nlinv / jsense

[uecker]

As pointed out by somebody to me at the Sedona Workshop who was confused about this: SigPy calls the regularized bilinear recon JSENSE. But the version with regularization for image and coil sensitivities is NLINV. This is a bit unfortunate, in my opinion.

[frankong]

Hi Martin :)
nlinv is cited in the references:

sigpy/sigpy/mri/app.py

Lines 234 to 237 in 8d2030d

	Uecker, M., Hohage, T., Block, K. T., & Frahm, J. (2008).
	Image reconstruction by regularized nonlinear inversion-
	joint estimation of coil sensitivities and image content.
	Magnetic Resonance in Medicine, 60(#), 674-682.

By default, regularization is turned off with lambda=0, so that's JSENSE.

I can merge a pull request, If you have suggestions on a clearer docstring.

[uecker]

Am Montag, dem 03.04.2023 um 22:41 -0700 schrieb Frank Ong:

Hi Martin :) nlinv is cited in the references: https://github.com/mikgroup/sigpy/blob/8d2030d98fe1d6199f430828f1d8c67c1502ad56/sigpy/mri/app.py#L234-L237 By default, regularization is turned off with lambda=0, so that's JSENSE. I can merge a pull request, If you have suggestions on a clearer docstring.

Maybe add something like this: "This formulation with regularization corresponds to the version described in the NLINV paper. Without regularization (which is the default) it corresponds to the version from the JSENSE paper but using a truncated Fourier representation of the coils (as in NLINV) instead of polynomials." (assuming this is what it does) Martin

…

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: ***@***.***>

[uecker]

Am Montag, dem 03.04.2023 um 22:41 -0700 schrieb Frank Ong:

Hi Martin :) nlinv is cited in the references: https://github.com/mikgroup/sigpy/blob/8d2030d98fe1d6199f430828f1d8c67c1502ad56/sigpy/mri/app.py#L234-L237 By default, regularization is turned off with lambda=0, so that's JSENSE. I can merge a pull request, If you have suggestions on a clearer docstring.

Maybe add something like this: "This formulation with regularization corresponds to the version described in the NLINV paper. Without regularization (which is the default) this corresponds to the version from the JSENSE paper but using a truncated Fourier representation of the coils (as in NLINV) instead of polynomials." (assuming this is what it does) Martin

…

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: ***@***.***>

[frankong]

Addressed in #122

[uecker]

Thanks Frank!