. For multiple comparison correction,\nweak family-wise error rate correction (FWER) is applied to perform\ncluster-level correction. The *mass of cluster*, sum of t-scores within\neach pixel cluster, is pooled from ALL the edges to determine the\nomnibus correction criterion.\n\n## 5.Show pre-selected ROIs\n\nThis plot shows the preselected pairwise dipole density (i.e.,\nunweighted graph edges). *The connectivity measure is this pairwise\ndipole density weighted by rPDC or dDTF*.\n\n## 6.View results \u0026 Export for movie (08/20/2020 updated)\n\nSpecify the \\*_tStatistics.mat and \\*_dipolePairDensity.mat files.\nEnter all the parameters. The 'Cluster-level correction' here determines\nthe threshold on *mass of cluster*. After pressing 'Plot connectivity\nmatrix', click a graph edge in the main connectivity matrix plot on the\nleft.\n\n\\-**MCC for graph edges** (checkbox)--MCC stands for multiple comparison\ncorrection. When checked, surrogate distributions of extreme values\n(minimum and maximum statistics) will be built using 10,000 values\n(default) of min/max value across ALL graph edges submitted (including\nthe ones with no significant results). Then, for the case of p \\\u003c 0.05\n(default), the 2.5- and 97.5-percentile values of the surrogate\ndistribution are obtained with which one can perform omnibus correction\nfor time-frequency values of any graph edge at the same time.\n\nIf unchecked, surrogate distributions will be built using 10,000\n(default) x \\[number_of_edges\\] values across all graph edges submitted.\nThis surrogate statistics distribution is NOT made of the max/min values\nacross all graph edges, so the 2.5- and 97.5-percentile values of the\nsurrogate distribution cannot address multiple comparison across graph\nedges, but it still does for any single graph edge. Thus, unchecking\nthis options should be used with caution, for example for\nhypothesis-driven ROI analysis.\n\n\\-**Use GFWER (u=1)**(checkbox)--GFWER stands for *generalized* (weak)\nfamily-wise error rate. 'Weak' means using cluster-level correction.\n\nThe trade off you make here is that you gain more detection power at the\ncost of the fact that you accept maximum 1 (hence u=1) false positive\nresult (in your case, one cluster of pixels) present in your result.\nThis may sound unusual and even scary, but remember that you are already\nalways accepting 5% of false positive results which is usually WAY\nlarger than 1. How it works is as follows. GFWER does not pick up the\nmax/min statistics from each iteration of permutation trial, but the\n*second to max/min* (only one next to the max/min, hence u=1). This\napproach is to gain statistical sensitivity at the cost of known number\nof false positive results (here, u=1 i.e., one *mass of cluster* in your\ndata is known to be a result of false positive). You may wonder if this\nis meaningful thing to do. It is, because a distributions of surrogate\nstatistics tend to have outliers in tails. Removing the leftmost and\nrightmost values from the tails can in most cases greatly ease the\nextreme value statistics. You can try to find out how effective this\ntrade could be, as it is calculated altogether anyway. By the way, this\noption is only usable when 'MCC for graph edges' option is checked.\n\nBelow, three statistical results are shown from the same data. From\nleft, graph-edge MCC on, graph-edge MCC on with GFWER, and graph-edge\nMCC off. Note the change of the number of significant edges. Data are\nfrom Loo et al. (2019).\n\n![Mccon_currentdefault2.png](images/Mccon_currentdefault2.png)\n![Mccon_gfwer2.png](images/Mccon_gfwer2.png)\n![Nomcc_previousdefault2.png](images/Nomcc_previousdefault2.png)\n\n# How to generate a group-level connectivity movie\n\n-   The button 'Save the data for movie' is located at the bottom right\n    corner of the visualization GUI.