Using the B0 tissue response for multi-tissue multicompartment models on higher b-shells only

[villalonreina]

Hi @rutgerfick,
Last we spoke we had discussed splitting up shells for calculations of various compartments. We have a dataset with b=0,500,1000,2000 and we were going to try and use the b=0,500,1000 for the ball and the b=500,1000,2000 for the stick. However, we were not sure if it was sound, in a multi-tissue model, to use the tissue response from the b=0 for the b=500,1000,2000 stick?

[rutgerfick]

Great question! and an interesting approach.

So, in dmipy's multi-tissue modeling the tissue response that you can put for any compartment is always the one that is calculated from the b=0 shell. This is because they are related to the TE-weighted (non-diffusion-weighted) signal response of a certain tissue.

Moreover, I don't think it is currently possible to fit data that has no b=0 measurements included...

Can you remind me why you wanted to avoid giving the b0 for the stick?

[rutgerfick]

To clarify on why the b0 is required. Regardless of whether we give the tissue response of any model, we still need an estimate of what is the S0 of the signal to properly relate the signal attenuation of the models with the actual measured diffusion weighted signal.

Currently this S0 estimate is recovered as the mean of the b0's in the dataset. With elbow grease, one could add S0 as an optimizable parameter and allow data without b0s, but let's first explore why you want to avoid using it :-).