-
Notifications
You must be signed in to change notification settings - Fork 29
PET raw data
Raw PET data (listmode, sinograms and normalisation files) need to be exported via the scanner console. This can be achieved with the following steps:
- Select the desired study in the patient browser.
- Inside the selected study, highlight the raw PET data to be exported.
- In the patient browser, choose
Applications > DICOM Tools > Export Large List Mode
. - Select a desired export destination. The default is typically
C:\MedCom\temp\CDR_OFFLINE
. It is strongly recommended that you append a new folder name to the path to avoid exporting all files to the root of the directory. If you are able to mount USB devices, these can also be used as a destination. - Click
Export
. - The console will state that the transfer is complete. However, especially when exporting large listmode files, the transfer can take considerably longer than this. It is possible to check whether there is any further disk activity through Windows Explorer when in
Advanced User
mode. - Once complete, the data can either be moved via network shares if available on the scanner or by safely unmounting the attached drive. Data exported using this method should be in the form of a DICOM header (
.dcm
) and binary file (.bf
). There should be one pair per raw data element.
To perform PET reconstruction with SIRF, the exported data need to be converted first.
It is necessary to extract the Interfile header (and occasionally the raw data) contained within the DICOM header. There are 2 possibilities:
If you have a research agreement with Siemens you may have access to their offline processing tools which can perform the extraction. How to use these tools is beyond the scope of this wiki page.
Our open-source project (https://github.com/UCL/pet-rd-tools) includes the nm_extract
tool which can be used to extract the Interfile headers and data from the DICOM file.
Basic usage
nm_extract -i <MYDCM>
where <MYDCM>
is the path to the DICOM header. The program uses the following file name conventions:
- Listmode data will have the extension
.l
and the associated Interfile header will be.l.hdr
. - Sinograms will be
.s
and.s.hdr
. - Normalisation files will have extensions
.n
and.n.hdr
.
NOTE: Header information that was previously converted via the STIR script convertSiemensToInterfile.sh
is no longer supported. STIR now reads the Siemens mMR Interfile header natively. It is necessary to re-extract any previously converted headers from the original raw data to be able to perform PET reconstruction with SIRF.
Unfortunately, sinograms are usually compressed by the scanner and must be decompressed before reconstruction. This can only be performed via the scanner console or with Siemens offline processing tools. You will need to use intfcompr
command line tool on the console, run it to get a usage message. You might be able to copy it to another Windows machine for convenience. To our knowledge, listmode data are never compressed. We therefore recommend to use listmode data with SIRF.
Occasionally, the Siemens mMR scanner can produce list mode files that do not start with a timing tag set to zero. Handling this is STIR Issue #70. At present, you have to
work-around this by specifying a frame start time corresponding to the first timing tag. You can use STIR's list_lm_events
. From within SIRF, you could use ListModeToSinograms.get_time_at_which_num_prompts_exceeds_threshold
. This was originally designed for dynamic data to find where the activity enters the FOV, but should work for static data as well. Check examples/Python/PET/reconstruct_from_listmode.py
.