\n-   All the parameters determined in **View results \u0026 Export for movie**\n    will take effect on the movie data.\n-   Provide any one of the .set files used for the analysis. The copy of\n    the specified .set file serves as a 'donor', and its\n    EEG.CAT.Conn.RPDC/dDTF08 is replaced with the group-mean values\n    after thresholding.\n-   From EEGLAB main window, Tools -\\\u003e SIFT -\\\u003e Visualiation -\\\u003e\n    BrainMovie3D. I use my customized movie function for this. This\n    opens the propertyGrid interface, which is known for multiple uses\n    and dependencies. Care must be taken to set Matlab path before using\n    this function. Use 'which -all propertyGrid' to ensure you use the\n    right one, otherwise it won't work.\n-   \"FrequenciesToCollapse\" may need to be adjusted so that in my case\n    instead of 2:50 I need to set 2:49.9 to make it work.\n-   Do not subtract baseline. It is taken care of by groupSIFT.\n    Otherwise, zero values (i.e., masked by statistical results) will be\n    non-zeros.\n-   \"FooterPanelDisplaySpec\", \"GraphMetric\". This will show you envelope\n    time-series.\n-   \"InitialView\" \\[50 36\\] is the default value. For axial slice, \\[0\n    90\\]; for sagittal slice, \\[90 0\\]; for coronal slice, \\[0 0\\].\n-   \"Theme\", \"darkdream\" (optional)\n-   \"ImageOutputDirectory\", \"prompt\" (you need to type it) By the way,\n    \"Save all picture frames\" currently does not work, but you still\n    need to enter 'prompt' twice for picture and movie; one for the path\n    and the other for the file name.\n-   \"MovieOutputFinename\", \"prompt\" (you need to type it)\n\n# How to generate a group-level connectivity movie (12/23/2019 update)\n\nOn the 'pop_viewResultsAndExportForMovie' GUI, there is a 'Output\nindividual data' button on the bottom right. When you click it,\nuigetdir() GUI pops up and asks you to select '_allSubjStack.mat' file.\nAfter selecting this file, it generates singificant blob-by-blob mean\nvalue for individual subjects with which one can perform correlation\nanalysis. The output is saved as 'dataSheet' on Matlab 'base' workspace,\nwhich you can export as csv file. For the case of A-B, you should\nperform for A and B separately (both A and B have the same graph edges).\n\n# How to output individual subject data in the case of subtraction (02/26/2020 Updated)\n\nCurrently, this is not supported by GUI. But with fairly simple command\nline operation, you can do it IF THE TWO CONDITIONS ARE WITHIN-SUBJECT.\n\n1.  Empty your workspace.\n2.  Load XXXX_allSubjStack.mat for Condition A.\n3.  Rename 'allConnectivityStack' to something else (here, 'tmp1')\n4.  Load YYYY_allSubjStack.mat for Condition B. Note that if A and B are\n    within-subject condition (i.e., same ICA results), they should have\n    the same data except for 'allConnectivityStack'.\n5.  Rename 'allConnectivityStack' to something else (here, 'tmp2')\n6.  Perform allConnectivityStack = tmp1-tmp2;\n7.  Save everything in the workspace with a new name\n    ZZZZ_allSubjStack.mat.\n8.  Feed ZZZZ_allSubjStack.mat during GUI operation.\n\nThe same method can be used to take subtraction between within-subject\nconditions. For example, if you have mismatch negativity data for two\ngroup of subjects, you want to subtract Deviant-Standard for each group\nusing the method explained above, then perform group-level analysis.\nNote that in this case, you have to obtain individual data list for each\ngroup separately.\n\n# Layout issue (06/21/2018 update)\n\nFor unknown reason, GUI layout can be collapsed in your environment.\nSince I can't replicate the issue in my environment, for the time being\nI would like the users to fix it themselves following these steps.