Brainframe

Chris Mezias & Justin Torok

MATLAB functionality for creating 3D human & mouse brain renderings, with many options. This is a tool for visualizing per-voxel or per-region data, as well as connectome data, in a 3D surface representation of a brain. This function has been tested for and works using mice and human brain atlases, but may well work with other species' brains. Further help including use case examples can be found in the script brainframe_Help.m.

To begin using the Brainframe package, please clone or download the entire repository into a folder accessible to MATLAB on your computer. Rearranging folder contents before getting comfortable with use could lead to package functionality breaking.

1. Functions, Scripts, & Data Files

Below is a short description of the various functions, scripts, and data files included in the Brainframe package.

Functions

• brainframe.m: the rendering function, which takes an input_struct, with fields described in the below Fields of input_struct subsection.

• brainframe_inputs_human.m: creates a human input_struct for brainframe.m to render. There are two types of input this functions takes. The first argument to brainframe_inputs_human.m, fpath, specifies where the .mat dependencies are located. All further inputs are optional, keyword arguments that allow for customization of the visualization. See Fields of input_struct for more information. Only specifying the filepath will result in the default input_struct being created.

• brainframe_inputs_mouse.m: creates a mouse input_struct. This function operates exactly the same way as the human function, with exactly the same input_struct fields.

• arrow3: a widely used Mathworks exchange function, which the authors of Brainframe did not create. This function is originally coded by Tom Davis. This function is only called internally by brainframe.m during rendering and it is not necessary for the user to interact with it. Please refer to the following link for more information about this function: arrow3.m

Scripts

• brainframe_Help.m: an extensively commented example script which goes through the basic syntax of using the Brainframe package. This script also contains almost a dozen use case examples, and if run will generate images following the code examples.

• brainframe_humandefault_creator.m: this script is an extensively commented example demonstrating both the default human input_struct field values and how to change or declare input_struct fields directly by using an input_struct.fieldname syntax. This can be found in the ExampleDefaults folder.

• brainframe_mousedefault_creator.m: this script is an extensively commented example demonstrating both the default mouse input_struct field values and how to change or declare input_struct fields directly by using an input_struct.fieldname syntax. This can be found in the ExampleDefaults folder.

Data Files

• brainframe_defaultHuman_datinput.mat: this function contains the volumetric 86-region Desikan atlas, 86-region Desikan connectome, and an 86 element vector with example tau pathology data, per-region, from an average of human patients' tau PET data from ADNI. It is necessary to have this data file for both the brainframe_humandefault_creator.m script and brainframe_inputs_human function to create the standard human default input_struct.

• brainframe_defaultMouse_datinput.mat: this function contains the volumetric 426-region AIBS atlas, 426-region AIBS connectome, and a 426 element vector with example tau pathology data, per-region, from an average of mice 1 month after tau injection. It is necessary to have this data file for both the brainframe_mousedefault_creator.m script and brainframe_inputs_mouse function to create the standard human default input_struct.

• default_human.mat: this data file is the default input_struct created by both the brainframe_humandefault_creator.m script and brainframe_inputs_human function when no Name and Value pair arguments are specified.

• default_mouse.mat: this data file is the default input_struct created by both the brainframe_mousedefault_creator.m script and brainframe_inputs_mouse function when no Name and Value pair arguments are specified.

• PerVox_ExampleData.mat: this is the example per-voxel data that can be rendered. Please see the Brief Examples subsection below and lines 207-216 in the script brainframe_Help.m for further information and example renderings.

2. Fields of input_struct

This section describes the fields of input_struct, the structure object which brainframe.m takes for brain rendering. Modifying the fields of input_struct will change the rendering. You can change these fields directly in an already specified input_struct. Alternately, you can create a new input_struct with either brainframe_inputs_human or brainframe_inputs_mouse. This can be done by specifying each field and the value you want to change it to as a Name and Value pair, following standard MATLAB syntax. This explanation section can also be duplicated within MATLAB by publishing or reading the brainframe_Help.m script.

• voxUreg: Binary flag for per-voxel or per-region visualizations. Human & mouse defaults are both 1. A value of 1 indicates per-region visualizations, and a value of 0 indicates per-voxel visualizations.

• brain_atlas: 3D volumetric atlas with numeric regional IDs per-voxel. Human default is the 86-region Desikan & mouse default is the 213-region AIBS CCF.

• bgcolor: Image background color. Options are 'k', 'w', 'other' (which produces gray). Human & mouse defaults are 'k'.

• savenclose: Binary flag indicating whether the user wants to have on-axis view saved image files (savenclose = 1) with no open MATLAB figure interface or whether the user wants to open the MATLAB figure interface to further interact with or customize their image (savenclose = 0). Mouse and human defaults are both 0.

• img_labels: Your desired filename, as a string, if savenclose == 1. Default is 'yourfilename'. You can also specify a full filepath.

• img_format: Desired image file format as a string, if savenclose == 1. Default is 'png'.

