This page provides some help and explanation on a few simple tools which can be used on the networked Unix workstations running the Solaris or Linux operating systems or the Windows 7 PCs of the Wolfson Molecular Imaging Centre at the University of Manchester.
All applications described here are available on the Unix application disk. This disk should normally be mounted to client workstations at the mount point /opt/wmic.
Platform independent executables (such as Bourne shell scripts) as well as wrapper scripts are installed in the directory /opt/wmic/bin.
The platform specific binary executables are located in six separate subdirectories.
Some of the binary programs require dynamic linking with the AVW library. For those, a wrapper script in /opt/wmic/bin sets the appropriate environment variables and then invokes the platform specific binary from one of the subdirectories listed above.
Exit status
The following exit values are returned:
0
Successful completion.
1
Incorrect usage, command help is displayed.
2
Memory allocation failed.
3
AVW licence validation error.
>3
Programme specific error, a corresponding error message is printed.
In the following examples, a per cent sign % at the beginning of the line indicates the command line prompt of a C shell compatible shell.
| Type | binary executable |
| Location | /opt/wmic/bin/sparc, /opt/wmic/bin/sparcv9, /opt/wmic/bin/i386, /opt/wmic/bin/x86_64, \\MMIC-FS1\WMIC_Apps, \\MMIC-FS1\WMIC_Apps\win64 |
| Version | 3.4 |
Description:
ShowHdr is a command line program to print the contents
of Analyze
7.5 Image header files on the
standard output stream.
ShowHdr tries
to identify the
endianness of the header files. If required, the bytes are swapped
and a notice it printed on the
standard error stream.
ShowHdr's output
can be piped through the
stream editor
sed to extract
the value of a specific field.
With the command line option -c
a field by field comparison of two header files is performed.
If a field's content is equal in both files, it is printed
as in the standard mode of ShowHdr.
Otherwise, the contents from both files is shown, and the line
is marked with 5 stars *****.
This allows the use of
fgrep to
filter those lines out where the contents differs between
the two files (see Example 3 below).
Usage:
ShowHdr [-f format] h01234.hdr t56789.hdr ...
shows the contents of all files specified
ShowHdr -c [-f format] a.hdr b.hdr
shows a field by field comparison of the two files
-f format a string specifying the output conversion of the
floating point values (default: %g)
See the Standard C Library Functions printf(3C)
manpage for examples.
% ShowHdr h00005_20060623_160920_s09_nf02.hdr h00005_20060623_160920_s15_nf29.hdr Notice: bytes swapped. Analyze header dump of: <h00005_20060623_160920_s09_nf02.hdr> hk.sizeof_hdr : <348> hk.data_type : <dsr > hk.db_name : < > hk.extents : <0> hk.session_error : <0> hk.regular : <r> hk.hkey_un0 : <0> dime.dim[0] : <4> dime.dim[1] : <256> dime.dim[2] : <256> dime.dim[3] : <81> dime.dim[4] : <2> dime.dim[5] : <0> dime.dim[6] : <0> dime.dim[7] : <0> dime.vox_units : <mm > dime.cal_units : <kBq/ml > dime.unused1 : <0> dime.datatype : <4> dime.bitpix : <16> dime.dim_un0 : <0> dime.pixdim[0] : <0> dime.pixdim[1] : <2.66615> dime.pixdim[2] : <2.66615> dime.pixdim[3] : <2> dime.pixdim[4] : <0> dime.pixdim[5] : <0> dime.pixdim[6] : <0> dime.pixdim[7] : <0> dime.vox_offset : <0> dime.funused1 : <0.00326873> dime.funused2 : <122.24> dime.funused3 : <0> dime.cal_max : <107.106> dime.cal_min : <0> dime.compressed : <0> dime.verified : <0> dime.glmax : <32767> dime.glmin : <0> hist.descrip : <PET Water Ranging Study - Pelvis (Adult) > hist.aux_file : <none > hist.orient : <> hist.originator : <petct1 > hist.generated : <EM2D 4i16s> hist.scannum : <9 > hist.patient_id : <h00005 > hist.exp_date : <2006-06-23> hist.exp_time : <16:09:20.9> hist.hist_un0 : <FFS> hist.views : <0> hist.vols_added : <0> hist.start_field : <0> hist.field_skip : <0> hist.omax : <0> hist.omin : <0> hist.smax : <0> hist.smin : <0> ================================================================================ Notice: bytes swapped. Analyze header dump of: <h00005_20060623_160920_s15_nf29.hdr> hk.sizeof_hdr : <348> hk.data_type : <dsr > hk.db_name : < > hk.extents : <0> hk.session_error : <0> hk.regular : <r> hk.hkey_un0 : <0> dime.dim[0] : <4> dime.dim[1] : <256> dime.dim[2] : <256> dime.dim[3] : <81> dime.dim[4] : <29> dime.dim[5] : <0> dime.dim[6] : <0> dime.dim[7] : <0> dime.vox_units : <mm > dime.cal_units : <kBq/ml > dime.unused1 : <0> dime.datatype : <4> dime.bitpix : <16> dime.dim_un0 : <0> dime.pixdim[0] : <0> dime.pixdim[1] : <2.66615> dime.pixdim[2] : <2.66615> dime.pixdim[3] : <2> dime.pixdim[4] : <0> dime.pixdim[5] : <0> dime.pixdim[6] : <0> dime.pixdim[7] : <0> dime.vox_offset : <0> dime.funused1 : <0.0972375> dime.funused2 : <122.24> dime.funused3 : <0> dime.cal_max : <3186.18> dime.cal_min : <-2806.08> dime.compressed : <0> dime.verified : <0> dime.glmax : <32767> dime.glmin : <-28858> hist.descrip : <PET Water Ranging Study - Pelvis (Adult) > hist.aux_file : <none > hist.orient : <> hist.originator : <petct1 > hist.generated : <DIFT > hist.scannum : <15 > hist.patient_id : <h00005 > hist.exp_date : <2006-06-23> hist.exp_time : <16:09:20.9> hist.hist_un0 : <FFS> hist.views : <0> hist.vols_added : <0> hist.start_field : <0> hist.field_skip : <0> hist.omax : <0> hist.omin : <0> hist.smax : <0> hist.smin : <0>
% ShowHdr h00005_20060623_160920_s09_nf02.hdr h00005_20060623_160920_s15_nf29.hdr | sed -n '/^ dime.dim\[4\] : <\([0-9]*\)>$/s//\1/p' 2 29
% ShowHdr -c h00450_mr2pet.hdr spm5/h00450_mr2pet.hdr | fgrep "*****"
hist.originator : <> <D> *****
spm.origx : <0> <68> *****
spm.origy : <0> <98> *****
spm.origz : <0> <60> *****
ShowHdr can also be used to convert the single precision floating point values using any desired format string. For example,
% ShowHdr -f %.7e t00321_45min.hdr | sed -n '/ dime.funused1 : <\(.*\)>/s//\1/p' 3.0592648e-08
prints the ROI Scale Factor in single precision. In contrast, the answer by od would depend on whether the current machine has the same endianness as the Analyze 7.5 Image header file.
% od -A n -j 112 -N 4 -t f4 t00321_45min.hdr
3.0592648e-08
looks all right whereas
%od -A n -j 112 -N 4 -t f4 t00321_45min.hdr
-7.3994286e+34
appears to be the wrong endian.
The command
% ShowHdr -f %18.16g t00321_45min.hdr | sed -n '/ dime.funused1 : <\(.*\)>/s//\1/p' 3.059264841454024e-08
prints the double precision representation of the ROI Scale Factor.
| Type | binary executable |
| Location | /opt/wmic/bin/sparc, /opt/wmic/bin/sparcv9, /opt/wmic/bin/i386, /opt/wmic/bin/x86_64, \\MMIC-FS1\WMIC_Apps, \\MMIC-FS1\WMIC_Apps\win64 |
| Version | 2.2 |
Description:
EditHdr is a command line program to edit the fields in
an Analyze 7.5
Image header file.
EditHdr tries
to identify the
endianness of the header file. If required, the bytes are swapped
internally. Regardless of the system where
EditHdr is run,
the original
endianness of the header file will remain unchanged.
Usage: EditHdr image.hdr field0=value0 field1=value1 ... fieldn=valuen
image.hdr the name of the Analyze Image 7.5 header file to be
edited.
fieldi=valuei the name of the header field with the value to be
assigned.
This is a list of the header fields and the data types of their values:
hk.sizeof_hdr = int
hk.data_type = char[10]
hk.db_name = char[18]
hk.extents = int
hk.session_error= short int
hk.regular = char[1]
hk.hkey_un0 = char[1]
dime.dim[0] = short int
dime.dim[1] = short int
dime.dim[2] = short int
dime.dim[3] = short int
dime.dim[4] = short int
dime.dim[5] = short int
dime.dim[6] = short int
dime.dim[7] = short int
dime.vox_units = char[4]
dime.cal_units = char[8]
dime.unused1 = short int
dime.datatype = short int
dime.bitpix = short int
dime.dim_un0 = short int
dime.pixdim[0] = float
dime.pixdim[1] = float
dime.pixdim[2] = float
dime.pixdim[3] = float
dime.pixdim[4] = float
dime.pixdim[5] = float
dime.pixdim[6] = float
dime.pixdim[7] = float
dime.vox_offset = float
dime.funused1 = float
dime.funused2 = float
dime.funused3 = float
dime.cal_max = float
dime.cal_min = float
dime.compressed = int
dime.verified = int
dime.glmax = int
dime.glmin = int
hist.descrip = char[80]
hist.aux_file = char[24]
hist.orient = char[1]
hist.originator = char[10]
spm.origx = short int
spm.origy = short int
spm.origz = short int
hist.generated = char[10]
hist.scannum = char[10]
hist.patient_id = char[10]
hist.exp_date = char[10]
hist.exp_time = char[10]
hist.hist_un0 = char[3]
hist.views = int
hist.vols_added = int
hist.start_field= int
hist.field_skip = int
hist.omax = int
hist.omin = int
hist.smax = int
hist.smin = int
Note that in most cases the brackets used to specify the subscripts of array fields have to be protected from interpretation by the shell! This can be done with back slash \ or double quote " characters. The same applies to white spaces in character arrays.
% EditHdr h01234.hdr dime.pixdim\[1\]=2.092 "dime.pixdim[2]"=2.092 "dime.pixdim[3]=2.42" dime.vox_units=mm dime.cal_units=kBq/ml hist.descrip=our\ example\ with\ spaces\ in\ between hist.patient_id="=my uncle"
Use the dd command on a Unix system to first create an empty 348-bytes header file new.hdr:
% dd if=/dev/zero bs=348 count=1 of=new.hdr 1+0 records in 1+0 records out
Then set all the required fields to the appropriate values. For example, the datatype value for AVW_FLOAT (32-bit single precision float) is defined as 16 for an Analyze 7.5 Image and the corresponding number of bits per pixel bitpix is 32:
% EditHdr new.hdr hk.sizeof_hdr=348 hk.regular=r "dime.dim[0]=4" "dime.dim[1]=256" "dime.dim[2]=256" "dime.dim[3]=207" "dime.dim[4]=1" dime.datatype=16 dime.bitpix=32 "dime.pixdim[1]=1.21875" "dime.pixdim[2]=1.21875" "dime.pixdim[3]=1.21875"
Paired up with a 256 x 256 x 207 voxels HRRT Interfile image file new.img, the resulting Analyze 7.5 Image can now be loaded into any image display software. As no Y or Z flip has been applied to the data, the image will not be shown in standard radiological orientation.
This example transfers the value of the ROI Scale factor in double precision from the image pre.hdr to the image post.hdr:
% EditHdr post.hdr dime.funused1=`ShowHdr -f %18.16g pre.hdr | sed -n '/ dime.funused1 : <\(.*\)>/s//\1/p'`
Check that the binary values of dime.funused1 in the two images are identical:
% ShowHdr -c pre.hdr post.hdr | fgrep dime.funused1 dime.funused1 : <0.00138057>
The output should only show one value and no asterisks*****at the end of the line.
| Type | binary executables |
| Location | /opt/wmic/mriwarp/1.52/sparcv9, /opt/wmic/mriwarp/1.52/sparc, /opt/wmic/bin/x86_64, /opt/wmic/mriwarp/1.52/i386, \\wmic-san1\secondary_data\appdir\mriwarp\1.52\win32, \\wmic-san1\secondary_data\appdir\mriwarp\1.52\win64 |
| Version | 1.52 |
Description: MRIWarp is a warping tool for intersubject registration of brain images constructed by Ulrik Kjems. Apart from warping there are also (command line) functions for ANALYZE header information, mirroring, translation, subsampling, ... The package cannot only be used as a preprocessing step in function neuroimaging but also as a step in deformation-based morphometry, see reference.
Quick reference to the 18 utility programmes:
| Type | binary executables |
| Location | /opt/wmic/air/5.3.0/sparcv9, /opt/wmic/air/5.3.0/sparc, /opt/wmic/air/5.3.0/x86_64, /opt/wmic/air/5.3.0/i386, \\wmic-san1\secondary_data\appdir\air\5.3.0\win32, \\wmic-san1\secondary_data\appdir\air\5.3.0\win64 |
| Version | 5.3.0 |
Description:
The Automated Image
Registration (AIR) package written by
Roger P. Woods
includes automated routines for aligning and reslicing tomographic
image data. In addition, a number of utility routines and error checking
routines are provided. The routines are invoked as standard UNIX commands.
The AIR package does not provide any graphics or image display capabilities.
The default image file format is compatible with the
ANALYZE format
developed at the Mayo Clinic. No file format converters are provided.
Programs that will display, generate or modify header files are
| Type | binary executables |
| Location | /opt/wmic/bin/sparc, /opt/wmic/bin/x86_64, /opt/wmic/bin/i386 |
| Version | 0.5 |
Description: NIfTI-1 is a new Analyze-style data format. A collection of I/O routines for the NIfTI-1 neuroimage data format is available in niftilib. From the C package nifticlib version 0.5 (31 May 2007), three tools are available.
Description: FSL is a comprehensive library of analysis tools for brain imaging data created by the Analysis Group, FMRIB, Oxford, UK. Most of the tools can be run both from the command line and as GUIs ("point-and-click" graphical user interfaces).