\n\n1.  Type 'guide' in Matlab command line.\n2.  Top tab 'Open existing GUI' and 'Browse' button to specify the\n    groupSIFT GUI in question under your eeglab plugin folder.\n3.  'Open' button to open the GUI in question.\n4.  Manually fix the layout and save. I even heard that one of the\n    windows were hidden by another window... so if you don't see what\n    you are looking for, do not forget to check the background of the\n    things on surface.\n\n# Link to the latest workshop material\n\n[EEGLAB workshop 2017 in\nTokyo](https://sccn.ucsd.edu/mediawiki/images/7/7c/GroupSIFT.pdf) [Link\nto a movie\nexample](https://sccn.ucsd.edu/mediawiki/images/6/6e/GroupSIFT_sagital.zip)\n\n# About the custom anatomical labels\n\ngroupSIFT uses anatomical labels defined in Automated Anatomical\nLabeling solution (Tzourio-Mazoyar et al., 2002). However, instead of\nthe original 88 regions, I reduced it to 76 regions by integrating 16\nsmall regions in limbic and basal regions into umbrella ROIs 'Upper\nBasal' and 'Lower Basal'. In\npop_groupSIFT_convertToGroupAnatomicalRois.m line 307-402, there is the\nfollowing description.\n\n``` matlab\n% These regions are to be included\n%     'Precentral_L'\n%     'Precentral_R'\n%     'Frontal_Sup_L'\n%     'Frontal_Sup_R'\n%     'Frontal_Sup_Orb_L'\n%     'Frontal_Sup_Orb_R'\n%     'Frontal_Mid_L'\n%     'Frontal_Mid_R'\n%     'Frontal_Mid_Orb_L'\n%     'Frontal_Mid_Orb_R'\n%     'Frontal_Inf_Oper_L'\n%     'Frontal_Inf_Oper_R'\n%     'Frontal_Inf_Tri_L'\n%     'Frontal_Inf_Tri_R'\n%     'Frontal_Inf_Orb_L'\n%     'Frontal_Inf_Orb_R'\n%     'Rolandic_Oper_L'\n%     'Rolandic_Oper_R'\n%     'Supp_Motor_Area_L'\n%     'Supp_Motor_Area_R'\n%     'Frontal_Sup_Medial_L'\n%     'Frontal_Sup_Medial_R'\n%     'Frontal_Med_Orb_L'\n%     'Frontal_Med_Orb_R'\n%     'Rectus_L'\n%     'Rectus_R'\n%     'Insula_L'\n%     'Insula_R'\n%     'Cingulum_Ant_L'\n%     'Cingulum_Ant_R'\n%     'Cingulum_Mid_L'\n%     'Cingulum_Mid_R'\n%     'Cingulum_Post_L'\n%     'Cingulum_Post_R'\n%     'Calcarine_L'\n%     'Calcarine_R'\n%     'Cuneus_L'\n%     'Cuneus_R'\n%     'Lingual_L'\n%     'Lingual_R'\n%     'Occipital_Sup_L'\n%     'Occipital_Sup_R'\n%     'Occipital_Mid_L'\n%     'Occipital_Mid_R'\n%     'Occipital_Inf_L'\n%     'Occipital_Inf_R'\n%     'Fusiform_L'\n%     'Fusiform_R'\n%     'Postcentral_L'\n%     'Postcentral_R'\n%     'Parietal_Sup_L'\n%     'Parietal_Sup_R'\n%     'Parietal_Inf_L'\n%     'Parietal_Inf_R'\n%     'SupraMarginal_L'\n%     'SupraMarginal_R'\n%     'Angular_L'\n%     'Angular_R'\n%     'Precuneus_L'\n%     'Precuneus_R'\n%     'Paracentral_Lobule_L'\n%     'Paracentral_Lobule_R'\n%     'Temporal_Sup_L'\n%     'Temporal_Sup_R'\n%     'Temporal_Pole_Sup_L'\n%     'Temporal_Pole_Sup_R'\n%     'Temporal_Mid_L'\n%     'Temporal_Mid_R'\n%     'Temporal_Pole_Mid_L'\n%     'Temporal_Pole_Mid_R'\n%     'Temporal_Inf_L'\n%     'Temporal_Inf_R'\n%\n% These regions are to be combined\n%     'Hippocampus_L'\n%     'Hippocampus_R'\n%     'ParaHippocampal_L'\n%     'ParaHippocampal_R'\n%     'Amygdala_L'\n%     'Amygdala_R'\n%                --\u003e Lower Basal\n%\n%     'Olfactory_L'\n%     'Olfactory_R'\n%     'Caudate_L'\n%     'Caudate_R'\n%     'Putamen_L'\n%     'Putamen_R'\n%     'Pallidum_L'\n%     'Pallidum_R'\n%     'Thalamus_L'\n%     'Thalamus_R'\n%               --\u003e Upper Basal\n%\n% One can visualize these regions by running visualizeAnatomicalRoiWithNHimasBlobs.m contained by the groupSIFT folder.