• data: Desired data input, specified as either a vector of per-voxel or per-region entries. Per voxel data can also be specified as a 3D matrix. Default for humans is per-region tau PET pathology data and the default for mice is per-region semi-quantitative tau IHC/IF data.

• nbin: Number of bins per voxel data is divided into for colormap visualization. This field only applies to per-voxel data. Default is 1.

• voxthresh: Threshold restricting the per-voxel data plotted based on the cumulative sum of density. Bounded between 0 and 1 and represents the fraction of total density to be plotted; default is 0.85.

• nreg: Number of regions to go through in specified atlas. This must be a number equal to the number of unique region IDs in brain_atlas. This field is only relevant for per-region data. Default is 86 for humans and 426 for mice.

• region_groups: This is a vector specifying the group each region is part of for colormap purposes. This vector must be the of length nreg. Default is an 86-element vector of 1s for humans and a 426 element vector of 1s for mice.

• sphere: Binary flag specifying whether or not to visualize spheres at region centers. This field is relevant only for per-region visualizations. Default is 0.

• sphere_npts: This field specifies how many points are used to construct the spheres, relative to each sphere's radius. Default is 35.

• centered: This is a two element vector, with fields specifying different elements of pointcloud functionality. The first element is a binary flag specifying whether or not the pointclouds are centered at each regional centroid. The second element denotes the degree of centering, with higher numbers resulting in more centering. Default is [1 2].

• cmap: The colormap used in either per-region or per-voxel visualizations. For per-region visualizations, this must be a number of unique region_groups X 3 matrix of RGB vectors per row. For per-voxel visualizations, this must be an nbin X 3 matrix of RGB vectors per row. Default is [1 0 0].

• xfac: Universal multiplier for sphere radius sizes or point cloud size and density for both per-region and per-voxel visualizations. Default = 1.

• pointsize: Specifies the size of points in the visualizations. Default is 50 for humans and 1 for mice.

• iscon: Binary flag specifying whether to visualize connectivity. Default is 0.

• conmat: Connectivity matrix that is nreg X nreg. Default is an 86 X 86 connectome in Desikan space for humans and the 426 X 426 AIBS connectome for mice. If your connectome is directional, it must have outgoing connectivity for each region per-row and incoming connectivity for each region per-column.

• con_rescale: Universal connectivity multiplier that scales the number of and spread of the ellipses plotted to visually simulate neural connectivity. Ellipses are visualized per region pairs in a number proportional to the C(i,j) entry of the connectome. The default for humans is 1 and the default for mice is 0.01.

• con_width: Specifies the line width of each ellipse visualized. The default is 0.01.

• con_regiongroups: A vector that specifies region groups as integers for the connectivity visualization colormap. This field works analogously to region_groups. Defaults are the same as for region_groups.

• con_cmap: The colormap for connectivity visualizations. This field works analogously to cmap. The default is [0 0.447 0.741].

• con_arch: This field specifies the degree of curvature in the ellipses. Default is 0.5.

• conarrow_WL: Specifies the width and length of the cone arrows used to indicate direction of connection between each region pair. Defaults are [1.5 2.5] in humans and [0.5 0.8] in mice.

3. Brief Examples

Below are some brief examples on the basics of the functionality of the Brainframe package, including using the brainframe_inputs_human and brainframe_inputs_mouse functions to create user specified input_struct objects. For more extensive examples through various probable use cases, including more human use cases, please see the brainframe_Help.m script. For a breakdown of the default input_struct objects that are created under the hood inside the brainframe_inputs_human and brainframe_inputs_mouse functions, please refer to the brainframe_humandefault_creator.m and brainframe_mousedefault_creator.m as example scripts and default_mouse and default_human.mat as example input_struct files. These default examples can be found within the ExampleDefaults folder.

• Visualizing the human default input_struct:

  • Change this to alter the path you load from:

  matpath = cd;

  • Create input_struct:

  input_struct = brainframe_inputs_human(matpath);

  brainframe(input_struct)

  

• Visualizing the mouse default input_struct:

  • Change this to alter the path you load from:

  matpath = cd;

  • Create input_struct:

  input_struct = brainframe_inputs_mouse(matpath);

  brainframe(input_struct)

  view([-1 0 0]);

  

• Saving and closing images, rather than opening image GUI:

  • Create input_struct with savenclose set to 1 rather than 0:

  matpath = cd;

  input_struct = brainframe_inputs_mouse(matpath,'savenclose',1)

  brainframe(input_struct)

  • brainframe.m will now create 3 on-axis view image files but will not render an image in the MATLAB figure GUI. These files will be saved into cd or into the specified filepath.