The version and platform specific distributions are located in separate subdirectories within /opt/wmic/fsl reflected by the environment variable FSLDIR. For csh and similar shells, the following command can be used to define FSLDIR
Then add the bin
subdirectory to the PATH
environment variable and define the two
FSL environment variables
FSLTCLSH
and FSLWISH,
respectively
setenv PATH ${PATH}:${FSLDIR}/bin (or for version 6.0.6 and newer: setenv PATH ${PATH}:${FSLDIR}/share/fsl/bin )The last statement defines in which file format FSL will save output images.
setenv FSLTCLSH ${FSLDIR}/bin/fsltclsh
setenv FSLWISH ${FSLDIR}/bin/fslwish
setenv FSLOUTPUTTYPE NIFTI
Of the FSL tools, only FSLUTILS is mentioned here as it provides a set of useful command-line utilities which allow the conversion, processing etc of Analyze and Nifti format data sets. Three examples are
| Type | Bourne shell script, requires the binaries ShowHdr and EditHdr |
| Location | /opt/wmic/bin |
| Version | 3.4 |
Description: chopima is a
command line program to cut dynamic
Analyze 7.5 Image file[s]
(multiple frames) into series of individual single-frame
Analyze 7.5 Image files.
This might be required for instance in order to perform a frame-to-frame
realignment of the data.
By default, the minimum and maximum values in the file headers are only copied
from the source file. This often eases the reassemblance of the individual frames
into one big dynamic image file later on.
If correct minimum and maximum values in the individual header files are required,
then the -u option should be used.
However, this only produces correct results if
chopima is run on a system with
the same endianness as the
Analyze 7.5 Image file!
Usage: chopima [-0] [-o] [-s] [-u] h01234_lm1_1_CZ.hdr t56789dy1_Cal.hdr ...
-0 use leading zeros in the frame numbers (i.e. "001" instead of "1").
-o overwrite ON (default: if the output files already exist, they will
not be overwritten)
-s suppress scanner type identification (set this flag for volumes with
nonstandard image matrix size)
-u update the file headers with the new, frame-specific min/max values
(default: copy the min/max values from the header of the dynamic)
Using this option will dramatically increase the execution time!!!
./h00005_20060623_160920_s15_nf29.img: 29 frames, 307888128 bytes, 10616832 bytes per frame (HiREZ scanner). ./h00005_20060623_160920_s15_nf29_v001.img -> written, ./h00005_20060623_160920_s15_nf29_v001.hdr -> written. ./h00005_20060623_160920_s15_nf29_v002.img -> written, ./h00005_20060623_160920_s15_nf29_v002.hdr -> written. ./h00005_20060623_160920_s15_nf29_v003.img -> written, ./h00005_20060623_160920_s15_nf29_v003.hdr -> written. ./h00005_20060623_160920_s15_nf29_v004.img -> written, ./h00005_20060623_160920_s15_nf29_v004.hdr -> written. ./h00005_20060623_160920_s15_nf29_v005.img -> written, ./h00005_20060623_160920_s15_nf29_v005.hdr -> written. ./h00005_20060623_160920_s15_nf29_v006.img -> written, ./h00005_20060623_160920_s15_nf29_v006.hdr -> written. ./h00005_20060623_160920_s15_nf29_v007.img -> written, ./h00005_20060623_160920_s15_nf29_v007.hdr -> written. ./h00005_20060623_160920_s15_nf29_v008.img -> written, ./h00005_20060623_160920_s15_nf29_v008.hdr -> written. ./h00005_20060623_160920_s15_nf29_v009.img -> written, ./h00005_20060623_160920_s15_nf29_v009.hdr -> written. ./h00005_20060623_160920_s15_nf29_v010.img -> written, ./h00005_20060623_160920_s15_nf29_v010.hdr -> written. ./h00005_20060623_160920_s15_nf29_v011.img -> written, ./h00005_20060623_160920_s15_nf29_v011.hdr -> written. ./h00005_20060623_160920_s15_nf29_v012.img -> written, ./h00005_20060623_160920_s15_nf29_v012.hdr -> written. ./h00005_20060623_160920_s15_nf29_v013.img -> written, ./h00005_20060623_160920_s15_nf29_v013.hdr -> written. ./h00005_20060623_160920_s15_nf29_v014.img -> written, ./h00005_20060623_160920_s15_nf29_v014.hdr -> written. ./h00005_20060623_160920_s15_nf29_v015.img -> written, ./h00005_20060623_160920_s15_nf29_v015.hdr -> written. ./h00005_20060623_160920_s15_nf29_v016.img -> written, ./h00005_20060623_160920_s15_nf29_v016.hdr -> written. ./h00005_20060623_160920_s15_nf29_v017.img -> written, ./h00005_20060623_160920_s15_nf29_v017.hdr -> written. ./h00005_20060623_160920_s15_nf29_v018.img -> written, ./h00005_20060623_160920_s15_nf29_v018.hdr -> written. ./h00005_20060623_160920_s15_nf29_v019.img -> written, ./h00005_20060623_160920_s15_nf29_v019.hdr -> written. ./h00005_20060623_160920_s15_nf29_v020.img -> written, ./h00005_20060623_160920_s15_nf29_v020.hdr -> written. ./h00005_20060623_160920_s15_nf29_v021.img -> written, ./h00005_20060623_160920_s15_nf29_v021.hdr -> written. ./h00005_20060623_160920_s15_nf29_v022.img -> written, ./h00005_20060623_160920_s15_nf29_v022.hdr -> written. ./h00005_20060623_160920_s15_nf29_v023.img -> written, ./h00005_20060623_160920_s15_nf29_v023.hdr -> written. ./h00005_20060623_160920_s15_nf29_v024.img -> written, ./h00005_20060623_160920_s15_nf29_v024.hdr -> written. ./h00005_20060623_160920_s15_nf29_v025.img -> written, ./h00005_20060623_160920_s15_nf29_v025.hdr -> written. ./h00005_20060623_160920_s15_nf29_v026.img -> written, ./h00005_20060623_160920_s15_nf29_v026.hdr -> written. ./h00005_20060623_160920_s15_nf29_v027.img -> written, ./h00005_20060623_160920_s15_nf29_v027.hdr -> written. ./h00005_20060623_160920_s15_nf29_v028.img -> written, ./h00005_20060623_160920_s15_nf29_v028.hdr -> written. ./h00005_20060623_160920_s15_nf29_v029.img -> written, ./h00005_20060623_160920_s15_nf29_v029.hdr -> written.
./image_10.img: 33 frames, 895352832 bytes, 27131904 bytes per frame (HRRT scanner). ./image_10_v1.img -> written, ./image_10_v1.hdr -> written (min=0,max=12). ./image_10_v2.img -> written, ./image_10_v2.hdr -> written (min=0,max=3743). ./image_10_v3.img -> written, ./image_10_v3.hdr -> written (min=0,max=10680). ./image_10_v4.img -> written, ./image_10_v4.hdr -> written (min=0,max=23744). ./image_10_v5.img -> written, ./image_10_v5.hdr -> written (min=0,max=21012). ./image_10_v6.img -> written, ./image_10_v6.hdr -> written (min=0,max=21774). ./image_10_v7.img -> written, ./image_10_v7.hdr -> written (min=0,max=17885). ./image_10_v8.img -> written, ./image_10_v8.hdr -> written (min=0,max=18652). ./image_10_v9.img -> written, ./image_10_v9.hdr -> written (min=0,max=21477). ./image_10_v10.img -> written, ./image_10_v10.hdr -> written (min=0,max=29326). ./image_10_v11.img -> written, ./image_10_v11.hdr -> written (min=0,max=32767). ./image_10_v12.img -> written, ./image_10_v12.hdr -> written (min=0,max=18903). ./image_10_v13.img -> written, ./image_10_v13.hdr -> written (min=0,max=21618). ./image_10_v14.img -> written, ./image_10_v14.hdr -> written (min=0,max=12352). ./image_10_v15.img -> written, ./image_10_v15.hdr -> written (min=0,max=8566). ./image_10_v16.img -> written, ./image_10_v16.hdr -> written (min=0,max=6991). ./image_10_v17.img -> written, ./image_10_v17.hdr -> written (min=0,max=4975). ./image_10_v18.img -> written, ./image_10_v18.hdr -> written (min=0,max=5190). ./image_10_v19.img -> written, ./image_10_v19.hdr -> written (min=0,max=2770). ./image_10_v20.img -> written, ./image_10_v20.hdr -> written (min=0,max=2845). ./image_10_v21.img -> written, ./image_10_v21.hdr -> written (min=0,max=3613). ./image_10_v22.img -> written, ./image_10_v22.hdr -> written (min=0,max=1687). ./image_10_v23.img -> written, ./image_10_v23.hdr -> written (min=0,max=2417). ./image_10_v24.img -> written, ./image_10_v24.hdr -> written (min=0,max=1454). ./image_10_v25.img -> written, ./image_10_v25.hdr -> written (min=0,max=1274). ./image_10_v26.img -> written, ./image_10_v26.hdr -> written (min=0,max=1223). ./image_10_v27.img -> written, ./image_10_v27.hdr -> written (min=0,max=1085). ./image_10_v28.img -> written, ./image_10_v28.hdr -> written (min=0,max=1273). ./image_10_v29.img -> written, ./image_10_v29.hdr -> written (min=0,max=1195). ./image_10_v30.img -> written, ./image_10_v30.hdr -> written (min=0,max=1580). ./image_10_v31.img -> written, ./image_10_v31.hdr -> written (min=0,max=1275). ./image_10_v32.img -> written, ./image_10_v32.hdr -> written (min=0,max=1189). ./image_10_v33.img -> written, ./image_10_v33.hdr -> written (min=0,max=1115).
Note that the dynamic image was reconstruced using an iterative algorithm. Hence there are no negative voxel values. The maximum value was observed in frame (volume) 11.
| Type | Bourne shell script, requires the binaries ShowHdr and EditHdr |
| Location | /opt/wmic/bin |
| Version | 2.3 |
Description: cmrgima is a Bourne shell script to create parametric images of the Cerebral Metabolic Rate of Glucose (CMRGl) from parametric images of IRF45 min (obtained from Spectral Analysis) or KI (e.g. created with plasma input function Patlak Analysis). The parametric images must be in Analyze image 7.5 file format and can currently be provided by any of the following three applications:
- MICKPM's pixel Spectral Analysis (scanid_45min.hdr/.img),
- MICKPM's pixel Patlak Analysis with plasma input function (scanid_Ki.hdr/.img),
- Piwave's Patlak Analysis with plasma input function (scanid_wave_par3D.hdr/.img).
To calculate the mean plasma glucose concentration during the PET scan, the user is prompted to enter a series of plasma glucose concentration measurements in millimolar units (i.e. in mmol/l= μmol/ml). The relative molecular mass of glucose C6H12O6 is 180 g/mol, plasma glucose measurements in gram glucose per liter plasma have to be converted accordingly.
The default lumped constant in cmrgima is 0.48 (Sokoloff et al., 1977). The command line option -l LC can be used to specify a different value LC as lumped constant.
The CMRGl parametric images are saved as signed 16-bit Analyze image file in units of μmol/ml/min = mmol/l/min. To convert them into mg/100 ml/min, all values have to be multiplied by 18.
The output images have the same endianness as the input files independent of the machine type on which cmrgima is run.
Reference:
Sokoloff, L.; Reivich, M.; Kennedy, C.; Des Rosiers, M.H.; Patlak, C.S.;
Pettigrew, K.D.; Sakurada, O.; Shinohara, M.
The [14C]deoxyglucose method for the measurement of local
cerebral glucose utilization: theory, procedure, and normal values in the
conscious and anesthetized albino rat.
J. Neurochem. 28 (1977), 897 - 916.
Usage: cmrgima [-l LC] [-o] h01234_45min.hdr t56789_PPM_Ki.hdr ...
-l LC define an alternative value for the lumped constant (default
value: LC=0.48)
-o overwrite ON (default: if the output files `h01234_45min_cmrg.hdr'
and `h01234_45min_cmrg.img' already exist, they will not be over-
written)
% cmrgima t00321_45min.hdr
Input file : "t00321_45min.hdr" (type 0)
ROIScale = 3.0592648e-08 = 0x330364f9 (single precision)
cal_max = 1.0024293e-03 = 0x3a8363f2 (single precision)
Enter Plasma Glucose Samples in Millimolar Units (concentration in mmol/l):
*** press Ctrl-D (^D) to finish input ***
Sample 0 = 5.70
Sample 1 = 5.54
Sample 2 = 5.50
Sample 3 = 5.36
Sample 4 = 5.50
Sample 5 = done.
Mean = 5.52000000000000e+00 mmol/l
Lumped Con= 0.48
Output Image= Input Image * 1.15000000000000e+01 * 60
ROIScale = 2.11089271200000e-05 (double precision)
cal_max = 6.91676217000000e-01 (double precision)
Output files : "./t00321_45min_cmrg.img" -> written.
"./t00321_45min_cmrg.hdr" -> written.
ROIScale = 2.1108926e-05 = 0x37b11313 (single precision)
cal_max = 6.9167620e-01 = 0x3f3111b1 (single precision)
The conversion was performed on a machine with the same endianness as the input image.
% cmrgima t00321_Ki.hdr
WARNING: t00321_Ki.hdr: glmax= 32766 (should normally be 32767).
Input file : "t00321_Ki.hdr" (type 1)
ROIScale = 1.1460926e-05 = 0x4d484037 (single precision - bytes need to be swapped)
cal_max = 3.7554014e-01 = 0xcc46c03e (single precision - bytes need to be swapped)
cal_min = -3.7077731e-01 = 0x86d6bdbe (single precision - bytes need to be swapped)
Enter Plasma Glucose Samples in Millimolar Units (concentration in mmol/l):
*** press Ctrl-D (^D) to finish input ***
Sample 0 = 5.70
Sample 1 = 5.54
Sample 2 = 5.50
Sample 3 = 5.36
Sample 4 = 5.50
Sample 5 = done.
Mean = 5.52000000000000e+00 mmol/l
Lumped Con= 0.48
Output Image= Input Image * 1.15000000000000e+01 * 1
ROIScale = 1.31800649000000e-04 (double precision)
cal_max = 4.31871161000000e+00 (double precision)
cal_min = -4.26393906500000e+00 (double precision)
Output files : "./t00321_Ki_cmrg.img" -> written.
"./t00321_Ki_cmrg.hdr" -> written.
