Difference between single versus multi shell diffusion weighted imaging


I’m having a hard time distinguishing between single versus multi-shell schemes. I know in the bval.txt for the dwi data I’m looking at there are multiple sets of the following values: 100, 200, 250, 350, 450, 550, 650, 900, and 1000. Would I be able to tell from the bval.txt doc if the dwi scans were multi shell (if there multiple different values) or single shell (if there were only three types of values)?


The number of shells is how many unique non b=0 b-values you have. So it looks like you have a multi-shell acquisition.


Another question, if the bvalue types were just 995 and 1000, would that be a single shell acquisition? If so, would the write-up be something like → For each subject, we collected 2 runs of diffusion-weighted imaging per session with two single shells schemes (b = 995 and 1000s/mm2) using echo planar imaging along collinear direction, and scanning each direction 116 and 140 times, respectively.

Also for the multi shell part: I noticed the values were interleaved with the 200 and 250 values (e.g., 200 250 200 200 250), but all other values were in sequential format (e.g., 350 350 350 450 450 450 450 550 550 550 550 650 650 650 650 650 900 900 900 1000 1000 1000). Would the write up be something like → we collected 5 runs of diffusion-weighted imaging per session with multiple shells (i.e., b-values) (b = [100] s/mm2, [200, 250] s/mm2, [350] s/mm2, [450]s/mm2, [550]s/mm2, [650]s/mm2, [950]s/mm2, and [1000] s/mm2) using fast-spin echo imaging along a collinear direction, scanning each direction 6, 12, 8, 6, 24, 24, 12, and 30 times.

No, those small differences will still constitute a single shell.

Each direction was scanned 116 or 140 times? Are you sure you do not mean there were 116 or 140 directions? A b-value is not a direction, but a gradient strength. The directions come from the bvec file.

Typically directions are non-collinear, I would confirm whether you meant to say collinear.

A b-val difference of 50, as opposed to 5, might be representing a different shell, so in that case I would not group 200/250 together. I would talk to anyone else in your group who has used or designed these sequences.

With diffusion directions, how would I go about counting the number of bvecs to each bvalue?

The number of occurrences of each b-value is the number of directions tested at that b-value.

So if there are about 50 1000-level b-values, then that would 50 diffusion directions that correspond to the 1000 b-value?

Yup! Assuming non of the directions overlap with each other, but that is usually not the case.