Spinal cord MRI acquisition (improvement and validation)

MRI Acquisition
#1

Hi Julien and everyone,

Hope you are fine. The MRI scanner that we have been using so far is: 3T Siemens Trio (VB17A), and we have been utilizing it to acquire different MRI modalities.
Are there any standard pulse sequence parameters to use in 3T Siemens that are specific to acquire each MRI modality of the spinal cord (some fixed parameters for DWI, T2w, MT, etc.)? The SOP which you have made available is great and very helpful to position the slice, the saturation band and the shim box, etc. But in the end, the quality of the acquired images sometimes do not seem to be great for further analysis.
Also if one wants to understand that the acquired image is good enough to work on, and to make a meaningful decision based on, how it can be validated?
It would be appreciated if you could suggest anything to lead us to acquire more useful MRI images.

Regards,
Maryam

#2

Hi @Maryam,

Does the SOP you are referring to belong to the generic protocol? If so, could you please elaborate on “the quality of the acquired images sometimes do not seem to be great for further analysis”? E.g., what exactly is not working, what kind of analysis are you doing? Are you using the protocol “out of the box” or are you modifying it and if so how?

Regarding your question about how can we validate an imaging protocol, this is a tough question to answer, as it depends on what metrics researchers are interested in computing down the line, and what is the expected effect size. To give you an example, if you are interested in detecting a 50% change in FA in the dorsal column, then the proposed protocol is likely sufficient for this purpose, given that the precision (intra- and inter-subject, within vendor) will give you statistical significance, according to our recent results from the multi-center dataset (see pages 31-41).

So, the recommended approach is to identify a few hypotheses, anticipate an effect size (from literature or pilot study), and compare against the precision of the proposed protocol. Then, adapt the protocol from there.

Best,
Julien

1 Like
#3

Hi Julien,

I had a meeting with the radiologist last week, so now I can answer your question better. They use the generic protocol, but there might be some changes depending on the MRI machine (do not know what changes exactly). They also mentioned some difficulties during imaging, because patients cannot stay fixed, so there are sometimes some motions in the images, that’s why images might not seem great always. By the way, the images would be used as a portion of the data set for an ML/DL model, that’s why I want the acquisition to be good enough and make sure that the images are completely correct (are taken correctly).
Thanks very much for your response and useful recommendations.

Regards,
Maryam

#4

They also mentioned some difficulties during imaging, because patients cannot stay fixed, so there are sometimes some motions in the images, that’s why images might not seem great always.

If motion is a problem, then I recommend you focus on sequences that are less sensitive to motion, such as the T2 CUBE (GE), VISTA (Philips) or SPACE (Siemens).

1 Like