ROIScale = 1.3180065e-04 = 0xf8330a39 (single precision - bytes need to be swapped)
cal_max = 4.3187118e+00 = 0xe3328a40 (single precision - bytes need to be swapped)
cal_min = -4.2639389e+00 = 0x307288c0 (single precision - bytes need to be swapped)
The conversion was performed on a machine with an endianness different to the input image. This can be seen at the comment bytes need to be swapped for the hexadecimal screen output of the 32-bit single precision float values. However, this has no effect on the output images created, they are the same independent of the machine type on which cmrgima is run.
| Type | binary executable, AVW library required |
| Location |
/opt/wmic/bin (Solaris/Linux wrapper script) \\MMIC_FS1\WMIC_Apps, \\MMIC-FS1\WMIC_Apps\win64 (32-bit/64-bit Windows binaries) |
| Version | 3.4 |
Description: ima2obj is a simple command line tool to convert image files into Analyze Object Map files. The input image files can be of any format that can be read by the AVW library. However, each file must contain a single volume (frame). If the data type is not AVW_UNSIGNED_CHAR, then the image is first converted into an unsigned 8-bit integer image. If the command line option -r is specified, then all unused objects are removed from the Object Map file.
Usage: ima2obj [-o] [-r] [-y] [-n 1...255] file[s]
-o overwrite ON (default: if the output file `file.obj'
already exists, it will not be overwritten)
-r remove all the unused objects. This is any object
which does not have at least one corresponding voxel
in the object map.
-y y flip. All input image volumes will be flipped
vertically.
-n 1...255 the number of objects to create. Every Object Map
has one background object "Original" representing
all voxels with a value of zero.
By default, the number of objects created is equal
to the maximum voxel value found in the image file.
file[s] the name[s] of the image file[s] to be converted.
They can be of any file format that can be opened
by AVW_OpenImageFile(). However, each file must
not contain more than a single volume. If the data
type is not AVW_UNSIGNED_CHAR, then the image is first
converted into an unsigned 8-bit integer image.
Input file : "h00116_mask.hdr" Data format = 4 AnalyzeImage(7.5) Data type = 1 AVW_UNSIGNED_CHAR NumVols = 1 Width = 181 Height = 217 Depth = 181 Voxel width = 1 mm Voxel height= 1 mm Voxel depth = 1 mm glmin = 0 glmax = 1 Output file : "h00116_mask.obj" File format = OBJMAP, version 20050829 NumObjects = 2 ============================================================================= Input file : "h00116_atlas.hdr" Data format = 46 NIFTI Data type = 1 AVW_UNSIGNED_CHAR NumVols = 1 Width = 135 Height = 175 Depth = 130 Voxel width = 1.21875 mm Voxel height= 1.21875 mm Voxel depth = 1.21875 mm glmin = 0 glmax = 83 Output file : "h00116_atlas.obj" File format = OBJMAP, version 20050829 NumObjects = 84 ============================================================================= Input file : "h00116_test.avw" Data format = 0 AnalyzeAVW Data type = 1 AVW_UNSIGNED_CHAR NumVols = 1 Width = 120 Height = 160 Depth = 120 Voxel width = 1 mm Voxel height= 1 mm Voxel depth = 1 mm glmin = 0 glmax = 5 Output file : "h00116_test.obj" File format = OBJMAP, version 20050829 NumObjects = 6
The image c2h01293_pam.hdr
was the product of the thresholded probabilistic white matter map (class 2)
generated with SPM and the 83-region brain atlas warped into individual space
(similar to this example).
However, the output file created by
mulimg
had a global maximum value of only 82 and not 83 indicating that there
was no surviving voxel with the value 83 corresponding to the structure
"G_sup_temp_ant_r" in the 83-region brain atlas.
Therefore
ima2obj needs to be
called with the command line option
-n 83
to create an Object Map with 83 objects:
Input file : "c2h01293_pam.hdr"
Data format = 4 AnalyzeImage(7.5)
Data type = 1 AVW_UNSIGNED_CHAR
NumVols = 1
Width = 130
Height = 140
Depth = 124
Voxel width = 1.21875 mm
Voxel height= 1.21875 mm
Voxel depth = 1.21875 mm
glmin = 0
glmax = 82
--> 83 Objects will be added.
Output file : "c2h01293_pam.obj"
File format = OBJMAP, version 20050829
NumObjects = 84
The command showobjn can now be called to double check the number of objects in the Object Map file:
% showobjn -v c2h01293_pam.obj
Object Map : "c2h01293_pam.obj"
Version = 20050829
Width = 130
Height = 140
Depth = 124
NumObjects = 84
NumVoxels = 2256800
1: " Original", 2064819 voxels, R=255, G=255, B=255
2: " Object_2", 0 voxels, R=255, G= 0, B= 0
3: " Object_3", 16 voxels, R= 0, G=255, B= 0
4: " Object_4", 0 voxels, R=255, G=255, B= 0
5: " Object_5", 0 voxels, R= 0, G= 0, B=255
6: " Object_6", 0 voxels, R=255, G= 0, B=255
7: " Object_7", 637 voxels, R= 0, G=255, B=255
77: " Object_77", 259 voxels, R= 70, G=255, B=255
78: " Object_78", 272 voxels, R=185, G=185, B=185
79: " Object_79", 5 voxels, R=255, G= 77, B= 77
80: " Object_80", 0 voxels, R= 77, G=255, B= 77
81: " Object_81", 50 voxels, R=255, G=255, B= 77
82: " Object_82", 40 voxels, R= 77, G= 77, B=255
83: " Object_83", 229 voxels, R=255, G= 77, B=255
84: " Object_84", 0 voxels, R= 77, G=255, B=255
As expected, "Object_84" is listed as an empty object with 0 voxels
at the end of the list.
Now,
panames can be called to label the objects according to
the 83-region Probabilistic Brain Atlas:
Object Map file : c2h01293_pam.obj Number of objects: 84 object names succesfully changed using atlas 4.
Another call of showobjn finally confirms that the empty object "G_sup_temp_ant_r" with 0 voxels exists at the end of the list:
% showobjn -v c2h01293_pam.obj
Object Map : "c2h01293_pam.obj"
Version = 20050829
Width = 130
Height = 140
Depth = 124
NumObjects = 84
NumVoxels = 2256800
1: " Original", 2064819 voxels, R=255, G=255, B=255
2: " Hippocampus_r", 0 voxels, R=255, G= 0, B= 0
3: " Hippocampus_l", 16 voxels, R= 0, G=255, B= 0
4: " Amygdala_r", 0 voxels, R=255, G=255, B= 0
5: " Amygdala_l", 0 voxels, R= 0, G= 0, B=255
6: " Ant_TL_med_r", 0 voxels, R=255, G= 0, B=255
7: " Ant_TL_med_l", 637 voxels, R= 0, G=255, B=255
77: " Subgen_antCing_l", 259 voxels, R= 70, G=255, B=255
78: " Subgen_antCing_r", 272 voxels, R=185, G=185, B=185
79: " Subcall_area_l", 5 voxels, R=255, G= 77, B= 77
80: " Subcall_area_r", 0 voxels, R= 77, G=255, B= 77
81: " Presubgen_antCing_l", 50 voxels, R=255, G=255, B= 77
82: " Presubgen_antCing_r", 40 voxels, R= 77, G= 77, B=255
83: " G_sup_temp_ant_l", 229 voxels, R=255, G= 77, B=255
84: " G_sup_temp_ant_r", 0 voxels, R= 77, G=255, B=255
In fact, there are several other empty objects ("Hippocampus_r", "Amygdala_r", "Amygdala_l", "Ant_TL_med_r" and "Subcall_area_r") shown in this part of the screen output. Each empty object creates a tacstats warning message similar to that shown in examples 1 and 2 below:
A call of tacstats with the command line switch -w suppresses this warning message and inserts a line with NaN (not a number) entries in the output for each of the empty objects as shwon in example 3 below.
| Type | binary executable, AVW library required |
| Location |
/opt/wmic/bin (Solaris/Linux wrapper script) \\MMIC_FS1\WMIC_Apps, \\MMIC-FS1\WMIC_Apps\win64 (32-bit/64-bit Windows binaries) |
| Version | 1.4 |
Description: showobjn is a
command line program to display the number of objects in an Analyze Object Map
file.
If the verbosity flag -v is set,
a list of the objects with name string, voxel numbers and RGB values is printed.
Usage: showobjn [-v|l] file[s]
-v verbose: list the names and voxel numbers of the individual objects
-l long format: list full attributes of the individual objects
(default: only the total number of objects and voxels is reported)
file[s] the name[s] of the Analyze Object Map file[s]
Object Map : "h00116.obj" Version = 20050829 Width = 120 Height = 160 Depth = 120 NumObjects = 6 NumVoxels = 2304000 ================================================================================ Object Map : "t00799.obj" Version = 910926 Width = 128 Height = 128 Depth = 63 NumObjects = 10 NumVoxels = 1032192
Object Map : "h00116.obj"
Version = 20050829
Width = 120
Height = 160
Depth = 120
NumObjects = 6
1: " Original", 2285660 voxels, R=255, G=255, B=255
2: " cerebellum", 8816 voxels, R=255, G= 0, B= 0
3: " thalamus_r", 1169 voxels, R= 0, G=255, B= 0
4: " thalamus_l", 1312 voxels, R=255, G=255, B= 0
5: " soft", 3305 voxels, R= 0, G= 0, B=255
6: " bone", 3738 voxels, R=255, G= 0, B=255
NumVoxels = 2304000
================================================================================
Object Map : "t00799.obj"
Version = 910926
Width = 128
Height = 128
Depth = 63
NumObjects = 10
1: " Original", 992754 voxels, R=255, G=255, B=255
2: " frontal_r", 3862 voxels, R=255, G= 0, B= 0
3: " frontal_l", 3804 voxels, R= 0, G=255, B= 0
4: " temppar_r", 8572 voxels, R=255, G=255, B= 0
5: " temppar_l", 8400 voxels, R= 0, G= 0, B=255
6: " occipit_r", 6236 voxels, R=255, G= 0, B=255
7: " occipit_l", 5583 voxels, R= 0, G=255, B=255
8: " cerebel_r", 1535 voxels, R=255, G=255, B=255
9: " cerebel_l", 1446 voxels, R=255, G= 7, B= 7
10: " brainstem", 0 voxels, R= 7, G=255, B= 7
NumVoxels = 1032192
Note: Object 10 "brainstem" is an entry in the object list, however, it actually does not contain any voxel.
| Type | binary executable, AVW library required |
| Location |
/opt/wmic/bin (Solaris/Linux wrapper script) \\MMIC_FS1\WMIC_Apps, \\MMIC-FS1\WMIC_Apps\win64 (32-bit/64-bit Windows binaries) |
| Version | 1.4 |
Description: The object name copy program
objncopy is a small utility
that transfers the object name list from a source Analyze Object
Map file into destination Object Map file(s). As an example, this is useful
after an Analyze Image
file has been converted with ima2obj into an
Analyze Object Map file and the object name list has automatically been
set to the default of "Object_1", "Object_2", "Object_3" etc. This happens,
for instance, after an Object Map originally defined on the MR image has
been resliced from the subject's MRI space into the subject's PET space.
Then objncopy can be used
to define the object name list of the resliced (destination) Object Map file
using the original object name list of the original (source) Object Map
file.
objncopy only works if the
numbers of objects in the source and the destination Object Map files are
the same. Otherwise, it exits with an error message and no change is performed.
Therefore, it is important that the original Object Map does not contain
any "empty" objects (see the example of the "brainstem"
object above), because these empty objects are removed when the Object
Map is saved as an
Analyze 7.5 image file.
Usage: objncopy source.obj dest1.obj [dest2.obj ...]
Copies the object names from the `sourc.obj' file into the `dest.obj' file if the numbers of objects in both Analyze Object Map files are equal. Otherwise nothing is changed. Multiple destination Object Map files can be specified on the command line.
Source file : "t00799.obj" Version = 910926 Width = 128 Height = 128 Depth = 63 NumObjects = 10 Destination fi: "t00799_resliced.obj" Version = 20050829 Width = 128 Height = 128 Depth = 63 NumObjects = 9 The number of objects must not differ between the files! ==> NOTHING CHANGED Exit 5
Source file : "t00799_original.obj"
Version = 20050829
Width = 128
Height = 128
Depth = 63
NumObjects = 9
Destination fi: "t00799_resliced.obj"
Version = 20050829
Width = 128
Height = 128
Depth = 63
NumObjects = 9
1: was "Original" is now "Original"
2: was "Object_2" is now "frontal_r"
3: was "Object_3" is now "frontal_l"
4: was "Object_4" is now "temppar_r"
5: was "Object_5" is now "temppar_l"
6: was "Object_6" is now "occipit_r"
7: was "Object_7" is now "occipit_l"
8: was "Object_8" is now "cerebel_r"
9: was "Object_9" is now "cerebel_l"
Modified Object Map "t00799_resliced.obj" succesfully saved.
| Type | binary executable, AVW library required |
| Location |
/opt/wmic/bin (Solaris/Linux wrapper script) \\MMIC_FS1\WMIC_Apps, \\MMIC-FS1\WMIC_Apps\win64 (32-bit/64-bit Windows binaries) |
| Version | 2.3 |
Description: panames
is a command line program which applies a predefined object name list
to an Analyze Object Map file. Currently, six such
predefined object name lists are implemented, with either 50, 51, 68, 74,
84 or 86 entries. When panames
opens an Object Map file, it checks whether the number of objects defined
in that file matches any of these numbers and writes the corresponding
object name list into that Object Map file. If the number of objects
does not match, then nothing is changed.
panames can be thought of as
a special case of the
objncopy utility that uses built in lists of object names instead
of reading the object name list from a source Object Map file.
References:
Hammers, A.; Koepp, M.J.; Free, S.L.; Brett, M.; Richardson, M.P.;
Labbé, C.; Cunningham, V.J.; Brooks, D.J.; Duncan, J.
Implementation and application of a brain template
for multiple volumes of interest.
Hum Brain Mapp 15 (2002), 165 - 174.
Hammers, A.; Allom, R.; Koepp, M.J.; Free, S.L.; Myers, R.; Lemieux, L.;
Mitchell, T.N.; Brooks, D.J.; Duncan, J.S.
Three-dimensional maximum probability atlas
of the human brain, with particular reference to the temporal lobe.
Hum Brain Mapp 19 (2003), 224 - 247
Hammers, A.; Chen, C.-H.; Lemieux, L.; Allom, R.; Vossos, S.; Free, S.L.;
Myers, R.; Brooks, D.J.; Duncan, J.S.; Koepp, M.J.
Statistical neuroanatomy of the human inferior frontal gyrus and
probabilistic atlas in a standard stereotaxic space.
