search

oesteban / diffantom

A whole-brain diffusion phantom using phantomas
MIT License
2 stars 1 forks source link

A more scientific way to set weights (APPENDIX) #9

Closed oesteban closed 8 years ago

oesteban commented 9 years ago

Do you know some references or a methodology to decide the values for w{cgm,dgm,wm} and w{f1, f2, f3} in the equations of the appendix?

oesteban commented 8 years ago

Reading these two references:

I find that we could define an additional contribution. We have so far the free-diffusion compartment (liquid) and the hidered-diffusion compartment (fibers). The additional fraction would be the no-diffusion compartment. This compartment contains the cell membranes and myelin sheaths.

I can imagine that this new compartment would provide the realistic point we lack now, because this no-diffusion compartment is prevalent across gray matter. It is also interesting to set it in the white matter, and in this case we have values from literature to be set.

oesteban commented 8 years ago

@mledesmacarbayo pointed out concerns about the way to set the weights in L141 "assessed visually" and L143 (in default w_i). I label this issue as a paper comment that we need to address.

oesteban commented 8 years ago

On 6 August 2015 at 14:23, @ecaruyer wrote:

I think those terms were originally coined by Assaf in his CHARMED paper [1], and the definitions therein are the most widely accepted in the community. As Ale mentioned in his comment, "hindered" refers to semi-free diffusion, where there are obstacles, such as in the extra-axonal compartment in white matter; while "restricted" refers to diffusion within a confining geometry, may it be a cylinder (intra-axonal diffusion) or a sphere (which is not relevant to us since not modeled in Phantomas).

Thanks a lot Manu, very illustrative :)

Then, if I understood, in general we can have isotropic and anisotropic diffusion if we talk about the "shape" of diffusion. If we talk in terms of micro-structure, we have hindered (that will be in general isotropic with low "speed" or low ADC, but can be anisotropic if obstacles define a prolate shape), restricted diffusion (generally due to axons, and always anisotropic), and free diffusion (isotropic and "fast").

Q1: Please, correct me if I'm wrong :)

Q2: Then, where should we place the concept of "no-diffusion"?. I've seen this in some NODDI papers and usually it refers to those regions containing cell membranes and other structures with no diffusion at all.

Q3: How do you think we could improve the model of the diffantom with these concepts (hindered-, restricted-, free-, no- diffusion)?. So far, I use Phantomas to simulate the restricted diffusion with the single fiber response and the free diffusion in areas of CSF. Finally, in GM we have a mixture of hindered- and no- diffusion using a free diffusion model with very low diffusivity. Is this correct in general? Am I missing something?

Q4: How do you see the FA as a surrogate of the volume fraction of the first (most prominent) fiber in a voxel?. I try to use it as such in the model.