def _crop_to_param(aff0, aff, shape):
dim = aff0.shape[-1] - 1
shape = shape[:dim]
layout0 = spatial.affine_to_layout(aff0)
layout = spatial.affine_to_layout(aff)
if (layout0 != layout).any():
raise ValueError('Input and Ref do not have the same layout: '
f'{spatial.volume_layout_to_name(layout0)} vs '
f'{spatial.volume_layout_to_name(layout)}.')
size = shape
layout = None
unit = 'vox'
center = torch.as_tensor(shape, dtype=torch.float).sub_(1).mul_(0.5)
like_aff = spatial.affine_lmdiv(aff0, aff)
center = spatial.affine_matvec(like_aff, center)
return size, center, unit, layout
def get_readout(readout, affine, shape):
"""
Convert the provided readout dir from semantic (R/A/S) to index (0/1/2)
or
Guess the readout direction as the one with largest number of voxels
"""
dim = len(shape)
layout = spatial.affine_to_layout(affine)
layout = spatial.volume_layout_to_name(layout).lower()
if readout is None:
readout = py.argmax(shape)
else:
readout = readout.lower()
for i, l in enumerate(layout):
if l in readout:
readout = i
break
print(f'Layout: {layout.upper()} | readout direction: {layout[readout].upper()}')
if readout > 0:
readout = readout - dim
return readout
def crop(inp,
size=None,
center=None,
space='vx',
like=None,
bbox=False,
output=None,
transform=None):
"""Crop a ND volume, while preserving the orientation matrices.
Parameters
----------
inp : str or (tensor, tensor)
Either a path to a volume file or a tuple `(dat, affine)`, where
the first element contains the volume data and the second contains
the orientation matrix.
size : [sequence of] int, optional
Size of the patch to extract.
Its unit and axes are defined by `units` and `layout`.
center : [sequence of] int, optional
Coordinate of the center of the patch.
Its unit and axes are defined by `units` and `layout`.
By default, the center of the FOV is used.
space : [sequence of] {'vox', 'ras'}, default='vox'
The space in which the `size` and `center` parameters are expressed.
bbox : bool or float, default=False
Crop at the bounding box of `inp > threshold`.
If `bbox` is a float, it is the threshold to use.
If `bbox` is `True`, the threshold is 0.
like : str or (tensor, tensor), optional
Reference patch.
Either a path to a volume file or a tuple `(dat, affine)`, where
the first element contains the volume data and the second contains
the orientation matrix.
output : [sequence of] str, optional
Output filename(s).
If the input is not a path, the unstacked data is not written
on disk by default.
If the input is a path, the default output filename is
'{dir}/{base}.{i}{ext}', where `dir`, `base` and `ext`
are the directory, base name and extension of the input file,
`i` is the coordinate (starting at 1) of the slice.
transform : [sequence of] str, optional
Input or output filename(s) of the corresponding transforms.
Not written by default.
If a transform is provided and all other parameters
(i.e., `size` and `like`) are None, the transform is considered
as an input transform to apply.
Returns
-------
output : list[str or (tensor, tensor)]
If the input is a path, the output paths are returned.
Else, the unstacked data and orientation matrices are returned.
"""
dir = ''
base = ''
ext = ''
fname = None
transform_in = False
use_bbox = bool(bbox or isinstance(bbox, float))
# --- Open input ---
is_file = isinstance(inp, str)
if is_file:
fname = inp
f = io.volumes.map(inp)
inp = (f.data(numpy=True) if use_bbox else f, f.affine)
if output is None:
output = '{dir}{sep}{base}.crop{ext}'
dir, base, ext = py.fileparts(fname)
dat, aff0 = inp
dim = aff0.shape[-1] - 1
shape0 = dat.shape[:dim]
layout0 = spatial.affine_to_layout(aff0)
# save input space in case we reorient later
aff00 = aff0
shape00 = shape0
if bool(size) + bool(like) + bool(bbox or isinstance(bbox, float)) > 1:
raise ValueError('Can only use one of `size`, `like` and `bbox`.')
# --- Open reference and compute size/center ---
if like:
like_is_file = isinstance(like, str)
if like_is_file:
f = io.volumes.map(like)
like = (f.shape, f.affine)
like_shape, like_aff = like
like_layout = spatial.affine_to_layout(like_aff)
if (layout0 != like_layout).any():
aff0, dat = spatial.affine_reorient(aff0, dat, like_layout)
shape0 = dat.shape[:dim]
if torch.is_tensor(like_shape):
like_shape = like_shape.shape
size, center, unit, layout = _crop_to_param(aff0, like_aff, like_shape)
space = 'vox'
elif bbox or isinstance(bbox, float):
if bbox is True:
bbox = 0.