Hum. Brain Mapp. 28 (2007), 34 - 48.
Rylands, A.J.; Hinz, R.; Jones, M.; Holmes, S.E.; Feldmann, M.;
Brown, G.; McMahon, A.W.; Talbot, P.S.
Pre- and postsynaptic serotonergic differences in males with
extreme levels of impulsive aggression without callous unemotional traits:
a PET study using 11C-DASB and 11C-MDL100907.
Biol. Psychiatry 72 (2012), 1004 - 1011.
Usage: panames [-p 0|1|2|3|4|5] filename[s]
-p n print the names of the objects in the probabilistic atlas `n' and exit
% panames -p 2
The following object names are defined in the Probabilistic Brain Atlas 2: Object 1: "Original" Object 2: "Hippocampus_r" Object 3: "Hippocampus_l" Object 4: "Amygdala_r" Object 5: "Amygdala_l" Object 6: "Ant_TL_med_r" Object 7: "Ant_TL_med_l" Object 8: "Ant_TL_lat_r" Object 9: "Ant_TL_lat_l" Object 10: "G_paraH_amb_r" Object 11: "G_paraH_amb_l" Object 12: "G_sup_temp_r" Object 13: "G_sup_temp_l" Object 14: "G_tem_midin_r" Object 15: "G_tem_midin_l" Object 16: "G_occtem_la_r" Object 17: "G_occtem_la_l" Object 18: "Cerebellum_r" Object 19: "Cerebellum_l" Object 20: "Brainstem" Object 21: "Insula_l" Object 22: "Insula_r" Object 23: "OL_rest_lat_l" Object 24: "OL_rest_lat_r" Object 25: "G_cing_ant_l" Object 26: "G_cing_ant_r" Object 27: "G_cing_post_l" Object 28: "G_cing_post_r" Object 29: "FL_mid_fr_G_l" Object 30: "FL_mid_fr_G_r" Object 31: "PosteriorTL_l" Object 32: "PosteriorTL_r" Object 33: "PL_rest_l" Object 34: "PL_rest_r" Object 35: "CaudateNucl_l" Object 36: "CaudateNucl_r" Object 37: "NuclAccumb_l" Object 38: "NuclAccumb_r" Object 39: "Putamen_l" Object 40: "Putamen_r" Object 41: "Thalamus_l" Object 42: "Thalamus_r" Object 43: "Pallidum_l" Object 44: "Pallidum_r" Object 45: "Corp_Callosum" Object 46: "FrontalHorn_r" Object 47: "FrontalHorn_l" Object 48: "TemporaHorn_r" Object 49: "TemporaHorn_l" Object 50: "ThirdVentricl" Object 51: "FL_precen_G_l" Object 52: "FL_precen_G_r" Object 53: "FL_strai_G_l" Object 54: "FL_strai_G_r" Object 55: "FL_orbitofr_l" Object 56: "FL_orbitofr_r" Object 57: "FL_inf_fr_G_l" Object 58: "FL_inf_fr_G_r" Object 59: "FL_sup_fr_G_l" Object 60: "FL_sup_fr_G_r" Object 61: "PL_postce_G_l" Object 62: "PL_postce_G_r" Object 63: "PL_sup_pa_G_l" Object 64: "PL_sup_pa_G_r" Object 65: "OL_ling_G_l" Object 66: "OL_ling_G_r" Object 67: "OL_cuneus_l" Object 68: "OL_cuneus_r"
% panames t00782.obj
Object Map file : t00782.obj Number of objects: 68 object names succesfully changed using atlas 2.
| Type | binary executable, AVW library required |
| Location |
/opt/wmic/bin (Solaris/Linux wrapper script) \\MMIC_FS1\WMIC_Apps, \\MMIC-FS1\WMIC_Apps\win64 (32-bit/64-bit Windows binaries) |
| Version | 1.2 |
Description: ima2obj
implements the two setp sequence "Load As" of an Object Map file (*.obj)
followed by "Save As" with the "Format" set to "AnalyzeImage(7.5)" in
the Analyze Main Workspace.
As the Analyze Object Map files do not carry information about their pixel
sizes, the pixel dimensions of the saved output images is set to the default
value of 1.0 mm as width, height and depth of the voxels.
By specifying an optional reference image the voxel width, height and depth
of the reference image can be transcribed into the output images. If the
reference image also contains a valid SPM orign, then this is transcribed
into the output images, too.
Usage: obj2ima [-o] [-r refer.hdr] file1.obj [file2.obj [file3.obj ...]]
-o overwrite ON, default: if the output files `filen.img'
or `filen.hdr' (n= 1, 2, ...) already exist, then they
will not be overwritten.
-r refer.hdr optional reference image in Analyze 7.5 Image format.
Without specifying this reference image, all output
images have the pixel dimensions of 1 mm and no SPM
origin set. If a reference image is given, then the
voxel width, height and depth and the SPM origin of
the output images are set to those read from the header
file of the reference image.
filen.obj Analyze Object Map file(s) to be converted (n= 1, 2,
...). The output files are all in Analyze 7.5 Image
format, data type AVW_UNSIGNED_CHAR (unsigned 8-bit
integer).
Reference file: "h00840_mr2pet.hdr" Data format = 4 AnalyzeImage(7.5) Data type = 4 DT_SIGNED_SHORT NumVols = 1 Width = 130 Height = 160 Depth = 124 Voxel width = -1.21875 mm Voxel height= 1.21875 mm Voxel depth = 1.21875 mm SPM Origin = [ 66, 96, 54] Object Map 1 : "wh00840_mask.obj" Version = 20050829 NumObjects = 2 Output file 1: "wh00840_mask.hdr" glmin = 0 glmax = 1
| Type | binary executable, AVW library required |
| Location |
/opt/wmic/bin (Solaris/Linux wrapper script) \\MMIC_FS1\WMIC_Apps, \\MMIC-FS1\WMIC_Apps\win64 (32-bit/64-bit Windows binaries) |
| Version | 2.13 |
Description: tacstats is a command
line program alternative to the Sample Images procedure in the Region Of
Interest module of
Analyze.
It saves a lot of time when time-activity curves are to be generated from
sampling dynamic images (multiple frames) with an Analyze Object Map.
It is also convenient to sample the series of single frame
Analyze files
which are created as the output of various parametric imaging programs.
The images to be sampled can be of data type 16-bit signed integer,
single precision (32-bit) float or double precision (64-bit)
floating point.
The tacstats
output in WMIC CSV format is printed on the
standard output stream. It can easily be piped through filter
programs and redirected into files. The format of the floating
point values can be configured via the
-f command line option.
As opposed to the STATS files created with the Region Of Interest module of
Analyze,
with tacstats it is possible to
incorporate information on the frame times and total counts in the header
section of the output. This requires to specify the Scan Information
File in WMIC CSV format of the dynamic image as the third command line
argument when invoking tacstats.
Caveat: tacstats does not issue a warning or error message if planes which are entirely zero (as often the outer planes of 3D images are) are sampled! This is likely to result in wrong results for minval, mean and stddev. The parameters skewness and excess derived from the third and fourth order sample central moments may be more sensitive to this problem and should therefore be carefully checked!
Let xj denote the voxel value (intensity) in the voxel of the ROI with the subscript j and let N be the total number of voxels in the ROI. Let the ordered set of voxel values be x1* ≤ x2* ≤ ... ≤ xN*. Then the following equations are used by tacstats:
Notes:
For the terminology positive skewness and negative skewness and examples,
refer to webpages such as Wikipedia
or Wolfram MathWorld.
For the terminology mesokurtic, leptokurtic and platykurtic and examples,
refer to webpages such as Wikipedia
or Wolfram MathWorld.
For the statistical median, refer to the the
Wolfram MathWorld
webpage and the excellent notes of
Nicolas Devillard.
The command line option -v can be used to investigate the distribution of the voxel values within a region of interest. Note, however, that the created output of comma-separated values (CSV) does not comply with the WMIC CSV file format guideline. Rather it is recommended to pipe the output through filter tools to create an output that can easily be imported into standard statistical analysis software.
Usage: tacstats [-f format] [-s] [-v] [-w] objectmap.obj dynamic.hdr [id_sif.csv] [> tac_stats.csv]
-f format a string specifying the output conversion of the
floating point values (default: %.6f)
See the Standard C Library Functions printf(3C)
manpage for examples.
-s summing ON (default: an output line for every
slice will be printed)
-v print the voxel values instead of performing the
statistics. This option can be used for subsequent
data analyis with statistical stand alone software.
-w alter the default behaviour of printing a message
"WARNING: object XY does not mask anything - skipped"
which results in no line for this object in the output.
If this warning message is switched off, then a
line with NaN (not a number) entries is inserted
in the output for each empty object encountered.
objectmap.obj the name of the Analyze Object Map file defining
the volumes of interest
dynamic.hdr the name of the dynamic image file in Analyze file format
to be sampled
id_sif.csv optional argument, the name of the Scan Information
File in CSV format
If given, the scan information will be transcribed as
CSV file metadata.
tac_stats.csv optional redirection of the output into a TAC
statistics file in CSV format
If omitted, the standard output will be used which
can be piped through filters (such as grep) for
reformatting.
% tacstats rct1.obj hirez_20060623_160920_s09_nf02.hdr hirez_20060623_160920_s09_nf02_sif.csv > h00005_stats.csv
WARNING: object 11 "11.Object" does not mask anything - skipped.The output has been saved in the file h00005_stats.csv. The information from the Scan Information File has been transcribed into the header section scan information.
WARNING: object 11 "11.Object" does not mask anything - skipped.The output has been saved in the file h00005s_stats.csv. The output is summed across image planes such that there is one line for each region and frame.
The output has been saved in the file h00005sw_stats.csv. In contrast to the previous example, no WARNING message about "11.Object" is printed. Rather this object is now included in the output file with a voxel number of 0 in the last column and all other numeric readouts set to NaN (not a number).
The voxel statistics file h00313c1r17_vstats.csv can be further processed with vstats2vec.
After the results are imported into spread sheet software, sometimes the question arises: What would the results be if two regions were already combined in the Analyze ROI tool and the results were calculated for that combined region?
Under the assumption that the two regions do not overlap, i.e. no
pixel is member of both regions, the following formulae may be used:
| N1... number of voxels region 1 | N2... number of voxels region 2 | N1+2... number of voxels regions 1 and 2 combined |
| mean1... mean value from region 1 | mean2... mean value from region 2 | mean1+2... mean value from regions 1 and 2 combined |
| var1... variance from region 1 | var2... variance from region 2 | var1+2... variance from regions 1 and 2 combined |
The Excel file t00813_Ki_rl.xls gives an example how the results for left and right regions combined can be calculated.
| Type | binary executable (deployed Matlab R2009b SP1 application) |
| Location | \\MMIC_FS1\WMIC_Apps, \\MMIC-FS1\WMIC_Apps\win64 (32-bit/64-bit Windows binaries) |
| Version | 4.1 |
Description:
This function loads from the current working directory all WMIC CSV
(*.csv) files. Each of the WMIC CSV STATS files must be produced from a
single frame image with the `tacstats -s' command to generate one line
of statistics output per region. It is a requirement that the list of
the region names is the same in all
the STATS files. This is usually the case if all STATS files were
generated with the same Analyze Object Map file, for example in template
space.
A a separate Worksheet in an Excel Workbook is inserted for each of the
outcome parameters 'maxval', 'minval', 'mean', 'stddev', 'skewness',
'excess' and 'median'. The first column of each Worksheet contains the
names of the volume files sampled. The two top rows of each Worksheet
consist of the names of the regions and the number of voxels per
region. If the number of voxels per region is not the same for all the
STATS files loaded, then an additional Worksheet 'voxels' is added to
the Excel Workbook.
The user is prompted to name an output file to save (default:
"output.xlsx"). If a file with the selected name already exists, a
dialogue requests confirmation that it can be replaced with the new file.
Note that the version of the saved Excel file depends on the Microsoft
Excel installed on the local machine and not on the Matlab version.
Optional input: A pattern for the regular expression to select the WMIC CSV files to read can be specified. The default pattern is '\w+\.csv$' i.e. all files with the extension '.csv' are selected. See Regular Expressions in the MATLAB Programming Fundamentals documentation for more information on how to match patterns. For example in the default pattern, '\w' matches any alphabetic, numeric, or underscore character and the quantifier '+' matches that character one or more times. The positional operator '$' defines that the file name must end with the four characters '.csv'.
Examples:
\\MMIC-FS1\WMIC_Apps\win64\stats2xls \w+_brain.*\.csv$ brain.xlsx
\\MMIC-FS1\WMIC_Apps\win64\stats2xls \w+_cereb.*\.csv$
The pattern '.*' matches any single character (including white space)
zero or more times.
Run in a DOS Command Window:
\\MMIC-FS1\WMIC_Apps\stats2xls11 files matching the pattern "\w+\.csv" found. 1 file "17015i_brain.csv" read. 2 file "17016i_brain.csv" read. 3 file "17017i_brain.csv" read. 4 file "17018i_brain.csv" read. 5 file "17019i_brain.csv" read. 6 file "17020i_brain.csv" read. 7 file "17021i_brain.csv" read. 8 file "17022i_brain.csv" read. 9 file "17023i_brain.csv" read. 10 file "17024i_brain.csv" read. 11 file "17025i_brain.csv" read. Processing data from 28 region(s) for each file... - done. Saving the output in "output.xlsx" ... 1 worksheet 'maxval' appended. 2 worksheet 'minval' appended. 3 worksheet 'mean' appended. 4 worksheet 'stddev' appended. 5 worksheet 'skewness' appended. 6 worksheet 'excess' appended. 7 worksheet 'median' appended.
| Type | GAWK or NAWK script |
| Location | /opt/wmic/bin |
| Version | 1.2 |
Description: This GAWK script reformats the voxel statistics
output from the
tacstats program and creates a vector file for each region in the
current working directory. The vector files are plain text files containing
one column with the (integer) voxel values separated by commas. These vector
files can easily be loaded into standard statistics software to perform
further analyses on the distribution of the voxel values.
The program
vec2xls can be used to load a cohort of vector files and
then write the data into one Microsoft Excel Workbook file.
Usage:
vstats2vec n01234_vstats.csv
or:
tacstats -v objectmap.obj image.hdr | vstats2vec -
1: " MesialTemp_r.vec" written: 1607 voxels.
2: " MesialTemp_l.vec" written: 935 voxels.
3: " TemporaLobe_r.vec" written: 27906 voxels.
