Getting started

Prior to running MSM you will need to have passed your data through a surface extraction and inflation pipeline such as FreeSurfer, or the HCP minimal processing pipeline. This is because, if MSM is to work for your data you must have cortical surface meshes that have been mapped to the sphere. In addition, you will require a data file for each mesh, where the data may be scalar (such as sulcal depth, curvature or myelin features) or multivariate (RSNs or fMRI task maps).

Data can be supplied as GIFTI (.func.gii or .shape.gii), ASCII (.asc) files, or as a simple text file (with '.txt' extension) provided the text file has as many columns as there are mesh vertices. Surface files may be supplied as GIFTI (surf.gii) or ASCII (.asc). In general the key files required to run MSM are the:

Examples of the most basic types of call to msm (using these inputs):

Where this assumes you are calling msm from the directory where the data exists. The final option (-o) is the stem of the path where you wish to output your data; we suggest ~/mydirname/L. or ~/mydirname/R. as an example of how you can input left and right hemisphere results into the same directory. Case B shows an example of how, when both datasets have been resampled to a population average surface (such as the HCP's 32k FS_LR surface), it is possible to enter just the average sphere as the input mesh.

MSM Output

The most relevant outputs of MSM are:

where GIFTI outputs are used here only as examples. The program also supports output as ASCII and VTK using the command line option -f.

Template Spaces

For cortical surface alignment it is common practice to align to a population average template space. For adults there are two prominent examples: the FreeSurfer fsaverage mesh and the FS_LR164k and FS_LR32k spaces for the HCP. The FS_LR164k population average space (see here for more details) is based upon fsaverage but has right and left vertex equivalence (it is symmetric). The FS_LR32k surface is a downsampled version of this for f/dMRI processing.

An important point to note about the HCP average space is that the process used to achieve symmetry generates a mesh space rotated with respect to the fsaverage and native mesh spaces. To account for this the HCP have introduced a rotated native sphere called Therefore, by example for some subject 'BOB' using HCP notation in directory /path/to/mystudy/:

In each subject's Native space there will be several spherical mesh representations. Using left as an example, these three meshes will be important for any further processing:

The important thing to take away here is that the HCP provides spheres aligned using MSM, with cortical folding as the features that drive the alignment. But, as the goal of the HCP is fMRI alignment this is highly constrained. This means the regularisation is strong (see configuration files ). Therefore users may wish to define their own MSMSulc alignment. For this they must use the as the input mesh. There are also two more spheres that represent FreeSurfer alignment:

These should not be required, other than for comparisons of results to FreeSurfer-based processing pipelines.

Advanced Command Line Features

In addition to the required inputs to msm, there are several useful options. The most important of these is the --conf call that allows users to supply a configuration file which modifies key parameters of the registration. For optimal running of the registration a configuration file should be supplied (parameters are described in more detail below).

Combining Warps

Another very useful feature is the --trans option. This allows users to specify the output mesh from a previous registration stage. For example, if you wished to initialise registration of some other Native space features (such as myelin maps) by first aligning coarse folding structure using sulcal maps (as performed in our NeuroImage paper), you could run registration in two stages as:

Running registration in this way, rather than simply taking the output from the sulc registration and using it as an input mesh for the RSN registration, allows distortions for all the stages combined (e.g. sulc + myelin here) to be penalised during alignment.

Combining warps of different resolutions

Sometimes the warps you wish to combine are defined on different mesh resolutions, and or the data used to drive the registration has been downsampled to a template mesh. This is common within the HCP when working with functional or diffusion data, which is aligned using MSMsulc and MSMall protocols, then resampled to the FS_LR32k mesh. In these instances it is possible to combine these warps using the --inregister option. In this instance, we can rewrite Step 2 above to include the HCP standard Sulc registration /path/to/mystudy/BOB/MNINonLinear/Native/ and we can incorporate the corresponding template mesh (/path/to/mystudy/BOB/MNINonLinear/fsaverage_LR32K/ in the following way:

File Formats

The output file format is controlled by the -f/--format option and the options are: GIFTI (surfaces are saved as .surf.gii and data as .func.gii); ASCII (surfaces are saved as .asc and data is saved as .dpv); ASCII_MAT (surfaces are saved as .asc and data is saved as a simple matrix in a textfile .txt); VTK (surfaces as .vtk and data as .txt). For more details on the .dpv format (which is FreeSurfer compatible, but differentiates surface from data files) please see the following blog post:

Costfunction Weighting

Costfunction weighting (CFw) can be controlled using --inweight and --refweight options. This allows you to supply a weighting mask for each of your source and reference meshes, although it is possible to run msm with only one. The CFw masks can be multivariate which allows you to vary the contribution of different features. For example in the "Multimodal alignment" section of our paper we use a single, multivariate CFw mask created on our template image (and therefore passed as a --refweight option) to vary the contribution of our different modalities to the registration.

Resolution and Smoothing

Two final useful parameters are --levels which allows you to control the number of resolution levels run during the course of the registration e.g. --levels=2. This supersedes the settings in the configuration file. Finally --smoothout controls the smoothing of the data after projection to the template image. By default the registration uses adaptive barycentric resampling (reference). However, this option will allow the user to smooth using a gaussian kernel with an input parameter equal to the standard deviation.


Therefore, with all command line parameters used an msm call might look like this:

This will repeat Step 2 above, but this time each of the meshes will have a corresponding weighting function supplied in the form of a GIFTI func.gii (but this could also be shape.gii, .asc or as matrix in a text file); the output of the registration will also be smoothed using a kernel of standard deviation 2. The registration will be stopped after two cycles or registration levels irrespective of the number of levels specified in the configuration file.