\n```\n\nThe reason why I created the umbrella ROIs is because these limbic and\nbasal regions (even including ventricles) are unlikely to be generators\nof scalp-measurable EEG due to lack of critical conditions, namely a\nlarge area of pyramidal cells aligned in parallel. However, because of\nerrors in dipole fitting, about 20% of fitted dipoles goes into these\nphysiologically invalid deep retions (for detail, see [this\npage](https://sccn.ucsd.edu/wiki/Makoto%27s_preprocessing_pipeline#Physiologically_invalid_deep_dipoles.3F_.28Special_contents_for_130.2C000_hit.2C_07.2F02.2F2020_Update.29)).\nIf we know the label 'thalamus' is completely inappropriate to be used\nto refer to the estimated EEG sources, should we still provide specific\nlabels that are now only misleading? Instead, I suggest that we use\numbrella terms 'Upper Basal' and 'Lower Basal' just to indicate how deep\nthey are. The depth information in dipole fitting could be related to\nthe area information in the actual dipole sheet, so making a minimal\ndistinction between 'upper' and 'lower' may be helpful.\n\n# Published works\n\nThe dedicated technical paper is not prepared yet. But a couple of\nclinical researches using groupSIFT are already published. Loo\net al. (2019) has relatively detailed description of the method in\nSupplement (which needs some update).\n\n[Loo et al. (2019) Neural activation and connectivity during cued eye blinks in Chronic Tic Disorders. *NeuroImage: Clinical* 24:101956](https://www.sciencedirect.com/science/article/pii/S2213158219303067?via%3Dihub)\n\n[Koshiyama et al. (2020) Abnormal effective connectivity underlying auditory mismatch negativity impairments in schizophrenia. *Biological Psychiatry CNNI* 5:1028-1039.](https://www.sciencedirect.com/science/article/abs/pii/S245190222030135X)\n\n[Koshiyama et al. (2020) Neurophysiologic Characterization of Resting State Connectivity Abnormalities in Schizophrenia Patients. *Front Psychiatry* 11:608154.](https://www.frontiersin.org/articles/10.3389/fpsyt.2020.608154/full)\n\n[Koshiyama et al. (2020) Auditory-Based Cognitive Training Drives Short- and Long-Term Plasticity in Cortical Networks in Schizophrenia. *Schizophrenia Bulletin Open* 1:sgaa065.](https://academic.oup.com/schizbullopen/article/1/1/sgaa065/5998109)\n\n[Miyakoshi et al. (2021) The AudioMaze: An EEG and motion capture study of human spatial navigation in sparse augmented reality. *European Journal of NeuroscienceI* Online ahead of print.](https://onlinelibrary.wiley.com/doi/10.1111/ejn.15131)\n\n[Jurgiel et al. (2021) Inhibitory control in children with tic disorder: aberrant fronto-parietal network activity and connectivity. *Brain Communications* 3:fcab067.](https://academic.oup.com/braincomms/article/3/2/fcab067/6219298)\n\n[Jurgiel et al. (2023) Additive and Interactive Effects of Attention-Deficit/Hyperactivity Disorder and Tic Disorder on Brain Connectivity. *Biological Psychiatry: Cognitive Neuroscience and Neuroimaging* 8:1094-1102.](https://www.sciencedirect.com/science/article/pii/S245190222200249X)\n\n[Tseng et al. (2024) Neural Network Dynamics and Brain Oscillations Underlying Aberrant Inhibitory Control in Internet Addiction. *IEEE Transactions on Neural Systems and Rehabilitation Engineering* 32:946-955.](https://ieeexplore.ieee.org/abstract/document/10431676)\n\n# Download\nGroupSIFT is NOT in the EEGLAB plugin manager. You may install GroupSIFT by downloading the zipped file from [the GitHub repository](https://github.com/sccn/groupSIFT), 'Code' (the green button) -\u003e 'Download ZIP' (the menu item at the botton). Unzip the file and place the resulting folder in the EEGLAB plugin folder. The current version is 0.51.\n\n# Support\ngroupSIFT was developed for a project for a study on chronic tic disorder (PI Sandra Loo) that was supported by NINDS 80160 and 97484.\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fsccn%2Fgroupsift","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fsccn%2Fgroupsift","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fsccn%2Fgroupsift/lists"}