4: " TemporaLobe_l.vec" written: 20902 voxels.
5: " Cerebellum_r.vec" written: 24895 voxels.
6: " Cerebellum_l.vec" written: 26466 voxels.
7: " Brainstem.vec" written: 4300 voxels.
8: " Insula_l.vec" written: 2037 voxels.
9: " Insula_r.vec" written: 2832 voxels.
10: " FrontalLobe_l.vec" written: 31774 voxels.
11: " FrontalLobe_r.vec" written: 35916 voxels.
12: " OccipitLobe_l.vec" written: 20105 voxels.
13: " OccipitLobe_r.vec" written: 21272 voxels.
14: " ParietalLob_l.vec" written: 25426 voxels.
15: " ParietalLob_r.vec" written: 28443 voxels.
16: " Central_l.vec" written: 3418 voxels.
17: " Central_r.vec" written: 3705 voxels.
NumVoxels = 281939
| Type | binary executable (deployed Matlab R2009b SP1 application) |
| Location | \\MMIC_FS1\WMIC_Apps, \\MMIC-FS1\WMIC_Apps\win64 (32-bit/64-bit Windows binaries) |
| Version | 3.2 |
Description:
This function loads from the current working directory all vector (*.vec)
files. These vector files are usually generated with the (N)AWK script
`vstats2vec' from a voxel statistics CSV file produced from a single
frame image with `tacstats -v'. They are plain text files with just a
single column of (integer) numbers.
Then these vectors are horizontally concatenated into a double matrix X.
To allow for vectors of variable length, the number of rows of X is
equal to the length of the longest vector. The other (shorter) vectors
are filled up with NaN (Not-a-Number) to the same length. The number of
columns of X is equal to the number of vector files loaded.
The cell array regions contains the names of the vector files loaded with
the '.vec' extension removed. It has one row and as many columns as X.
The cell array stat contains statistic output such as the numnber of
voxels n and the maximum data value, minimum data value, mean, standard
deviation, median and skewness for each data vector. Where applicable,
statistics are also provided for the combined left and right regions.
Output: A Microsoft Excel file 'output.xls' is written in the current working directory replacing any previous file with the same name that may already exist. It contains the regions name vector and the data matrix X.
Run in a DOS Command Window:
\\MMIC-FS1\WMIC_Apps\vec2xls
1: " Brainstem.vec" read: 4300 voxels.
2: " Central_l.vec" read: 3418 voxels.
3: " Central_r.vec" read: 3705 voxels.
4: " Cerebellum_l.vec" read: 26466 voxels.
5: " Cerebellum_r.vec" read: 24895 voxels.
6: " FrontalLobe_l.vec" read: 31774 voxels.
7: " FrontalLobe_r.vec" read: 35916 voxels.
8: " Insula_l.vec" read: 2037 voxels.
9: " Insula_r.vec" read: 2832 voxels.
10: " MesialTemp_l.vec" read: 935 voxels.
11: " MesialTemp_r.vec" read: 1607 voxels.
12: " OccipitLobe_l.vec" read: 20105 voxels.
13: " OccipitLobe_r.vec" read: 21272 voxels.
14: " ParietalLob_l.vec" read: 25426 voxels.
15: " ParietalLob_r.vec" read: 28443 voxels.
16: " TemporaLobe_l.vec" read: 20902 voxels.
17: " TemporaLobe_r.vec" read: 27906 voxels.
NumVoxels = 281939
"output.xls" written: 35930 rows x 18 columns.
| Type | binary executable, AVW library required |
| Location |
/opt/wmic/bin (Solaris/Linux wrapper script) \\MMIC_FS1\WMIC_Apps, \\MMIC-FS1\WMIC_Apps\win64 (32-bit/64-bit Windows binaries) |
| Version | 2.3 |
Description: decay
is a command line tool to either correct or uncorrect dynamic PET images
for radioactive decay relative to the start time of the scan.
This is done by either multiplying or dividing every voxel value of a frame
by the decay correction factor dc of the frame.
In the formula,
tframestart is the frame start time,
tframeend is the frame end time and
λ is the the radioactive decay constant.
The start and end times of the frames are retrieved from the
Scan Information File.
The following table links the half lives T1/2
with the decay constants λ for the isotopes most
commonly used in PET.
| Symbol | Half life T1/2 | Decay constant λ |
| decay-corrected data | infinity | 0 |
| 18F | 109.7 min | 0.0001053 s-1 |
| 11C | 20.4 min | 0.0005663 s-1 |
| 15O | 122.55 s | 0.005656 s-1 |
Usage: decay [-o] [-u] id_sif.csv dynamic.hdr altered.hdr
-o overwrite ON (default: if the output file `altered.hdr'
already exists, it will not be overwritten)
-u undecay `dynamic.hdr' (default: decay `dynamic.hdr')
id_sif.csv the name of the Scan Information File in CSV format
The old SIF file format is still supported.
dynamic.hdr the name of the dynamic image file in Analyze file
format to be [un]decayed
altered.hdr the name of the altered image file in Analyze Image 7.5
file format which will be created
Scan Information File : "h00005_20060623_160920_s15_nf29_sif.csv"
Number of frames in the scan: 29
Radioactive deacy : isotope= O-15, lambda= 0.005656 1/s, T???= 2.04251 min
Dynamic Image File : "h00005_20060623_160920_s15_nf29.hdr"
NumVols = 29
Width = 256
Height = 256
Depth = 81
Voxel width = 2.66615 mm
Voxel height= 2.66615 mm
Voxel depth = 2 mm
ROIScale = 0.0972374901175499
glmin = -28858
glmax = 32767
cal_min = -2806.0793
cal_max = 3186.1807
Frame frame start frame end undecay frame frame
number (s) (s) factor min max
1 0 140 1.44763288 -601 523
2 140 145 2.23881493 -8434 10600
3 145 150 2.30303237 -13008 14203
4 150 155 2.36909180 -15976 20896
5 155 160 2.43704607 -18523 15785
6 160 165 2.50694950 -13057 17024
7 165 170 2.57885803 -21830 20079
8 170 175 2.65282916 -20863 22907
9 175 180 2.72892205 -24369 28142
10 180 185 2.80719756 -26168 32767
11 185 190 2.88771831 -28858 24912
12 190 195 2.97054868 -25056 22895
13 195 200 3.05575494 -14920 18921
14 200 205 3.14340522 -20080 22271
15 205 210 3.23356963 -25854 28055
16 210 220 3.37335133 -16220 15410
17 220 230 3.56964700 -13225 14811
18 230 240 3.77736513 -18309 15483
19 240 250 3.99717040 -11845 12432
20 250 260 4.22976616 -13434 13802
21 260 280 4.60244132 -7414 8826
22 280 300 5.15365867 -10090 9807
23 300 320 5.77089329 -10667 7731
24 320 350 6.64297995 -7002 6608
25 350 380 7.87143710 -5285 5035
26 380 410 9.32706744 -3827 4407
27 410 440 11.05188111 -3352 4916
28 440 470 13.09565702 -3458 4308
29 470 500 15.51737944 -2976 3555
Reading the dynamic image and calculating the altered undecayed image
Volume 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, done.
Altered Image File : "h00005_20060623_160920_s15_nf29_dc.hdr"
NumVols = 29
Width = 256
Height = 256
Depth = 81
Voxel width = 2.66615 mm
Voxel height= 2.66615 mm
Voxel depth = 2 mm
ROIScale = 0.2729648649692535
glmin = -29781
glmax = 32767
cal_min = -8129.1666
cal_max = 8944.2397
| Type | binary executable, AVW library required |
| Location |
/opt/wmic/bin (Solaris/Linux wrapper script) \\MMIC_FS1\WMIC_Apps, \\MMIC-FS1\WMIC_Apps\win64 (32-bit/64-bit Windows binaries) |
| Version | 2.5 |
Description: addimg
is an efficient implementation (written in the programming language
C) to
create add images which can be corrected for radioactive decay
(command line option: -d).
See the description of the command
decay for further
details on the decay correction factors used.
Because addimg processes
the dynamic image slice by slice, it achieves a better
memory locality than many other programs which require the whole
dynamic data set to be loaded. This is a particular advantage for data
from the HRRT scanner which can be bigger than one gigabyte for a dynamic
image sequence with 37 frames.
A Scan Information File is mandatory. Add images will be calculated
correctly for noncontiguous frames.
The data can be devided by an activity value A0 (option:
-a) such that ratio images
are generated.
Usage: addimg [-a A0] [-d] [-f first_frame] [-l last_frame] [-o] id_sif.csv dynamic.hdr addimage.hdr
-a A0 divide the add image by the (constant) activity value A0
defined at scan start time of the Scan Information File
This option can be used to create ratio images of apparent
volumes of distribution or apparent binding potentials.
-d decay correction ON (default: no decay correction will
be applied to the data)
-f first_frame the number of the first frame (default: 1)
-l last_frame the number of the last frame (default: the last frame
of the scan)
-o overwrite ON (default: if the output files `addimage.hdr'
or `addimage.img' already exist, then they will not
be overwritten)
id_sif.csv the name of the Scan Information File in CSV format
The old SIF file format is still supported.
dynamic.hdr the name of the dynamic image file in Analyze file format
to be added
addimage.hdr the name of the add image file in Analyze file format
which will be created
Create an add image from 35 min (2100 s) to 55 min (3300 s) post injection.
Correct for radioactive decay back to the start of the emission scan.
Here it is assumed that the intravenous infusion of [18F]FDG commenced
440 s after the start of the emission scan.
Scan Information File : "image_10_sif.csv"
Number of frames in the scan: 33
Number of the first frame : 29
Number of the last frame : 32
Radioactive deacy : isotope= F-18, lambda= 0.0001053 1/s, T1/2= 109.71 min
Frame frame start frame end decay correc-
number (s) (s) tion factor
1 0 450 1.0238796
2 450 455 1.04880166
3 455 460 1.049354
4 460 465 1.04990663
5 465 470 1.05045955
6 470 475 1.05101277
7 475 480 1.05156627
8 480 485 1.05212007
9 485 490 1.05267415
10 490 495 1.05322853
11 495 500 1.0537832
12 500 505 1.05433817
13 505 510 1.05489342
14 510 520 1.05572681
15 520 530 1.05683908
16 530 560 1.05906674
17 560 590 1.06241762
18 590 620 1.0657791
19 620 680 1.07083995
20 680 740 1.07762693
21 740 800 1.08445693
22 800 920 1.09477772
23 920 1040 1.1086991
24 1040 1340 1.13344925
25 1340 1640 1.16982647
26 1640 1940 1.20737118
27 1940 2240 1.24612087
28 2240 2540 1.28611419
29 2540 2840 1.32739108 *** first frame ***
30 2840 3140 1.36999271
31 3140 3440 1.41396161
32 3440 3740 1.45934167 ### last frame ###
33 3740 4040 1.50617816
Dynamic Image File : "image_10.hdr"
NumVols = 33
Width = 256
Height = 256
Depth = 207
Voxel width = 1.21875 mm
Voxel height= 1.21875 mm
Voxel depth = 1.21875 mm
ROIScale = 0.2228407561779022
glmin = 0
glmax = 32767
cal_min = 0
cal_max = 7301.8232
Reading the dynamic image and calculating the add image
Slice 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, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, done.
Add Image File : "image_10_35to55min.hdr"
NumVols = 1
Width = 256
Height = 256
Depth = 207
Voxel width = 1.21875 mm
Voxel height= 1.21875 mm
Voxel depth = 1.21875 mm
ROIScale = 0.0043513062410057
glmin = 0
glmax = 32767
cal_min = 0
cal_max = 142.57925
| Type | binary executable, AVW library required |
| Location |
/opt/wmic/bin (Solaris/Linux wrapper script) \\MMIC_FS1\WMIC_Apps, \\MMIC-FS1\WMIC_Apps\win64 (32-bit/64-bit Windows binaries) |
| Version | 1.3 |
Description: mulimg
can conveniently be used to convert an image
input1 into a binary mask
and then applying this mask to another image
input2 saving the result as
output.
Currently, input2 is limited
to a single volume image, multi frame dynamic images are not yet supported.
Usage: mulimg [-l lower_thresh] [-o] [-u upper_thresh] input1 input2 output
input1 The first image file which will be converted into a
binary image by setting all of the voxels greater than or
equal to the lower_thresh and less than or equal to
the upper_thresh to 1 and the rest of the voxels to 0.
input2 The second image file.
Both image1 and image2 can be of any file format that can
be opened by AVW_OpenImageFile(). However, they each must
not contain more than a single volume.
output The name of the output file in Analyze Image 7.5 format
which will be created. The data type of output will be
the same as of input2. Equation:
output= ((input1 >= lower_thresh) & (input1 <= upper_thresh)) * input2
-l lower_thresh A scalar value defining the lower threshold for input1
(default: lower_thresh= 0.5).
-o overwrite ON (default: If the output file already exists,
it will not be overwritten.)
-l upper_thresh A scalar value defining the upper threshold for for input1
(default: upper_thresh= 127.0).
Apply an 80% threshold to the probabilistic grey matter map (class 1) generated with SPM5 and multiply with the 83-region brain atlas.
% mulimg -l 0.8 c1h00311_mr2pet.hdr wh00311_pa.hdr c1h00311_paInput1 file : "c1h00311_mr2pet.hdr" Data format = 46 NIFTI Data type = 64 AVW_FLOAT NumVols = 1 Width = 120 Height = 160 Depth = 124 Voxel width = 1.21875 mm Voxel height= 1.21875 mm Voxel depth = 1.21875 mm glmin = 0 glmax = 1.00011 Input1 thresholded: Lower thresh= 0.8 Upper thresh= 127 Data type = 1 AVW_UNSIGNED_CHAR VoxelsPerVol= 2380800 MaskedVoxels= 290960 Input2 file : "wh00311_pa.hdr" Data format = 46 NIFTI Data type = 1 AVW_UNSIGNED_CHAR NumVols = 1 Width = 120 Height = 160 Depth = 124 Voxel width = 1.21875 mm Voxel height= 1.21875 mm Voxel depth = 1.21875 mm glmin = 0 glmax = 83 Output file : "c1h00311_pa" Data type = 1 AVW_UNSIGNED_CHAR Data format = 4 AnalyzeImage(7.5) NumVols = 1 Width = 120 Height = 160 Depth = 124 Voxel width = 1.21875 mm Voxel height= 1.21875 mm Voxel depth = 1.21875 mm glmin = 0 glmax = 83
The output image c1h00311_pa.hdr can then be converted into an Object Map file using ima2obj and properly named with panames.
| Type | binary executable, AVW library required |
| Location |
/opt/wmic/bin (Solaris/Linux wrapper script) \\MMIC_FS1\WMIC_Apps, \\MMIC-FS1\WMIC_Apps\win64 (32-bit/64-bit Windows binaries) |
| Version | 2.5 |
Description: normi
is a command line tool to normalise dynamic PET images. Dynamic PET images
are normalised by subtracting from each frame its mean and dividing it by
its standard deviation.
