XA30_workaround

This repo contains some scripts to get around the limitations of the SIEMENS XA30A image database implementation, specifically system crashes or dramatically degraded responsivity when acquiring multi-echo BOLD.

When running multi-echo sequences with many measurements (like the CMRR BOLD sequence), the XA30A image database cannot handle the load.

Note

The log output in UTraceSrv.utr suggests that the problem arises right after the usual ICE reconstruction that ends with ImageSend, during which DICOMs are written into the image database, because the log file showed function names like “DICOM backend writer”.

Our solution for this was to send the reconstructed images for only the very first echo to the DICOM backend writer, with the rest of the reconstructed echoes (including the first echo for comparison purposes) written directly to a binary file that contains unsigned 16-bit integers. These binary files are of size #columns x #rows x #slices x #echoes, and contain just the reconstructed (image-domain) data for 1 measurement instance (either magnitude or phase)​.

Note

The solution is embedded into a modified version of the ICE program (.ipr file) for the CMRR BOLD sequence, and gets activated only if the scan has “MBME_DcmOnlyE1” anywhere in its name.

To copy the extra binary files written by the ICE program, you can use the dat_copier.ps1 PowerShell script provided. The user runs it and enters the patient ID to initiate the process.

This repo also contains files for converting the generated binary .dat files (not to be confused with the raw k-space .dat files) into .nii files. It uses dcm2niix and the header information available in the existing .dcm files (corresponding to the first echoes) to construct a NIFTI version of all the echoes.

Installation

This repo contains two sets of files:

1. The first is a PowerShell script located under host_scripts and is intended to be installed on your scanner console. This script is called dat_copier.ps1 and is intended to be run on the scanner console when the session is completed. It copies .dat files from the scanner to a location of your choice.

2. The second contains the dcmdat2niix script, which takes in .dat files alongside .dcm files and converts them to .nii files.

Installation instructions for host_scripts

STEPS FOR INSTALLING dat_copier.ps1.

Installation instructions for dcmdat2niix

First, ensure that dcm2niix is installed and on your path.

Note

dcm2niix is available from here. Version v1.0.20220720 or later is required.

To install dcmdat2niix, clone this repo and install using pip:

# install using pip and local path to repo
git clone https://github.com/vanandrew/XA30_workaround.git
cd XA30_workaround
pip install .

If the install is successful, you should see dcmdat2niix in your path. You can check this by running:

dcmdat2niix -h

Usage

Data organization

Once you've copied your data off the scanner, you should have a collection of .dcm and .dat files for your data. Each .dat file is named with a unique series instance UID. You can match this UID to the corresponding .dcm file by checking the header of one of the .dcm files for a particular series.

After identifying the corresponding series, place the correponding .dat files in the same folder. dcmdat2niix will auto detect the .dat files and use the corresponding .dcm files to construct the NIFTI files.

Running dcmdat2niix

You can use dcmdat2niix in the same way you would use dcm2niix. For example, if you invoked:

dcm2niix -z y -f %p_%t_%s -o /path/output /path/to/dicom/folder

you can invoke dcmdat2niix in the same way:

dcmdat2niix -z y -f %p_%t_%s -o /path/output /path/to/dicom/folder

Current limitations

1. dcmdat2niix currently only supports interleaved slices.
2. Since dcmdat2niix parses the verbose output of dcm2niix, the -v flag is forced on. Manually specifying -v on dcmdat2niix will result in an error.
3. Echo detection is currently done by looking at e# or echo# (e.g. e1 or echo1) in the filename. If you choose to use a different naming convention, this script will likely not work :(.
4. Metadata (JSON sidecar) for any subsequent echoes past the 1st echo is copied from the 1st echo. The only metadata information that is replaced is the EchoTime field, which is replaced with the appropriate echo time parameter for that echo. This is obtained from the alTE field in the DICOM header (this is not an actual DICOM tag so it is text parsed from the DICOM).