Dcmdb

Dcmdb is a simple system for retrieving data from the scanner. At 1AM every morning, dcmdb retrieves the previous day's studies to the shared volume dcmdb on odin. Studies are sorted into folders by Referring Physician Name. Studies without a referring physician name are put in the unknown folder. Dcmdb is not an archive service. There is no off-site backup. This page describes dcmdb as of June 2019.

Connecting to dcmdb
Dcmdb is a file server share. To connect from a Mac, use the Finder Go menu and choose connect to server. Enter smb://odin.ad.bangor.ac.uk in the server address. Click connect. The user name and password are both mruser. Mount the volume dcmdb. To connect from Windows 10, open Windows Explorer and choose Map Network Drive... from the Tools menu.

Enter \\saga.ad.bangor.ac.uk\dcmdb or \\odin.ad.bangor.ac.uk\dcmdb as the folder to map. Be sure to tick Connect using different credentials. Otherwise the PC will attempt to connect using your Bangor login and password. The correct username and password is mruser.

Finding Your Data
At the top level, files are sorted by Referring Physician. Set Referring Physician when you create a new exam on the scanner. Typically you will set Referring Physician to the PI's name but it can be any useful text string up to a maximum of 30 characters. For example, the weekly QA tests are in the QA folder. Do not use the characters ?,*, or / in the Referring Physician field. Automated preprocessing changes spaces to underscore '_'.

Inside the Referring Physician folders, studies are sorted by Patient Name and then by study date and series. For example, the path to a single study is dcmdb/Adams/SR0029/20110825.

Multiple studies from the same day using the same patient ID will break the sorting system. The work around is to change the patient name. For example SR0099_run01 and SR0099_run02. Dcmdb allows for special folders or patient comments with additional pre-processing. Currently neuropatient and bedpost are the special folders. All other folders use the standard pre-processing described in the next section.

Standard Preprocessing
After moving data to Odin, dcmdb uses dcm2dcm to transcode from transfer syntax 1.2.840.10008.1.2.4.90 (JPEG 2000 lossless) to transfer syntax 1.2.840.10008.1.2.1 (Explicit VR little-endian uncompressed). JPEG 2000 lossless should not be confused with JPEG or .jpg files. JPEG 2000 is a compute-intensive, rarely used format. It is however how our PACSs store data internally. Using lossless compression speeds up transfers and saves space on the server. Transcodeing (decompressing) to transfer syntax 1.2.840.10008.1.2.1 is often a first step in processing your data with any tool set. Standard preprocessing converts DICOM to NIfTI using mcverter, the command line version of MRIConvert.
 * -o
 * Set output directory
 * --format=fsl
 * Output fsl formatted NifTi
 * -d
 * Save output volumes as 4D files
 * -r
 * Apply rescale slope and intercept to data
 * --nii
 * Save output files with .nii at the end.

The output is gziped.

Output is 4D .nii.gz with rescale and slope applied.

Automated Bedpostx
Automated Bedposting is presently disabled. Please contact Andrew if you would like it turned back on.

To use Automated FSL bedpostx pre-proccessing, set Referring Physician to bedpost or set the patient comment to Bepost. This instructs dcmb to place your data in the bedpost folder. Dcmb converts all dicom data in the bedpost folder to NifTi format.

For DTI scans, dcmb will also For details look at the shell script or ask Andrew.
 * Run eddy_correct
 * Copy the first volume to nodiff.nii.gz
 * Remove the last volume.
 * Remove the last column from the bval and bvec files.
 * Run bet.
 * Run dtifit.
 * Run bedpostx.

Neuropatient folder
Neuropatient is deprecated. Data in the neuropatient folder receives standard preprocessing.