"""
Cartesian Kspace-Generators.
At each iteration, a column of kspace is yielded.
Transpose your data if you want to work row-wise.
"""
import numpy as np
from .base import KspaceGeneratorBase
[docs]class Column2DKspaceGenerator(KspaceGeneratorBase):
"""K-Space generator, at each step a new column fills the existing k-space.
Parameters
----------
full_kspace: numpy.ndarray
Complete kspace_data.
mask_cols: array_like
List of the column indices to use for the mask.
max_iter: int, optional
The maximum number of iteration to do, default is
the number of column provided.
mode: {"line", "current", "memory"}
If "line" : at step k, the generator yields only the
data of k-th column, and the column number associated.
If "current": at step k, the generator yields an array
with the shape of full_kspace, but containing only the data
of step k, and the mask associated.
If "memory", same as "current", but all the data that have
been previously acquired fills the kspace-array.
Default is "current".
start_center: bool, optional
Should the acquisition defined by mask_cols be reordered
to start from the center of kspace and move outward by alternating
left and right. (default: True)
"""
def __init__(self, full_kspace, mask_cols, max_iter=0, mode="current", start_center=True):
mask = np.zeros(full_kspace.shape[-2:])
if mode == "line":
self.acquire = self._getitem_line
elif mode == "current":
self.acquire = self._getitem_current
elif mode == "memory":
self.acquire = self._getitem_memory
else:
raise ValueError("Unknown mode of acquisition.")
self.cols = np.asarray(mask_cols)
if start_center:
center_pos = np.argmin(np.abs(mask_cols - full_kspace.shape[-1] // 2))
mask_cols = list(mask_cols)
left = mask_cols[center_pos::-1]
right = mask_cols[center_pos + 1:]
new_cols = []
while left or right:
if left:
new_cols.append(left.pop(0))
if right:
new_cols.append(right.pop(0))
self.cols = np.array(new_cols)
if max_iter == 0:
max_iter = len(self.cols)
super().__init__(full_kspace, mask, max_iter=max_iter)
[docs] def _getitem_memory(self, idx):
mask = np.zeros(self.shape[-2:])
mask[:, self.cols[:idx+1]] = 1
kspace = np.squeeze(self._full_kspace * mask[np.newaxis, ...])
return kspace, mask
[docs] def _getitem_current(self, idx: int):
mask = np.zeros(self.shape[-2:])
mask[:, self.cols[idx]] = 1
kspace = np.squeeze(self._full_kspace * mask[np.newaxis, ...])
return kspace, mask
[docs] def _getitem_line(self, idx: int):
col = self.cols[idx]
kspace = self.kspace[..., col]
return kspace, col
def __getitem__(self, it):
if it > self._len:
raise IndexError
idx = min(it, len(self.cols)-1)
return self.acquire(idx)
def __next__(self):
if self.iter > self._len:
raise StopIteration
self.iter += 1
return self[self.iter-1]