• Using the region_groups and cmap fields to make colorful per-region visualizations:

  • The below code creates region_groups based on major region IDs
  • Then it creates cmap based on the number of unique region_groups

  reggroups = zeros(213,1);

  amy = 1:11; cer = 12:23; sub = 24:26; hip = 27:37; hyp = 38:57;

  ncx = 58:95; med = 96:120; mid = 121:141; olf = 142:149;

  pal = 150:157; pon = 158:170; str = 171:178; tha = 179:213;

  reggroups(amy) = 1; reggroups(cer) = 2; reggroups(sub) = 3;

  reggroups(hip) = 4; reggroups(hyp) = 5; reggroups(ncx) = 6;

  reggroups(med) = 7; reggroups(mid) = 8; reggroups(olf) = 9;

  reggroups(pal) = 10; reggroups(pon) = 11; reggroups(str) = 12;

  reggroups(tha) = 13;

  reggroups = [reggroups;reggroups];

  cmap = hsv(length(unique(reggroups)));

  • Finally we specify input_struct fields to be customized using a Name and Value pair argument, as is usual in MATLAB. This creates a user specified input_struct with the attendant visualization below:

  matpath = cd;

  input_struct = brainframe_inputs_mouse(matpath,'region_groups',reggroups,'cmap',cmap);

  brainframe(input_struct)

  view([-1 0 0]);

  

• Visualizing only certain regions, rather than all of them:

  • Note in the below example that regions are visualized as spheres

  • Note in the below example that reggroups and cmap from above are used again here to reset region_groups and cmap.

  • Note that sphere radii are being reset using xfac.

  • Note that this visualizes hippocampal regions only!

  • Set regions you don't want visualized to 0 in the data vector, as below:

  datavec = zeros(426,1);

  datavec(27:37) = 1;

  • Reset the fields you want to reset, including using reggroups and cmap from the prior example to reset region_groups and cmap:

  input_struct = brainframe_inputs_mouse(matpath,'region_groups',reggroups,'cmap',cmap,'xfac',0.075,'sphere',1,'data',... datavec);

  brainframe(input_struct);

  view([-1 0 0]);

  

• Visualizing connectivity using the selected regions from above:

  • Turn on connectivity visualization binary flag:

  input_struct.iscon = 1;

  • Zeroing out the diagonal:

  input_struct.conmat = input_struct.conmat - diag(diag(input_struct.conmat));

  • Zeroing out all non-hippocampal connectivity data:

  input_struct.conmat(1:26,:) = 0;

  input_struct.conmat(27:37,[1:26 38:end]) = 0;

  input_struct.conmat(38:end,:) = 0;

  • This line thresholds conmat:

  input_struct.conmat(input_struct.conmat<0.5*mean(nonzeros(input_struct.conmat))) = 0;

  • Visualize connectivity:

  brainframe(input_struct);

  view([-1 0 0]);

  

• Per-voxel visualizations:

  • First load in and select the example per-voxel data:

  matpath = cd;

  load([matpath filesep 'PerVox_ExampleData.mat'],'pervoxdata');

  • Here we use Pvalb+ interneuron distributions

  datinput = pervoxdata.Pvalb;

  • Co-register your data to the reference atlas!

  datinput = imresize3(datinput,[133 81 115]);

  datinput(datinput<0) = 0;

  • Create your input_struct and visualize per-voxel data:

  input_struct = brainframe_inputs_mouse(matpath,'voxUreg',0,'data',datinput,'xfac',0.5,'voxthresh',0.75,'nbin',5,'cmap',autumn(5));

  brainframe(input_struct);

  view([-1 0 0]);

  

• More examples and further information: for more examples and further information about various use cases, including some human use case examples, please see the brainframe_Help.m file.

4. Best Practices, Citations, & Contact

This section recaps best practices, how to use & cite this package, and how to contact the creator.

• Best Practices: Download the contents of this package and keep all scripts in main folder together. Keep all items in main folder until comfortable with use. Moving items around is possible and can be useful, but is not recommended unless the user is both very familiar with MATLAB and has experience with Brainframe. Further help and more use case examples can be found in brainframe_Help.m and default examples can be found in the ExampleDefaults folder.

• Citations: Brainframe can be cited by the GitHub repository URL and the DOI associated with it. When citing Brainframe, please also cite arrow3.m, which was coded by Tom Davis, not the authors of the Brainframe package. Citation information and other information about arrow3.m can be found on the Mathworks exchange site in the hyperlinked text. Please see below for data source citations:

  • Connectome & Volumetric Atlas are from AIBS, originally published in and retrieved from:

    • Oh, SW, et al. 2014. A mesoscale connectome of the mouse brain. Nature. Volume 508, pp. 207-214.

  • Mouse tau pathology data from figures in published literature:

    • Iba, M, et al. 2013. Synthetic tau fibrils mediate transmission of neurofibrillary tangles in a transgenic mouse model of Alzheimer's-like tauopathy. Journal of Neuroscience. Volume 33(3), pp. 1024-1037.

• Contact: to contact the creators of Brainframe, Chris Mezias & Justin Torok, please use the following email addresses:

  • Chris Mezias: cmezias@gmail.com
  • Justin Torok: jlt46@cornell.edu