Configuration Files

Configuration files modify all tunable parameters of the registration. For a full list of all registration parameters you can enter:

Some parameters require inputs for every stage of the registration, and are input as comma separated lists e.g. --lambda=0,0.1,0.2,0.3 (for four levels). These are:

Other parameters need only be specified once:

We supply a series of configuration files, each tuned to work with different (sulcal depth, myelin, and RSN) data. An example of the sulcal depth config file (which also forms the default parameterisation in the absence of any supplied configuration file) is:

The comma separated lists above represent parameters per level, and the number of resolution levels run by msm can be controlled by the length of the lists specified here. Registration may also be initialised using an affine alignment step, run as an additional level at the beginning. Therefore, the above case is stating that the registration should run one affine step using: SSD as a similarity measure, 50 iterations, input mesh smoothing 4mm, reference mesh smoothing 2mm, on a data grid of resolution 2562 vertices; Following this discrete optimisation is run over 3 levels with 3 iterations at each level, using control point grid resolutions 162, 642, and 2562, where the sampling grid resolution is 2 subdivisions above this, and the data grids have resolution: 2562, 10242 and 40962 vertices. Smoothing is applied to the source image as 4, 2, then 1mm sigma smoothing kernels, and to the reference image as 2, 1 and 1mmm smoothing. --IN indicates that the source intensity distribution is matched to the target intensity distribution, once at the beginning of the registration.

If you choose to edit or optimise the config files then it is important to remember that all multiresolution level parameter lists must have the same length, else the program will throw the following error:

MeshREG ERROR:: config file parameter list lengths are inconsistent

In addition, as affine registration only implements the following parameters: --opt, --simval, --it, --sigma_in, --sigma_ref, --IN, --VN, --scale, --excl, for all other multi level parameters, it is necessary to supply a zero value for the AFFINE stage.

Regular Mesh Surfaces

The number of faces in an icosahedron is 20 and subsampling this gives rise to high resolution representations of a sphere that are used for controlling the grid spacing. Serial subsampling leads to polyhedra with the following number of faces: 42, 162, 642, 2562, 10242, 40962. These correspond to the codes: 1, 2, 3, 4, 5, 6. Below are examples of codes 0 (icosahedron), 1 and 2 in the first row and 4 and 5 in the second row.



Post Processing

Transforming Unseen Data

In msm a warp or deformation field is prescribed by two meshes representing the start and end point of a transformation. For example, using the common syntax of this user guide, the warp for the left hemisphere is between the and These represent the same data, so have the same number of vertices and vertex numbers correspond. Therefore, the change in coordinates from the input_mesh to sphere.reg tells us where each vertex on the input_mesh has to move so as to optimise overlap of the input and reference data. Nevertheless, the warp alone is not sufficient to enable direct comparisons between the alignment of the input and reference featurespace (i.e. sulc/myelin/RSN maps) following alignment. To achieve this it is also necessary to resample the input features onto the reference mesh surface. This will take the features for each input mesh vertex and project them onto to the reference mesh at their new location i.e. according to MSM generates the resampling of the input data automatically as L.transformed_and_reprojected.func.gii. However, if you wish to project new data through this transformation i.e. project myelin or RSN data through a sulc transformation the following functions will be necessary:

Two functions are supplied for the warping and resampling of unseen data:

msmapplywarp: of spheres

This can be useful for HCP data where fMRI data is resampled onto the low resolution 32k_FS_LR mesh, but other data lies on the high resolution 164k_FS_LR surface. It will assume that the starting point of the deformation is given by the un-deformed icosphere equivalent to However it is important to make sure with this formulation that the to-be-deformed mesh ( here) was in alignment with the input mesh used in the call to MSM that produces the warp

msmapplywarp: of anat

. msmapplywarp input_mesh.reg.gii -anat

Here the input_mesh.sphere.reg.gii is the deformed sphere output from MSM and the will be the projection of this deformed sphere onto the target white surface. Therefore the mesh will have the mesh topology of the input, with the shape of the target. To resample the target mesh onto the input anatomy apply:

. msmapplywarp -anat input_mesh.reg.gii


Where output_metric_basename refers to the desired output name without the file ending (i.e. .func.gii), and is the registered sphere output from MSM

Important Options

Estimating metric distortion

It is also possible to estimate the strength of the deformation in terms of how it distorts the input mesh. This can be estimated in terms of the areal distortion, or log2(A2/A1) where A1 and A2 represent mesh face areas before and after projection, respectively. Per Vertex values are taken by weighted averaging of the distortions for every mesh face connected to each vertex. Areal distortions can be estimated as:

. estimate_metric_distortion input_mesh.sphere.reg.gii distortion_basename 

In general it is undesirable for areal distortions to exceed 2 or more (four fold expansion)


M.F. Glasser, S.N. Sotiropoulos, J.A. Wilson, T.S. Coalson, B. Fischl, J.L. Andersson, J. Xu, S. Jbabdi, M. Webster, J. R. Polimeni, D.C. Van Essen, M. Jenkinson, The minimal preprocessing pipelines for the Human Connectome Project, NeuroImage, Volume 80, 15 October 2013, Pages 105-124, ISSN 1053-8119,


MSM/UserGuide (last edited 06:46:13 29-11-2017 by EmmaRobinson)