#1: The command below prints out the acquisition start time and the number
of rows for each valid DICOM file in the directories below the current one:
dicom_hinfo -tag 0008,0031 0028,0010 */*.dcm
One sample output line would be
TASK-A/image-00102-004243.dcm 141255.882508 256
#2: A more complicated example searches all the directories below the current one,
then prints out a list of summaries of what look like unique acquisitions.
This could be used to figure out what kind of data you have when someone gives
you a bunch of DICOM files with no obvious structure to their filenames.
find . -type f | xargs dicom_hinfo -tag 0008,0031 0028,0010 | uniq -f 1 -c
The output from the above example was
9 ./A/A/A/Z01 154116 256
9 ./A/A/A/Z10 154210 256
38 ./A/A/A/Z19 154245 64
126 ./A/A/C/Z05 154326 256
6000 ./A/A/H/Z01 154854 64
2400 ./A/J/D/Z21 155617 64
126 ./A/M/S/Z03 160228 256
40 ./A/M/W/Z25 160304 64
40 ./A/M/Y/Z13 160332 64
126 ./A/N/A/Z01 160404 256
126 ./A/N/E/Z23 160411 256
126 ./A/N/J/Z19 160417 256
1 ./A/N/O/Z15 161252 960
1 ./A/N/O/Z16 161403 640
9 ./A/N/O/Z17 150935 256
9 ./A/N/P/Z00 151039 256
37 ./A/N/P/Z10 151122 64
120 ./A/N/Q/Z21 151203 256
6000 ./A/N/V/Z11 151624 64
2400 ./A/W/S/Z05 153010 64
My goal was the find the structural and FMRI collections of images mixed
in with various localizers and other 'junk'. Based on the above, it seems:
* the 126 files starting with ./A/A/C/Z05 are a structural set
* the 6000 files starting with ./A/A/H/Z01 are an FMRI set
* the 2400 files starting with ./A/J/D/Z21 are an FMRI set
* the 126 files starting with ./A/M/S/Z03 are a structural set
and so on. This information makes it possible to extract the desired files
from the giant collection of un-informative filenames, create AFNI datasets
(using program Dimon and its '-infile_list' option appropriately), and then
look at them to make final decisions about what to keep.
#3: Continuing the above example with actual creation of AFNI dataset
from the collection of files, a script (in csh syntax):
#!/bin/tcsh
\rm -f qq*.out
find . -type f \
| xargs dicom_hinfo -tag 0008,0031 0028,0010 0028,0011 \
| awk '$3 == $4' >> qqa.out
uniq -f 1 -c qqa.out | awk '$1 > 99' > qqb.out
foreach ddd ( `cat qqb.out | awk '{print $3}'` )
echo 'Organizing files with time stamp $ddd'
grep $ddd qqa.out | awk '{print $1}' > qqc_${ddd}.out
Dimon -infile_list qqc_${ddd}.out -dicom_org -GERT_Reco \
-gert_create_dataset -gert_to3d_prefix ACQT_${ddd} -quit
end
As before, the find command gets all the DICOM files under the current
In this case, the awk command also filters out images that are not square.
The output of 'find' is piped into xargs to prevent creating a gigantic
command line, since there are over 17,000 files in this directory tree.
The uniq command finds files with unique time stamps, and the
awk command filters out those lines that don't have more than 99
such files.
The foreach loop over variable ddd (the time stamp) creates a file list
that matches the given value, then runs Dimon to create an AFNI dataset.
[Not all of these datasets are actually useful, but it is easy to]
[delete the ones that are not relevant to the research underway. ]
Note the use of the '-dicom_org' option to Dimon to have it organize
the files to be in the correct order -- otherwise, it would take the
files in their filename alphabetical order, which is not always right.
This ordering is done using various DICOM fields, including
0054,1330 = Image Index
0020,0013 = Instance Number
This example solved a real problem with image files dumped from a PACS.
You might have to change things around to solve your problem, but I
hope that this sample script will give you an idea of how to start.
--- RWCox - 15 Nov 2011 ---