Normalised images are needed to perform certain operations independently of
the absolute scale of the image. An example is cluster analysis which
aims at classifying time series by using similarity measures.
During the run normi
generates an output table with the minimum voxel value, the maximum voxel
value, the mean voxel value and the standard deviation of the voxel
values for each frame of the dynamic image sequence.
If the -h option is specified,
a histogram of the voxel values is saved in the binary file
voxhist.dat in the current
directory.
Using the -m option,
a mask image can be loaded. The lower and upper thresholds for the
conversion of the mask image into a binary mask can be set with the
-l and
-u options, respectively.
Usage: normi [-a] [-h] [-l lower_thresh] [-m mask] [-o] [-u upper_threshd] dynamic output
-a Calculate all voxels in a masked volume (default:
only the masked voxels will be calculated and all other
voxels are set to zero in the output image). Warning:
This option is only provided for compatibility and it
may cause data overflow!
-h histogram ON. If this flag is set, a histogram file
`voxhist.dat' will be saved to the disk which contains
voxel histograms (32-bit unsigned integer values) for
each frame of the dynamic.
-l lower_thresh A scalar value defining the lower threshold for the
optional mask image (default: lower_thresh= 0.5).
-m mask the name of an optional mask image file. If no
mask file is specified, then all voxels will be
normalised (the default).
-o overwrite ON (default: if either of the .hdr or .img
output files already exist, they will not be
overwritten)
-u upper_thresh A scalar value defining the upper threshold for the
optional mask image (default: upper_thresh= 3.40282e+38).
dynamic the name of the dynamic image file to be normalised.
Data type must be 16-bit signed integer.
output the name of the normalised image to be created.
Data type will be 16-bit signed integer, file format
will be AnalyzeImage(7.5).
Dynamic PET images are normalised by subtracting from each frame its
mean and dividing it by its standard deviation.
If an optional mask image is specified, then a binary mask is calculated
binary_mask= (mask_image >= lower_thresh) & (mask_image <= upper_thresh)
and applied to select the voxels to include in the calculation of 'frame
mean' and 'frame std dev'.
This dynamic image has been obtained from iterative image reconstruction. Therefore, there are no negative values in the image, and the distribution of the voxel values is positively skewed.
% normi -h h00094.hdr normi22.hdr
Dynamic Image : "h00094.hdr"
Data format = 4 AnalyzeImage(7.5)
Data type = 8 AVW_SIGNED_SHORT
NumVols = 18
Width = 130
Height = 168
Depth = 126
Voxel width = 1.21875 mm
Voxel height= 1.21875 mm
Voxel depth = 1.21875 mm
ROIScale = 0.0126550681889057
glmin = 0
glmax = 32767
cal_min = 0
cal_max = 414.66861
MaskedVoxels= 2751840
Frame frame frame frame frame normalised normalised
number min max mean std dev min max
1 0 104 0.030940 0.517831 -0.059749 200.777842
2 0 12033 126.659932 316.673485 -0.399970 37.598159
3 0 32767 258.966779 682.848639 -0.379245 47.606499
4 0 20002 354.607806 752.219985 -0.471415 26.119211
5 0 11638 317.900548 526.445300 -0.603862 21.502898
6 0 8452 262.985571 377.996232 -0.695736 21.664275
7 0 7348 235.643544 340.028644 -0.693011 20.916933
8 0 6574 224.081126 323.560196 -0.692548 19.625155
9 0 7421 213.315857 309.648351 -0.688897 23.276998
10 0 4467 189.161766 226.948224 -0.833502 18.849402
11 0 4519 156.171273 188.513314 -0.828436 23.143345
12 0 3843 125.902737 154.463503 -0.815097 24.064567
13 0 2533 99.907360 125.660566 -0.795057 19.362420
14 0 2240 80.318421 104.083982 -0.771669 20.749414
15 0 2015 64.762914 86.259918 -0.750788 22.608845
16 0 1451 52.405403 71.646261 -0.731446 19.520832
17 0 1033 39.330155 51.149534 -0.768925 19.426763
18 0 711 26.604856 36.358061 -0.731746 18.823753
Histogram matrix "voxhist.dat" saved: 18 volumes with 65536 histogram bins (uint32) each.
Reading the dynamic image and calculating the normalised image
Volume 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, done.
Output Image : "normi22.hdr"
Data format = 4 AnalyzeImage(7.5)
Data type = 8 AVW_SIGNED_SHORT
NumVols = 18
Width = 130
Height = 168
Depth = 126
Voxel width = 1.21875 mm
Voxel height= 1.21875 mm
Voxel depth = 1.21875 mm
ROIScale = 0.006127440836280584
glmin = -136
glmax = 32767
cal_min = -0.83333195
cal_max = 200.77785
fid= fopen('voxhist.dat', 'rb');
hist= fread( fid, '*uint32');
fclose(fid);
hist= reshape( hist, 65536, 18); % 18 frames in total
plot(-32768:32767, hist(:,3)) % plot the histogram of the 3rd frame
plot(-32768:32767, sum(hist,2)) % plot the histogram of the whole dynamic
This example shows how a floating point image that represents probabilities for each voxel to belong to tissue class 1 (c1) is used as mask. The default threshold values apply for the conversion of the mask image into a binary mask.
% normi -m c1h00094_mr2pet.hdr h00094.hdr h00094_c1.hdr
Dynamic Image : "h00094.hdr"
Data format = 4 AnalyzeImage(7.5)
Data type = 8 AVW_SIGNED_SHORT
NumVols = 18
Width = 130
Height = 168
Depth = 126
Voxel width = 1.21875 mm
Voxel height= 1.21875 mm
Voxel depth = 1.21875 mm
ROIScale = 0.0126550681889057
glmin = 0
glmax = 32767
cal_min = 0
cal_max = 414.66861
Mask Image : "c1h00094_mr2pet.hdr"
Data format = 46 NIFTI
Data type = 64 AVW_FLOAT
NumVols = 1
Width = 130
Height = 168
Depth = 126
Voxel width = 1.21875 mm
Voxel height= 1.21875 mm
Voxel depth = 1.21875 mm
glmin = 0
glmax = 1.00011
Mask Image thresholded:
Lower thresh= 0.5
Upper thresh= 3.40282e+38
Data type = 1 AVW_UNSIGNED_CHAR
VoxelsPerVol= 2751840
MaskedVoxels= 384101
Frame frame frame frame frame normalised normalised
number min max mean std dev min max
1 0 91 0.031138 0.622275 -0.050038 146.187548
2 0 8439 292.406458 412.957301 -0.708079 19.727448
3 0 20455 576.535154 915.076042 -0.630041 21.723293
4 0 15446 797.909854 940.602859 -0.848296 15.573087
5 1 10421 732.788444 619.943415 -1.180412 15.627574
6 5 4722 613.718592 429.525821 -1.417187 9.564690
7 2 4624 540.455005 393.101172 -1.369762 10.388025
8 2 4800 506.328429 373.053254 -1.351894 11.509541
9 2 4444 474.912856 354.905986 -1.332502 11.183489
10 11 2124 410.395232 192.085905 -2.079253 8.921033
11 4 1664 310.230603 157.772481 -1.940963 8.580517
12 3 1381 229.109992 128.650644 -1.757550 8.953628
13 1 1373 168.198344 104.606418 -1.598356 11.517474
14 1 1195 125.893031 86.381660 -1.445828 12.376550
15 1 859 95.722576 71.784451 -1.319542 10.632907
16 0 860 73.452956 60.398240 -1.216144 13.022682
17 1 486 53.180627 38.684988 -1.348860 11.188303
18 0 428 34.808772 29.142094 -1.194450 13.492209
Reading the dynamic image and calculating the normalised image
Volume 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, done.
Output Image : "h00094_c1.hdr"
Data format = 4 AnalyzeImage(7.5)
Data type = 8 AVW_SIGNED_SHORT
NumVols = 18
Width = 130
Height = 168
Depth = 126
Voxel width = 1.21875 mm
Voxel height= 1.21875 mm
Voxel depth = 1.21875 mm
ROIScale = 0.004461425822228193
glmin = -466
glmax = 32767
cal_min = -2.0790244
cal_max = 146.18754
As the default values for thresholding the mask images are floating
point numbers, it is crucial to manually set with the
-l and
-u options values
that are representable by the actual data type of the particular
mask image.
In this example a 16-bit signed integer image is used as mask image.
Hence the lower threshold is set to
1, and the upper
threshold is set to 32767.
Dynamic Image : "h00094.hdr"
Data format = 4 AnalyzeImage(7.5)
Data type = 8 AVW_SIGNED_SHORT
NumVols = 18
Width = 130
Height = 168
Depth = 126
Voxel width = 1.21875 mm
Voxel height= 1.21875 mm
Voxel depth = 1.21875 mm
ROIScale = 0.0126550681889057
glmin = 0
glmax = 32767
cal_min = 0
cal_max = 414.66861
Mask Image : "h00094_MASK.hdr"
Data format = 4 AnalyzeImage(7.5)
Data type = 8 AVW_SIGNED_SHORT
NumVols = 18
Width = 130
Height = 168
Depth = 126
Voxel width = 1.21875 mm
Voxel height= 1.21875 mm
Voxel depth = 1.21875 mm
glmin = 0
glmax = 32767
Mask Image thresholded:
Lower thresh= 1
Upper thresh= 32767
Data type = 1 AVW_UNSIGNED_CHAR
VoxelsPerVol= 2751840
MaskedVoxels= 1359806
Frame frame frame frame frame normalised normalised
number min max mean std dev min max
1 0 104 0.034836 0.612600 -0.056866 169.711354
2 0 12033 237.877951 417.139866 -0.570259 28.276180
3 0 32767 474.067982 896.290813 -0.528922 36.029525
4 0 20002 652.117841 953.783184 -0.683717 20.287506
5 0 11638 590.802654 622.792046 -0.948636 17.738180
6 1 8452 491.344076 417.290909 -1.175065 19.076993
7 1 7348 440.917790 376.456739 -1.168575 18.347612
8 1 6574 418.291538 358.798592 -1.163024 17.156445
9 0 7421 398.003100 344.471547 -1.155402 20.387742
10 5 4467 355.107350 214.922929 -1.628990 19.131940
11 3 4519 293.136061 180.174594 -1.610305 23.454272
12 2 3843 235.087729 151.303000 -1.540536 23.845610
13 0 2533 185.407572 126.962241 -1.460336 18.490477
14 1 2240 148.117177 108.532391 -1.355514 19.274272
15 1 2015 118.779866 92.045828 -1.279579 20.600827
16 0 1451 95.643434 77.941281 -1.227122 17.389457
17 0 1033 71.491167 53.961845 -1.324847 17.818309
18 0 711 48.320360 39.647822 -1.218739 16.714150
Histogram matrix "voxhist.dat" saved: 18 volumes with 65536 histogram bins (uint32) each.
Reading the dynamic image and calculating the normalised image
Volume 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, done.
Output Image : "h00094_normi22.hdr"
Data format = 4 AnalyzeImage(7.5)
Data type = 8 AVW_SIGNED_SHORT
NumVols = 18
Width = 130
Height = 168
Depth = 126
Voxel width = 1.21875 mm
Voxel height= 1.21875 mm
Voxel depth = 1.21875 mm
ROIScale = 0.005179337225854397
glmin = -319
glmax = 32767
cal_min = -1.6522086
cal_max = 169.71134
| Type | script using /usr/bin/gawk, the GNU Project's implementation of the pattern scanning and processing language AWK |
| Location | /opt/wmic/bin |
| Version | 1.2 |
Description: hdr2sif, a simple script written in the pattern scanning and processing language AWK, reformats the contents of the Interfile header file created by the Inveon image reconstruction software. The output is formatted in the WMIC's comma-separated value (CSV) standard and is normally redirected into a Scan Information File (SIF).
Example: hdr2sif a00760.pet.img.hdr > a00760_sif.csv
| Type | binary executable (deployed Matlab R2009b SP1 application) |
| Location |
/data/secondary/rhinz/bin (Solaris/Linux wrapper script) \\MMIC_FS1\WMIC_Apps, \\MMIC-FS1\WMIC_Apps\win64 (32-bit/64-bit Windows binaries) |
| Version | 1.1 |
Description: inveon_i2ana converts the dynamic binary interfile created by the Inveon image reconstruction software (*.pet.img) into an Analyze 7.5 Image file pair .hdr/.img. To calibrate the image data, it requires the Scan Information File (SIF) in WMIC CSV format of the scan, typically generated with hdr2sif. The output image is in default radiological orientation, calibrated in kBq/ml not corrected for radioactive decay.
Based on the data read from the SIF, each frame of the dynamic image is scaled as follows.
The output image is saved as 32-bit single precision float data. Use the command sthg2ana to further convert the output image into 16-bit signed integer data for the subsequent use by other tools such as addimg or tacstats.
Usage: inveon_i2ana siffile iname oname
siffile A string containing the name of a Scan Information File
(SIF) in WMIC CSV format. Typically, this file is generated
from the Inveon dynamic interfile header (*.pet.img.hdr)
with the command `hdr2sif'.
iname A string containing the name of the Inveon dynamic binary
interfile (*.pet.img) to be converted.
oname A string defining the path for the Analyze 7.5 Image files
to be written. The file name extensions `.hdr' and `.img'
will be appended to the oname string. If the output files
(hdr/img) already exist, then an attempt will be made to
overwrite them.
Output:
The Analyze 7.5 Image file pair byte order (big endian or
little endian) is that of the machine on which the program
is run. The data type is 16 (32-bit single precision float).
Use the command `sthg2ana' to further convert the output
image into data type 4 (16-bit signed integer).
Scan Informat. File: "a00760_sif.csv"
Scan Date : "2011-06-14"
Scan Time : "11:09:46"
Number Of Time Sl. = 21
Original Width = 128
Original Height = 128
Original Depth = 159
Voxel Width = 0.776383 mm
Voxel Height = 0.776383 mm
Voxel Depth = 0.796 mm
Isotope : "C-11" (bf= 0.998)
....................... Z flip ..... Y flip ........................
