forked from BIC-MNI/mni_autoreg
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathFAQ
157 lines (123 loc) · 7.43 KB
/
FAQ
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
Frequently Asked Questions
Montreal Neurological Institute Automated Linear Registration Package
This is a list of frequently asked questions regarding the MNI_AutoReg
package.
Questions:
1- How do I get my data into MINC format?
2- How do I view the results of the registration?
3- How do I customize objective functions used for fitting?
4- How do I customize the optimization procedure used?
5- How do I build a new target model for registration?
6- How do I report bugs?
1- HOW DO I GET MY DATA INTO MINC FORMAT?
The simplest way is with rawtominc, one of the utilities
distributed with the MINC package. If you can extract just the raw
images from some data file (e.g. your scanner's native format),
then rawtominc can turn it into a MINC file. Note that you must
know at *least* the volume size, orientation (transverse, coronal,
or sagittal) and voxel size in all three dimensions to create a
MINC file that can be used for registration or display purposes.
See the rawtominc man page for more details.
2- HOW DO I VIEW THE RESULTS OF THE REGISTRATION?
When you run mritotal, the result is not a new image volume
containing the input MRI in Talairach space. Rather, mritotal
outputs a "transform file", which simply contains a homogeneous
transformation matrix that transforms points in native (MRI) space
to Talairach space. Note that these points and transforms are
specified as world coordinates, so are invariant across simple
operations done on the volumes (such as resampling, reshaping,
zero-padding, or removing data). If you wish to see the registered
data, you need to first apply the coordinate transform to the
native MRI data. This is done with mincresample, one of the core
programs of the MINC package. In this case, mincresample requires
that you specify both a transform file (as output by mritotal) and
a sampling grid. For example, say you have just registered the MRI
volume smith_john_mri.mnc with the command
mritotal smith_john_mri.mnc smith_john_mrital.xfm
creating the transform file smith_john_mrital.xfm. You now need to
apply the transform file *and* specify the sampling grid. The
transform file is specified with mincresample's "-transformation"
option. There are a variety of ways to specify the sampling grid,
but the easiest way to specify the sampling grid is with the
"-like" option, which simply tells mincresample to reuse the
sampling grid of another MINC volume. (You can also explicitly
specify all the start coordinates, voxel step sizes, dimension
lengths, and direction cosines of the sampling axes.) For example,
if you have installed the MNI_AutoReg model files in
/usr/local/bic/share/mni-models, you could use the
average305_t1_tal_lin.mnc file there as a resampling model with
the following command:
mincresample smith_john_mri.mnc smith_john_mrital.mnc \
-transform smith_john_mrital.xfm \
-like /usr/local/bic/share/mni-models/average305_t1_tal_lin.mnc
This works by taking every voxel in average305_t1_tal_lin.mnc, transforming
it's (x,y,z) coordinates *back* to native space using the inverse
of the specified transformation, and sampling the native MRI data
there (using trilinear interpolation). For more information, see
the mincresample man page.
Next, you'll want to look at the resampled MRI data, preferably
alongside the registration model (average305_t1_tal_lin.mnc). You can
obtain Register (in precompiled form) from packages.mni.mcgill.ca
When you have Register installed on your system, try the following
command (assuming you have already done the above resampling):
register /usr/local/bic/share/mni-models/average305_t1_tal_lin.mnc smith_john_mrital.mnc
2- HOW DO I CUSTOMIZE OBJECTIVE FUNCTIONS USED FOR FITTING?
3- HOW DO I CUSTOMIZE THE OPTIMIZATION PROCEDURE USED?
Customization of the fitting program is possible by modifying the source
files for minctracc. Notes for adding new objective functions or new
optimization procedures can be found in Doc/howto_add_obj_fn and
Doc/howto_add_opt_fn, repectively.
4- HOW DO I BUILD A NEW TARGET MODEL FOR REGISTRATION?
The default model used with the MNI AutoReg package is is based on
the 305 brain average MRI volume created at the Montreal Neurological
Institute. (This model is *not* included with the MNI AutoReg
package, but is available at
http://packages.bic.mni.mcgill.ca/tgz
Look for mni_autoreg_model-XXX.tar.gz, or possibly a more recent
version. There are four steps to use a new model file: you have to
build the features volumes, construct a brain mask, install the
model, and tell mritotal to use the new model.
1 - Building the feature volumes:
This package (mni_autoreg) installs a script called `make_model' that
contains the commands required to build the feature volumes for the
stereotaxic model. Your input volume must be in Talairach space, and
should have a simple name that will identify the model: for example,
the input volume supplied with the model package is
"average305_t1_tal_lin.mnc". Run make_model with the *base* name of your input
volume, eg. "make_model average305_t1_tal_lin". This will then perform the
necessary processing to make a set of feature volumes (16mm intensity
blur, 8mm intensity and gradient blurs) needed for fitting with
mritotal.
2 - Mask creation:
It is also necessary to have two mask volumes.
One is a mask of all the voxels in the head. This must be named
<BASE>_headmask.mnc, with <BASE> replaced by the base name of your model.
For example, the MNI's model package contains average305_t1_tal_lin_headmask.mnc.
The second mask is a mask of all voxels in the brain. This must be
named <BASE>_mask.mnc; e.g. the file average305_t1_tal_lin_mask.mnc included
with the model package. This particular mask was created
semi-automatically using the interactive program Display;
presumably, other 3-D volume display/manipulation programs have
similar capabilities. (Display is also available from the MNI, but
currently for SGI workstations only. Contact the author, David
MacDonald <[email protected]>, for more information.)
3 - Installation:
You should then install the input volume (<base>.mnc), feature volumes
(<base>_16_blur.mnc, <base>_8_blur.mnc, <base>_8_dxyz.mnc), and mask
volume (<base>_mask.mnc) in your system model directory.
4 - mritotal:
There are two ways to use the new model. The first is to run
mritotal -modeldir /foo/bar/baz -model <BASE> subject.mnc subject.xfm
where /foo/bar/baz is the directory in which the new model files
were installed. In fact, `modeldir /foo/bar/baz' is unnecessary
if you installed the model files into the default model directory.
You can check the default model directory using `mritotal -help'.
The second is to make a copy of mritotal.cfg in your home directory and
replace the -modeldir and -model options with the directory and
base name of your model, respectively. Mritotal will read your
personal mritotal.cfg, and use the new model by default.
5- HOW DO I REPORT BUGS?
Your bug reports play an essential role in making MNI AutoReg
reliable by making the next version of MNI AutoReg work better. Any
unexpected, anomalous, weird, or just plain wrong behaviour should be
reported to Louis Collins <[email protected]>.