Dear All,
we have recently seen examples on the mailbase of people collecting a large proportion of b=0 scans as part of their diffusion studies. We would therefore like to take the opportunity to give our humble opinion on the topic.
First of all: There isn't one correct number/proportion. It depends on
1. The purpose of the study: If you plan to do tractography you should use a smaller proportion of b=0 scans than if you want to measure mean diffusivity.
2. The SNR of your study: If you have VERY poor SNR (as in for example some in vitro studies) you might need a higher proportion of b=0 scans than would otherwise be the case.
For the rest of the discussion we will assume that you have a decent SNR and we will divide it into tractography and "TBSS style studies".
Tractography:
For tractography the important information lies in how the signal varies as you vary the direction of the diffusion weighting. You could do _almost_ as good tractography without having a single b=0 scan as part of your study. For this type of studies the "good old fmrib protocol" had 60 diffusion weighted volumes and 3 b=0, i.e. the proportion of b=0 scans were 1/21. We have seen no reason to revise that number upwards. In the past there might have been an argument that interspersing with b=0 scans helped with subject motion correction, but with the new eddy that is no longer so much of an issue.
TBSS style analysis:
With "TBSS style" we don't mean you have to use TBSS. We merely mean studies where the main objective is to measure the diffusivity (or some derived measure) for subsequent comparison to other subjects. If the main response variable is mean diffusivity (e.g. in Stroke diagnosis) it makes sense to have a larger number of b=0 scans, maybe even close to 1/2. If the main variable is FA we need to characterise the variability of the diffusion signal in addition to its mean and we will again benefit from a larger number of diffusion weighted volumes (i.e. a smaller proportion of b=0). For these types of studies we would recommend between 1/5 and 1/10 of b=0 scans.
Jesper, Stam and Saad
|