Frequently asked questions for the PSMD toolbox.
If your cannot find an answer to your question here, please feel free to post at Discussions on GitHub.
If you found a bug or have a feature request, you can create an Issue on GitHub.
Clinical use
Can I use PSMD in the clinic to assess small vessel disease burden in my patients?
No, unfortunately not. PSMD is not a medical device. Therefore, it can be used in research studies only! Any application for diagnosis, prognosis or monitoring in the clinic is forbidden by law, e.g. the medical device regulation (MDR) in the EU.
MRI acquisition
What MRI sequences do I need to calculate PSMD?
Only a diffusion sequence with multiple directions ("diffusion tensor imaging sequence") is required.
We recommend the following parameters
- B-value around 1000 s/mm2 (between 700 and 1200 has been tested)
- At least 20 directions, ideally 30 or more
- 2 mm isotropic resolution recommended, but PSMD will also work with thicker slices
Caution
- Coverage of the entire hemispheres is needed. Cropped brains (usually on top) can be problematic and may even introduce a bias into your analysis (see "known problems" below).
- PSMD seems to work best with data from Siemens or GE scanners. Data from Philipps scanners showed problems in some studies, especially if there was a software update during the study (see below).
Known issues
Are there any known problems that might affect PSMD?
-
Scanner software upgrades can be problematic. We have seen this especially with Philips scanners. Even minor software upgrades can have a strong effect on the MD histogram and therefore PSMD. We recommend to check for different software versions (even minor differences) in your data, especially if you have Philips data. A potential solution is to use mean skeletonized mean diffusivity (MSMD), which is available in version 1.5.
-
Another issue is when the hemispheres are not covered completely by the field-of-view. If e.g. the top is cropped, this will alter the histogram shape. If only large brains are cropped in your study, you might introduce a systematic bias.
-
Acute infarcts will affect the PSMD calculation. Version 1.5 and newer of the PSMD tool allows to automatically exclude DWI hyperintense lesions (acute infarcts) based on their appearance on the raw diffusion images. Consequently, this option is only available when starting with raw data. In addition, you can supply your own lesion mask (exclude mask) or analyse only the non-affected hemisphere (there is a built-in option to analyze left/right separately).
Different results using different FSL versions or operating systems
Different environments, such as FSL version or operating system, can indeed cause slight differences in PSMD/MSMD values. In order to get a consistent environment and results, consider using the PSMD Singularity container.
Also, make sure to always use the same version/release of PSMD to get consistent results.
Troubleshooting
- dc not installed: On some Linux operating systems, the PSMD script might abort. Using the verbose option
-v, you might see the error messagedc not found. In this case, the programdc(which is needed by FSL) is not installed. You can correct this by installing with your package manager, e.g. for Ubuntu:sudo apt -y install dc. Such dependency problems can be avoided completely by using the container version.