..... calibrating image data ..... into kBq/ml ..... decaying ......
Frame frame start frame end frame Decay prompts delayeds
number (s) (s) length Factor
(s)
1 0 30 30 1.00852 2923135 371251
2 30 60 30 1.02582 7985292 1392574
3 60 90 30 1.04341 9079716 1571832
4 90 120 30 1.0613 9318788 1563466
5 120 150 30 1.0795 9422925 1497454
6 150 180 30 1.09801 9538119 1460007
7 180 240 60 1.12633 18874475 2804010
8 240 300 60 1.16528 18334568 2586932
9 300 360 60 1.20559 17795986 2392233
10 360 480 120 1.26849 33874628 4362698
11 480 600 120 1.35775 31784337 3805424
12 600 720 120 1.4533 29794825 3301422
13 720 840 120 1.55557 28275999 2899854
14 840 960 120 1.66504 26270435 2584236
15 960 1080 120 1.78221 24943377 2253318
16 1080 1200 120 1.90763 23654348 1961143
17 1200 1500 300 2.14654 52573012 3943790
18 1500 1800 300 2.54434 45036378 2854052
19 1800 2400 600 3.27162 71418000 3514981
20 2400 3000 600 4.59659 51880015 1834110
21 3000 3600 600 6.45814 37228085 952282
Binary Interfile : "a00760.pet.img"
Analyze 7.5 Dynamic: "a00760.hdr|img"
Data Type : "single"
ROI Scale Factor = 1.0000000000000000 (single precision)
1 ifm=[ 0.208276; 0.000000] fsf= 6169.87753 ofm=[1285.038330; 0.000000]
2 ifm=[ 0.500311; 0.000000] fsf= 6065.82528 ofm=[3034.797852; 0.000000]
3 ifm=[ 0.598478; 0.000000] fsf= 5963.56647 ofm=[3569.066406; 0.000000]
4 ifm=[ 0.513827; 0.000000] fsf= 5863.04051 ofm=[3012.590820; 0.000000]
5 ifm=[ 0.491587; 0.000000] fsf= 5764.19165 ofm=[2833.600342; 0.000000]
6 ifm=[ 0.426429; 0.000000] fsf= 5667.02024 ofm=[2416.583252; 0.000000]
7 ifm=[ 0.711539; 0.000000] fsf= 2762.26545 ofm=[1965.459961; 0.000000]
8 ifm=[ 0.673439; 0.000000] fsf= 2669.9355 ofm=[1798.038940; 0.000000]
9 ifm=[ 0.656249; 0.000000] fsf= 2580.66378 ofm=[1693.557495; 0.000000]
10 ifm=[ 1.233394; 0.000000] fsf= 1226.34883 ofm=[1512.570679; 0.000000]
11 ifm=[ 1.293754; 0.000000] fsf= 1145.72729 ofm=[1482.289062; 0.000000]
12 ifm=[ 1.405996; 0.000000] fsf= 1070.39924 ofm=[1504.977051; 0.000000]
13 ifm=[ 1.332499; 0.000000] fsf= 1000.0265 ofm=[1332.533936; 0.000000]
14 ifm=[ 1.269559; 0.000000] fsf= 934.278589 ofm=[1186.122192; 0.000000]
15 ifm=[ 1.348324; 0.000000] fsf= 872.855176 ofm=[1176.891846; 0.000000]
16 ifm=[ 1.439322; 0.000000] fsf= 815.468001 ofm=[1173.720825; 0.000000]
17 ifm=[ 3.150705; 0.000000] fsf= 289.88255 ofm=[913.334290; 0.000000]
18 ifm=[ 3.278511; 0.000000] fsf= 244.560275 ofm=[801.793518; 0.000000]
19 ifm=[ 5.376124; 0.000000] fsf= 95.0973048 ofm=[511.254883; 0.000000]
20 ifm=[ 5.564751; 0.000000] fsf= 67.6854461 ofm=[376.652618; 0.000000]
21 ifm=[ 5.541458; 0.000000] fsf= 48.175209 ofm=[266.960907; 0.000000]
Calibration Maximum= 3569.0664 kBq/ml (single precision)
Calibration Minimum= 0 kBq/ml (single precision)
Conversion completed.
Please close the open figure to return to the command line prompt.
| Type | binary executable, AVW library required |
| Location |
/opt/wmic/bin (Solaris/Linux wrapper script) \\MMIC_FS1\WMIC_Apps, \\MMIC-FS1\WMIC_Apps\win64 (32-bit/64-bit Windows binaries) |
| Version | 2.7 |
Description: sthg2ana is a
command line program that is meant to convert any dynamic PET image into a
signed 16-bit Analyze
Image 7.5 file.
To open the image file, a call of the function
AVW_OpenImageFile() of the
AVW library is
made. Therefore, the capabilities of
sthg2ana with respect to supported input file formats are determined
by the AVW library.
sthg2ana takes care of properly
finding the global minimum value, the global maximum value and the scale
factor. Use the command line flag -c
to calculate input file minimum and maximum values as for some image file
formats (for example some versions of the
NIfTI Data Format) the value 0
is returned as global minimum and global maximum, respectively.
Usage: sthg2ana [-c] [-o] file[s]
file[s] the name[s] of the image file[s] to be converted. They must be "something" which is read by the AVW library as FLOAT (Data type= 64 AVW_FLOAT or 16384 AVW_DOUBLE_FLOAT). The output images will be 16-bit signed integer data (Data type= 8 AVW_SIGNED_SHORT) with the appropriate ROIScaleFactor stored in the header file. The original file name extensions will be replaced with `.hdr' and `.img', respectively. If the input files are already in AnalyzeImage(7.5) format, then a '1' will be appended to the base name of the files. -c calculate input file minimum and maximum values. For some image file formats, AVW_OpenImageFile() does not return the actual MinimumDataValue and MaximumDataValue for the input file[s]. Use this flag to calculate the global minimum and maximum values. -o overwrite ON (default: overwrite OFF - if the output files already exist, then they will not be overwritten)
Input file : "16800102_DD_PET_bsl_nonatten.iee"
Data format = 11 GEADVANCE
NumVols = 26
Width = 128
Height = 128
Depth = 35
Voxel width = 2.34375 mm
Voxel height= 2.34375 mm
Voxel depth = 4.25 mm
glmin = -0.002764
glmax = 0.007328
--> a scale factor of 0.0000002236396374 (double precision)
will be applied to convert the image data into signed short
volume 0: min= -0.0027635545 (-12357), max= 0.0032541733 ( 14551)
volume 1: min= -0.0021226688 ( -9491), max= 0.0073283855 (-32767)
Hmm, the integer max should actually be 32769. I try it again with an
updated glmax value.
glmin = -0.002764
glmax = 0.0073283855
--> a scale factor of 0.0000002236514019 (double precision)
will be applied to convert the image data into signed short
volume 0: min= -0.0027635545 (-12357), max= 0.0032541733 ( 14550)
volume 1: min= -0.0021226688 ( -9491), max= 0.0073283855 ( 32767)
volume 2: min= -0.0021626423 ( -9670), max= 0.0049025002 ( 21920)
volume 3: min= -0.0023246598 (-10394), max= 0.0065824515 ( 29432)
volume 4: min= -0.0023461808 (-10490), max= 0.0060981419 ( 27266)
volume 5: min= -0.0014876558 ( -6652), max= 0.0048609907 ( 21735)
volume 6: min= -0.0016661918 ( -7450), max= 0.0053821108 ( 24065)
volume 7: min= -0.0014612994 ( -6534), max= 0.0054523339 ( 24379)
volume 8: min= -0.0010024582 ( -4482), max= 0.0054889899 ( 24543)
volume 9: min= -0.001128411 ( -5045), max= 0.005620101 ( 25129)
volume 10: min= -0.0010891993 ( -4870), max= 0.005042186 ( 22545)
volume 11: min= -0.00073882204 ( -3303), max= 0.0044765132 ( 20016)
volume 12: min= -0.00074835977 ( -3346), max= 0.0038151864 ( 17059)
volume 13: min= -0.00070397067 ( -3148), max= 0.0036536546 ( 16336)
volume 14: min= -0.00065414753 ( -2925), max= 0.0033836644 ( 15129)
volume 15: min= -0.00072272785 ( -3231), max= 0.0029330335 ( 13114)
volume 16: min= -0.00058612489 ( -2621), max= 0.0028217703 ( 12617)
volume 17: min= -0.00051442458 ( -2300), max= 0.0028146172 ( 12585)
volume 18: min= -0.00049125892 ( -2197), max= 0.0026030187 ( 11639)
volume 19: min= -0.00060776836 ( -2717), max= 0.0020727254 ( 9268)
volume 20: min= -0.00046873366 ( -2096), max= 0.0017879589 ( 7994)
volume 21: min= -0.00045446554 ( -2032), max= 0.0019034527 ( 8511)
volume 22: min= -0.00042255118 ( -1889), max= 0.0016251947 ( 7267)
volume 23: min= -0.00038577712 ( -1725), max= 0.0015312384 ( 6847)
volume 24: min= -0.00042828752 ( -1915), max= 0.0014196591 ( 6348)
volume 25: min= -0.00036300253 ( -1623), max= 0.0013784897 ( 6164)
Output file : "16800102_DD_PET_bsl_nonatten.hdr"
Data format = 4 AnalyzeImage(7.5)
NumVols = 26
Width = 128
Height = 128
Depth = 35
Voxel width = 2.34375 mm
Voxel height= 2.34375 mm
Voxel depth = 4.25 mm
glmin = -12357
glmax = 32767
cal_min = -0.0027635545
cal_max = 0.0073283855
ROIScale = 0.0000002236514019 (double precision)
= 0.0000002236513978 (single precision)
Input file : "m99016.v"
Data format = 9 CTIECAT7
Data type = 64 AVW_FLOAT
NumVols = 18
Width = 128
Height = 128
Depth = 35
Voxel width = 2.34375 mm
Voxel height= 2.34375 mm
Voxel depth = 4.25 mm
glmin = -14308.289
glmax = 34631.726
--> a scale factor of 1.0569086569391033 (double precision)
will be applied to convert the image data into signed short
volume 0: min= -14308.289 (-13538), max= 15084.472 ( 14272)
volume 1: min= -12093.518 (-11442), max= 34631.727 ( 32767)
volume 2: min= -11032.296 (-10438), max= 32186.289 ( 30453)
volume 3: min= -5740.502 ( -5431), max= 26222.4 ( 24810)
volume 4: min= -5756.7603 ( -5447), max= 24164.234 ( 22863)
volume 5: min= -5182.6094 ( -4904), max= 23069.336 ( 21827)
volume 6: min= -4075.1941 ( -3856), max= 20426.467 ( 19327)
volume 7: min= -7647.3188 ( -7236), max= 19454.352 ( 18407)
volume 8: min= -6098.1265 ( -5770), max= 18679.182 ( 17673)
volume 9: min= -4733.5093 ( -4479), max= 17491.615 ( 16550)
volume 10: min= -5992.1484 ( -5670), max= 16008.05 ( 15146)
volume 11: min= -7412.2119 ( -7013), max= 15981.348 ( 15121)
volume 12: min= -7010.856 ( -6633), max= 16710.389 ( 15811)
volume 13: min= -7183.4961 ( -6797), max= 17113.162 ( 16192)
volume 14: min= -10562.646 ( -9994), max= 18414.24 ( 17423)
volume 15: min= -8743.4717 ( -8273), max= 18286.117 ( 17302)
volume 16: min= -8392.2109 ( -7940), max= 16023.494 ( 15161)
volume 17: min= -10225.085 ( -9675), max= 21682.314 ( 20515)
Output file : "m99016.hdr"
Data format = 4 AnalyzeImage(7.5)
Data type = 8 AVW_SIGNED_SHORT
NumVols = 18
Width = 128
Height = 128
Depth = 35
Voxel width = 2.34375 mm
Voxel height= 2.34375 mm
Voxel depth = 4.25 mm
glmin = -13538
glmax = 32767
cal_min = -14308.289
cal_max = 34631.727
ROIScale = 1.0569086569391033 (double precision)
= 1.0569086074829102 (single precision)
descrip = "2.3 (10030,10000) 1368550922: -o m99016.v" (41 characters)
Input file : "h03136dy1_BP.hdr"
Data format = 4 AnalyzeImage(7.5)
Data type = 64 AVW_FLOAT
NumVols = 1
Width = 128
Height = 128
Depth = 95
Voxel width = 2.096 mm
Voxel height= 2.096 mm
Voxel depth = 2.425 mm
glmin = -1.5091565
glmax = 6.3636827
--> a scale factor of 0.0001942101122131 (double precision)
will be applied to convert the image data into signed short
volume 0: min= -1.5091565 ( -7771), max= 6.3636827 ( 32767)
Output file : "h03136dy1_BP1.hdr"
Data format = 4 AnalyzeImage(7.5)
Data type = 8 AVW_SIGNED_SHORT
NumVols = 1
Width = 128
Height = 128
Depth = 95
Voxel width = 2.096 mm
Voxel height= 2.096 mm
Voxel depth = 2.425 mm
glmin = -7771
glmax = 32767
cal_min = -1.5091565
cal_max = 6.3636827
ROIScale = 0.0001942101122131 (double precision)
= 0.0001942101080203 (single precision)
descrip = "2.3 (10030,10000) 1368550925: -o h03136dy1_BP.hdr" (49 characters)
The example file h00013_01_hammers_gm.nii was written by SPM12 in NIfTI format. Because AVW_OpenImageFile() returns 0 as MinimumDataValue and MaximumDataValue, respectively, a scale factor of 1 is used to convert the float image data into signed short which is a simple round to the nearest integer operation without scaling of the values.
% sthg2ana h00013_01_hammers_gm.nii |& fgrep -v 'nifti_datatype:<4>'
Input file : "h00013_01_hammers_gm.nii"
Data format = 10 NIFTI
Data type = 64 AVW_FLOAT
NumVols = 1
Width = 256
Height = 256
Depth = 150
Voxel width = 1 mm
Voxel height= 1 mm
Voxel depth = 1 mm
glmin = 0
glmax = 0
--> a scale factor of 1.0000000000000000 (double precision)
will be applied to convert the image data into signed short
volume 0: min= 0 ( 0), max= 85 ( 85)
Output file : "h00013_01_hammers_gm.hdr"
Data format = 4 AnalyzeImage(7.5)
Data type = 8 AVW_SIGNED_SHORT
NumVols = 1
Width = 256
Height = 256
Depth = 150
Voxel width = 1 mm
Voxel height= 1 mm
Voxel depth = 1 mm
glmin = 0
glmax = 85
cal_min = 0
cal_max = 85
ROIScale = 1.0000000000000000 (double precision)
= 1.0000000000000000 (single precision)
descrip = "2.4 (10030,10000) 1629792985: h00013_01_hammers_gm.nii" (54 characters)
Now the output file h00013_01_hammers_gm.hdr can be converted into an Analyze Object Map file using ima2obj and then correctly labelled with panames.
