Skip to content

Latest commit

 

History

History
87 lines (55 loc) · 4.1 KB

README.md

File metadata and controls

87 lines (55 loc) · 4.1 KB
Raw

#Label DICOMS and check task data The first step of the MR processing pipeline is to identify what each DICOM files corresponds to, verify that the subject_id, PatientID and Patient Name match and match up E-prime output files with the DICOMs.

The main command is

gather_dicoms.py subject_id

This will generate a data file that lists key features of each DICOM series and which modality, task and run the series should correspond to.

##DICOM matching template

The dicom_dict.csv file provides a template for identifying scans based on the ProtocolName and DICOM tags. Each possible scan should be uniquely identified by a single row. The fields marked ID in the table below are used to identify the scan and associate it with the information in the meta fields.

Field Content Tag ID/meta Description
ImageType String ID The XXX element of the
EchoTime float (ms) ID The echo time
MB String ID The XXX elment of
ProtocolName String ID The name of the protocl
nimages int n/a ID The number of DICOM files in the folder for this series
modality String: T1w, dwi, fmap, func n/a meta
type String: T1w, sbref, dwi, AP,PA, magnitude[12]*, phasediff, bold n/a meta
task String: {cn,en,sp}{pma,wma,wmv,sma} n/a meta
acq String: multiband, singleband n/a meta
for String: dwi, func n/a meta

##Matching The script iterates through directories and uses all of the ID fields (read from a single DICOM file, which should not be assumed to be the first file in the directory) and the number of files in a directory to identify the series.

###Assigning runs

##Sanity Checks and Errors ###Subject ID The script will raise an exception and terminate if any of (1) the subject_id passed on the command line; (2) PatientID; and (3) PatientName are not identical.

The solution to this is to:

  1. Review the AcquisitionDate and AcquisitionTime fields in the DICOMS
  2. Review the scan schedule and log sheets to determine which subject should have been in the scanner at the time of acquisition
  3. Rename the subject folder and/or edit the DICOM fields

To edit the fields:

  1. Create a new folder for the DICOM export directories based on the existing folder with the suffix _Remapped
  2. Use DicomBrowser to correct the field(s), storing the corrected files in this new folder
  3. Rename the original folder by adding the prefix Orig_

Multiple Sessions

The folder structure for raw data is assumed to be of the form BILNNNN/Hoeftxxxx/series/*.dcm. Multiple Hoeft* directories will raise an exception. These directories should be merged.

Tasks

For scans of type bold, the gather script attempts to match the scan to a processed eprime log file by

  • Searching for an eprime file for the same subject, run and task
  • Checking that the runtime of the script and the AcquisitionDate/AcquistionTime are close (within 120secs)
  • Checking that there are the same number of bold and eprime runs for each task

If any of these conditions fail, the script will generate a warning and suffix the output with _eprime. If the time difference is >60 seconds, an information message will be printed. In practice, the stimulus computer and scanner clocks appear to be 50-100secs apart.

##Output

##Fieldmaps (phase-magnitude sets) DICOMS for magnitude images appear to be inconsistently exported in two different ways:

  • A single directory containing both echoes (IM*-0001.dcm and IM*-0002.dcm)
  • Two directories, one for each echo, having the same series number

The reconstructed phase difference is always in a separate directory. Currently, this is handled by having 5 entries for every fieldmap (for historical reasons):

  1. EchoTime = TE1; nimages = #slices; type = magnitude1
  2. EchoTime = TE2; nimages = #slices; type = magnitude2
  3. EchoTime = TE1; nimages = 2*#slices; type = magnitude
  4. EchoTime = TE2; nimages = 2*#slices; type = magnitude
  5. EchoTime = TE2; nimages = #slices; type = phasediff

If 3 or 4 are matched, we rely on dcm2niix (the command) to separate the echoes and use further hacks in mk_bids and dcm2niix

#Convert DICOMs to bids-raw

mk_bids.py subject_id