From af80b2cd13b465b52a05219f7e68228685800039 Mon Sep 17 00:00:00 2001 From: Philippe Ciuciu Date: Fri, 28 Jun 2024 12:44:46 +0200 Subject: [PATCH] Update mrinufft_convention.rst Fixed typos. --- docs/mrinufft_convention.rst | 20 ++++++++++---------- 1 file changed, 10 insertions(+), 10 deletions(-) diff --git a/docs/mrinufft_convention.rst b/docs/mrinufft_convention.rst index dcabab6e..2558cefc 100644 --- a/docs/mrinufft_convention.rst +++ b/docs/mrinufft_convention.rst @@ -9,9 +9,9 @@ The NUFFT Operator ================== -MRI-NUFFT provides a common interface for computing NUFFT, regardless of the computation backend choosen. All the backend implements the followings methods and attributes. More technical details are available in the API reference. +MRI-NUFFT provides a common interface for computing NUFFT, regardless of the chosen computation backend. All the backends implement the following methods and attributes. More technical details are available in the API reference. -All MRI-NUFFT operator inherits from :class:`FourierOperatorBase` . The minimum signature of an MRI nufft operator is: +All MRI-NUFFT operators inherit from :class:`FourierOperatorBase` . The minimum signature of an MRI nufft operator is: .. code-block:: python @@ -37,7 +37,7 @@ All MRI-NUFFT operator inherits from :class:`FourierOperatorBase` . The minimum .. tip:: - The precision of the samples array will determine the precision of the computation. See also :ref:`K-Space Trajectories` + The precision of the samples array will determine the precision of the computation. See also :ref:`K-Space Trajectories` Moreover, the two following methods should be implemented for each backend @@ -49,33 +49,33 @@ After initialization, defaults for the following methods are available, as well * ``data_consistency(image, obs_data)``: perform the data consistency step :math:`\cal{F}^*(\cal{F} x - y)` * ``get_lipschitz_cst(max_iter)``: Estimate the spectral radius of the auto adjoint operator :math:`\cal{F}^*\cal{F}` -If the NUFFT backend makes some optimization possible for these backend, these can be manually overriden. +If the NUFFT backend makes some optimization possible, these backends can be manually overriden. Extensions ---------- -The base NUFFT operators can be extended to add extra functionality. With MRI-NUFFT we already provides: +The base NUFFT operators can be extended to add extra functionality. With MRI-NUFFT we already provide: -- Off-Resonnance Correction operators, using subspace separation (Sutton et al. , ) +- Off-resonnance Correction operators, using subspace separation (Sutton et al. IEEE TMI 2005) - Auto-differentiation (for Deep Learning applications) Adding a NUFFT Backend ---------------------- -Adding a NUFFT backend to MRI-NUFFT should be easy. We recommend to check how other backend have been inplemented. CPU-based nufft interface can use the `FourierOperatorCPU` to minimize the boiler-plate. +Adding a NUFFT backend to MRI-NUFFT should be easy. We recommend to check how other backends have been inplemented. CPU-based nufft interface can use the `FourierOperatorCPU` to minimize the boiler-plate. K-Space Trajectories ==================== -The k-space sampling trajectories are generated in the :py:mod:`mrinufft.trajectories` module and then use in the differents backends. +The k-space sampling trajectories are generated in the :py:mod:`mrinufft.trajectories` module and then used in the differents backends. They are ``numpy`` arrays with the followings characteristics: - ``float32`` or ``float64`` precision (this will trigger the use of single or double precision in the computations using this trajectory). ``float32`` precision is recommended for computational efficiency. -- They are row major array (C Convention) with the following shape: ``(N_shot, N_samples, dim)`` (where dim is either 2 or 3). A "flatten" version of shape ``(N_shots * N_samples, dim)`` is also acceptable by operators. +- They are row-major array (C Convention) with the following shape: ``(N_shot, N_samples, dim)`` (where dim is either 2 or 3). A "flatten" version of shape ``(N_shots * N_samples, dim)`` is also acceptable by operators. -- By convention all k-space coordinate are contains in :math:`[-0.5,0.5)`, They will be rescaled internally by operator if required. +- By convention all k-space coordinates are in the range :math:`[-0.5,0.5)`. They will be rescaled internally by operator if required.