Note: Had ima2obj been used on the original SPM12 generated NIfTI file h00013_01_hammers_gm.nii, then the float values would have been truncated (rounded towards zero) for data type conversion into integer. For example, a float value of 16.998962 in the right cerebellum would have been cast into an integer value of 16 which then panames would have labelled as G_occtem_la_l in the Analyze Object Map file created.
The example file h00013_odi.nii was written with the code from the NIFTI Matlab library in NIfTI format. Because AVW_OpenImageFile() returns 0 as MinimumDataValue and MaximumDataValue, respectively, the command line option -c is used to calculate the global minimum and maximum values in the input image. Then the image data are scaled into the signed short range.
% sthg2ana -c h00013_odi.nii |& fgrep -v 'nifti_datatype:<4>'
Input file : "h00013_odi.nii"
Data format = 10 NIFTI
Data type = 16384 AVW_DOUBLE_FLOAT
NumVols = 1
Width = 128
Height = 128
Depth = 60
Voxel width = 1.875 mm
Voxel height= 1.875 mm
Voxel depth = 2.1 mm
Calculating the global minimum and maximum values:
volume 0: min= 0 , max= 1
glmin = 0
glmax = 1
--> a scale factor of 0.0000305185094760 (double precision)
will be applied to convert the image data into signed short
volume 0: min= 0 ( 0), max= 1 ( 32767)
Output file : "h00013_odi.hdr"
Data format = 4 AnalyzeImage(7.5)
Data type = 8 AVW_SIGNED_SHORT
NumVols = 1
Width = 128
Height = 128
Depth = 60
Voxel width = 1.875 mm
Voxel height= 1.875 mm
Voxel depth = 2.1 mm
glmin = 0
glmax = 32767
cal_min = 0
cal_max = 1
ROIScale = 0.0000305185094760 (double precision)
= 0.0000305185094476 (single precision)
descrip = "2.7 (10030,10000) 1637056973: -c h00013_odi.nii" (47 characters)
Note: After the data type conversion, the size of the output file h00013_odi.img is only a quarter of the size of the original image h00013_odi.nii.
| Type | binary executable (deployed Matlab R2008a application) |
| Location | /data/secondary/rhinz/bin (Solaris/Linux wrapper script) |
| Version | 3.1 |
Description: hrrt_gfilt
is a deployed Matlab application to perform Gaussian 3D filtering of
dynamic PET images in
Analyze 7.5 format.
Two methods are available: either convolution in images space (default)
using the function
imfilter of the
Matlab Image Processing Toolbox
or multiplication in the frequency domain after 3D Fourier transform
using the built-in Matlab functions
fftn and
ifftn, respectively.
Usage: hrrt_gfilt fname fwhm [fouriermeth]
Command line arguments:
fname A string containing the file name of the Analyze 7.5 image
file to filter. The file name can be without file name
extension or with either '.hdr' or '.img' extension.
Supported data types are 4 (16-bit signed integer) and
16 (single precision float).
fwhm A scalar floating point number specifying the Full Width
at Half Maximum (FWHM) value for the Gaussian 3D filter in
millimetres.
fouriermeth Set this flag to perform the convolution of the 3D images
with the Gaussian 3D filter as multiplication of the two
Fourier transforms in the frequency domain. This method
becomes efficient for big kernels (i.e. large FWHM) as
the execution time practically only depends on the image
volume size.
Outputs:
The Analyze 7.5 image with the Point Spread Function (PSF)
of the Gaussian 3D filter will be written into the current
working directory.
The filtered PET images in Analyze 7.5 format will be
written where the input files are located. The file names
will be derived from the fname argument extended by the
suffix '_g' and the value of the filter FWHM. If the output
files already exist, they will not be overwritten.
The execution times of the plot were obtained filtering a single frame
image volume from the HRRT scanner (140 x 160 x 128 voxels) on an IBM Blade
Server node with
Xeon 3.06 GHz processor and 2.5 Gigabytes RAM running SuSe Linux
Enterprise Server 10.
Note that for the image space convolution with a 3D Gaussian of
10 mm FWHM the resident set size was about
63 megabytes and the total virtual memory used was 111 megabytes. To perform
the same filtering in the Fourier domain, up to 815 megabytes of memory were used as
resident set size and a peak of 863 megabytes as total virtual memory.
The execution time of hrrt_gfilt performing the convolution in image space (green line) is proportional to the number of elements in the Gaussian filter kernel (red line) which in turn is approximately proportional to the third power of the width of the Gaussian. Performing the filtering as a multiplication in the Fourier domain, the execution time of hrrt_gfilt practically only depends on the image volume size and not on the Gaussian filter width (blue line).
In summary, for systems low on memory hrrt_gfilt is best run performing the convolution in images space (default method). On systems with sufficient physical memory, the Gaussian 3D filtering can be performed considerably faster in the Fourier domain for filter widths greater than about 3.5 mm FWHM. For large filters (width of 10 mm FWHM or more), Gaussian 3D filtering has to be performed in the Fourier domain because the time needed for the convolution in image spaces is prohibitively long.
This image has been iteratively reconstructed and therefore contains only nonnegative voxel values.
% hrrt_gfilt data/pk11195/h00050/h00050_20070417_nf02_i06_v2 2.5
PET Image:
----------
File Name : "data/pk11195/h00050/h00050_20070417_nf02_i06_v2"
Byte Order : Little Endian
Subject Identifier : "h00050 " (10 characters)
Description : "Default " (80 characters)
Series Number : "6 " (10 characters)
Series Date : "2007-04-17" (10 characters)
Series Time : "10:26:46 " (10 characters)
Patient Position : "HFS" (3 characters)
Data Type = 4 SIGNED_SHORT
Width = 140
Height = 180
Depth = 128
Number Of Time Sl. = 1
Voxel Width = 1.21875 mm
Voxel Height = 1.21875 mm
Voxel Depth = 1.21875 mm
ReconstructionMeth : "-OSEM3D-OP" (10 characters)
Maximum Data Value = 32767
Minimum Data Value = 0
ROI Scale Factor = 0.0005724382936023 (single precision)
Calibration Maximum= 18.757086 kBq/ml (single precision)
Calibration Minimum= 0 kBq/ml (single precision)
Radioactive decay : lambda= Inf 1/s, T1/2= 0 s= 0 min
Gaussian Lowpass Filter:
------------------------
FWHM = 2.5 mm
Gaussian's sigma = 1.06165 mm
Filter size = [9, 9, 9] voxels
Performing 3-D image filtering, volume 1, done.
Elapsed time : 95.0 seconds with convolution in image space using IMFILTER
1 ifm=[ 32767; 0] ofm=[ 19338; 0] is= 3.9770136e+09 os= 3.962529e+09
Filtered PET Image:
-------------------
File Name : "data/pk11195/h00050/h00050_20070417_nf02_i06_v2_g2.5"
Byte Order : Big Endian
Subject Identifier : "h00050 " (10 characters)
Description : "Default " (80 characters)
Series Number : "6 " (10 characters)
Series Date : "2007-04-17" (10 characters)
Series Time : "10:26:46 " (10 characters)
Patient Position : "HFS" (3 characters)
Data Type = 4 SIGNED_SHORT
Width = 140
Height = 180
Depth = 128
Number Of Time Sl. = 1
Voxel Width = 1.21875 mm
Voxel Height = 1.21875 mm
Voxel Depth = 1.21875 mm
ReconstructionMeth : "-OSEM3D-OP" (10 characters)
Maximum Data Value = 19338
Minimum Data Value = 0
ROI Scale Factor = 0.0005724382936023 (single precision)
Calibration Maximum= 11.069812 kBq/ml (single precision)
Calibration Minimum= 0 kBq/ml (single precision)
Radioactive decay : lambda= Inf 1/s, T1/2= 0 s= 0 min
Two files psf2.5.hdr and psf2.5.img have been saved in the current working directory. They are Analyze 7.5 Images of the 3D Gaussian filter used.
This image was generated using filtered backprojection and has hence positive and negative voxel values.
% hrrt_gfilt h00005/160920/h00005_20060623_160920_s15_nf29.hdr 10 fft
PET Image:
----------
File Name : "h00005/160920/h00005_20060623_160920_s15_nf29"
Byte Order : Little Endian
Subject Identifier : "h00005 " (10 characters)
Description : "PET Water Ranging Study - Pelvis (Adult) " (80 characters)
Series Number : "15 " (10 characters)
Series Date : "2006-06-23" (10 characters)
Series Time : "16:09:20.9" (10 characters)
Patient Position : "FFS" (3 characters)
Data Type = 4 SIGNED_SHORT
Width = 256
Height = 256
Depth = 81
Number Of Time Sl. = 29
Voxel Width = 2.66615 mm
Voxel Height = 2.66615 mm
Voxel Depth = 2 mm
ReconstructionMeth : "DIFT " (10 characters)
Maximum Data Value = 32767
Minimum Data Value = -28858
ROI Scale Factor = 0.0972374901175499 (single precision)
Calibration Maximum= 3186.1807 kBq/ml (single precision)
Calibration Minimum= -2806.0793 kBq/ml (single precision)
Radioactive decay : lambda= 0.00567038 1/s, T1/2= 122.24 s= 2.03733 min
Gaussian Lowpass Filter:
------------------------
FWHM = 10 mm
Gaussian's sigma = 4.24661 mm
Filter size = [15, 15, 19] voxels
Fourier domain vol = [512, 512, 128] complex elements
FFT of the filter : 14.4 seconds
Performing 3-D image filtering, volume 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, done.
Elapsed time : 891.4 seconds with FFT, multiplication and IFFT
1 ifm=[ 523; -601] ofm=[ 22; -15] is= 977772 os= 954499
2 ifm=[ 10600; -8434] ofm=[ 351; -277] is= -3509256 os= -3502999
3 ifm=[ 14203; -13008] ofm=[ 342; -378] is= -4060605 os= -4065483
4 ifm=[ 20896; -15976] ofm=[ 445; -330] is= -3469000 os= -3461777
5 ifm=[ 15785; -18523] ofm=[ 566; -443] is= -485374 os= -516673
6 ifm=[ 17024; -13057] ofm=[ 666; -472] is= 10444445 os= 10254220
7 ifm=[ 20079; -21830] ofm=[ 657; -546] is= 21493813 os= 21058945
8 ifm=[ 22907; -20863] ofm=[ 669; -503] is= 31352821 os= 30746225
9 ifm=[ 28142; -24369] ofm=[ 779; -550] is= 35761011 os= 35013359
10 ifm=[ 32767; -26168] ofm=[ 825; -760] is= 35300224 os= 34560502
11 ifm=[ 24912; -28858] ofm=[ 705; -662] is= 37771160 os= 37075487
12 ifm=[ 22895; -25056] ofm=[ 674; -583] is= 37645430 os= 36897564
13 ifm=[ 18921; -14920] ofm=[ 661; -535] is= 41979180 os= 41099870
14 ifm=[ 22271; -20080] ofm=[ 686; -551] is= 39901577 os= 39131415
15 ifm=[ 28055; -25854] ofm=[ 813; -712] is= 37881118 os= 37055749
16 ifm=[ 15410; -16220] ofm=[ 512; -488] is= 35518177 os= 34839504
17 ifm=[ 14811; -13225] ofm=[ 481; -379] is= 35008494 os= 34294511
18 ifm=[ 15483; -18309] ofm=[ 501; -379] is= 32835573 os= 32152621
19 ifm=[ 12432; -11845] ofm=[ 412; -271] is= 32253564 os= 31647440
20 ifm=[ 13802; -13434] ofm=[ 437; -309] is= 29836462 os= 29209363
21 ifm=[ 8826; -7414] ofm=[ 363; -234] is= 27125580 os= 26597523
22 ifm=[ 9807; -10090] ofm=[ 239; -180] is= 24652161 os= 24145258
23 ifm=[ 7731; -10667] ofm=[ 296; -178] is= 22665229 os= 22188779
24 ifm=[ 6608; -7002] ofm=[ 205; -153] is= 19604003 os= 19169156
25 ifm=[ 5035; -5285] ofm=[ 174; -136] is= 16927307 os= 16543720
26 ifm=[ 4407; -3827] ofm=[ 143; -93] is= 14280570 os= 13972855
27 ifm=[ 4916; -3352] ofm=[ 141; -93] is= 12310550 os= 12042626
28 ifm=[ 4308; -3458] ofm=[ 117; -86] is= 10262569 os= 10035164
29 ifm=[ 3555; -2976] ofm=[ 125; -75] is= 8741869 os= 8571112
Filtered PET Image:
-------------------
File Name : "h00005/160920/h00005_20060623_160920_s15_nf29_g10"
Byte Order : Little Endian
Subject Identifier : "h00005 " (10 characters)
Description : "PET Water Ranging Study - Pelvis (Adult) " (80 characters)
Series Number : "15 " (10 characters)
Series Date : "2006-06-23" (10 characters)
Series Time : "16:09:20.9" (10 characters)
Patient Position : "FFS" (3 characters)
Data Type = 4 SIGNED_SHORT
Width = 256
Height = 256
Depth = 81
Number Of Time Sl. = 29
Voxel Width = 2.66615 mm
Voxel Height = 2.66615 mm
Voxel Depth = 2 mm
ReconstructionMeth : "DIFT " (10 characters)
Maximum Data Value = 825
Minimum Data Value = -760
ROI Scale Factor = 0.0972374901175499 (single precision)
Calibration Maximum= 80.220932 kBq/ml (single precision)
Calibration Minimum= -73.90049 kBq/ml (single precision)
Radioactive decay : lambda= 0.00567038 1/s, T1/2= 122.24 s= 2.03733 min
Two files psf10.hdr and psf10.img have been saved in the current working directory. They are Analyze 7.5 Images of the 3D Gaussian filter used.
@(#)toolshelp.html 2.36 24/07/09 16:03:38 © Rainer Hinz, The University of Manchester.