mask = torch.as_tensor(dat > bbox)
while mask.dim() > 3:
mask = mask.any(dim=-1)
mins = []
maxs = []
for d in range(dim):
n = mask.shape[d]
idx = utils.movedim(mask, d,
0).reshape([n, -1
]).any(-1).nonzero(as_tuple=False)
mins.append(idx.min())
maxs.append(idx.max())
mins = utils.as_tensor(mins)
maxs = utils.as_tensor(maxs)
size = maxs + 1 - mins
center = (maxs + 1 + mins).float() / 2
space = 'vox'
del mask
# --- Open transformation file and compute size/center ---
elif not size:
if not transform:
raise ValueError('At least one of size/like/transform must '
'be provided')
transform_in = True
t = io.transforms.map(transform)
if not isinstance(t, io.transforms.LinearTransformArray):
raise TypeError('Expected an LTA file')
like_aff, like_shape = t.destination_space()
size, center, unit, layout = _crop_to_param(aff0, like_aff, like_shape)
# --- use center of the FOV ---
if not torch.is_tensor(center) and not center:
center = torch.as_tensor(shape0[:dim], dtype=torch.float)
center = center.sub_(1).mul_(0.5)
# --- convert size/center to voxels ---
size = utils.make_vector(size, dim, dtype=torch.long)
center = utils.make_vector(center, dim, dtype=torch.float)
space_size, space_center = py.make_list(space, 2)
if space_center.lower() == 'ras':
center = spatial.affine_matvec(spatial.affine_inv(aff0), center)
if space_size.lower() == 'ras':
perm = spatial.affine_to_layout(aff0)[:, 0]
size = size[perm.long()]
size = size / spatial.voxel_size(aff0)
# --- compute first/last ---
center = center.float()
size = (size.ceil() if size.dtype.is_floating_point else size).long()
first = center - size.float().sub_(1).mul_(0.5)
first = first.round().long()
last = (first + size).tolist()
first = [max(f, 0) for f in first.tolist()]
last = [min(l, s) for l, s in zip(last, shape0[:dim])]
verb = 'Cropping patch ['
verb += ', '.join([f'{f}:{l}' for f, l in zip(first, last)])
verb += f'] from volume with shape {shape0[:dim]}'
print(verb)
slicer = tuple(slice(f, l) for f, l in zip(first, last))
# --- do crop ---
if use_bbox:
dat = dat.numpy()
dat = dat[slicer]
if not torch.is_tensor(dat):
dat = dat.data(numpy=True)
aff, _ = spatial.affine_sub(aff0, shape0[:dim], slicer)
shape = dat.shape[:dim]
if output:
if is_file:
output = output.format(dir=dir or '.',
base=base,
ext=ext,
sep=os.path.sep)
io.volumes.save(dat, output, like=fname, affine=aff)
else:
output = output.format(sep=os.path.sep)
io.volumes.save(dat, output, affine=aff)
if transform and not transform_in:
if is_file:
transform = transform.format(dir=dir or '.',
base=base,
ext=ext,
sep=os.path.sep)
else:
transform = transform.format(sep=os.path.sep)
trf = io.transforms.LinearTransformArray(transform, 'w')
trf.set_source_space(aff00, shape00)
trf.set_destination_space(aff, shape)
trf.set_metadata({
'src': {
'filename': fname
},
'dst': {
'filename': output
},
'type': 1
}) # RAS_TO_RAS
trf.set_fdata(torch.eye(4))
trf.save()
if is_file:
return output
else:
return dat, aff
def crop(inp,
size=None,
center=None,
space='vx',
like=None,
output=None,
transform=None):
"""Crop a ND volume, while preserving the orientation matrices.
Parameters
----------
inp : str or (tensor, tensor)
Either a path to a volume file or a tuple `(dat, affine)`, where
the first element contains the volume data and the second contains
the orientation matrix.
size : [sequence of] int, optional
Size of the patch to extract.
Its unit and axes are defined by `units` and `layout`.
center : [sequence of] int, optional
Coordinate of the center of the patch.
Its unit and axes are defined by `units` and `layout`.
By default, the center of the FOV is used.
space : [sequence of] {'vox', 'ras'}, default='vox'
The space in which the `size` and `center` parameters are expressed.
like : str or (tensor, tensor), optional
Reference patch.
Either a path to a volume file or a tuple `(dat, affine)`, where
the first element contains the volume data and the second contains
the orientation matrix.
output : [sequence of] str, optional
Output filename(s).
If the input is not a path, the unstacked data is not written
on disk by default.
If the input is a path, the default output filename is
'{dir}/{base}.{i}{ext}', where `dir`, `base` and `ext`
are the directory, base name and extension of the input file,
`i` is the coordinate (starting at 1) of the slice.
transform : [sequence of] str, optional
Input or output filename(s) of the corresponding transforms.
Not written by default.
If a transform is provided and all other parameters
(i.e., `size` and `like`) are None, the transform is considered
as an input transform to apply.
Returns
-------
output : list[str or (tensor, tensor)]
If the input is a path, the output paths are returned.
Else, the unstacked data and orientation matrices are returned.
"""
dir = ''
base = ''
ext = ''
fname = None
transform_in = False
# --- Open input ---
is_file = isinstance(inp, str)
if is_file:
fname = inp
f = io.volumes.map(inp)
inp = (f.data(numpy=True), f.affine)
if output is None:
output = '{dir}{sep}{base}.crop{ext}'
dir, base, ext = py.fileparts(fname)
dat, aff0 = inp
dim = aff0.shape[-1] - 1
shape0 = dat.shape[:dim]
if size and like:
raise ValueError('Cannot use both `size` and `like`.')
# --- Open reference and compute size/center ---
if like:
like_is_file = isinstance(like, str)
if like_is_file:
f = io.volumes.map(like)
like = (f.shape, f.affine)
like_shape, like_aff = like
if torch.is_tensor(like_shape):
like_shape = like_shape.shape
size, center, unit, layout = _crop_to_param(aff0, like_aff, like_shape)
# --- Open transformation file and compute size/center ---
elif not size:
if not transform:
raise ValueError('At least one of size/like/transform must '
'be provided')
transform_in = True
t = io.transforms.map(transform)
if not isinstance(t, io.transforms.LinearTransformArray):
raise TypeError('Expected an LTA file')
like_aff, like_shape = t.destination_space()
size, center, unit, layout = _crop_to_param(aff0, like_aff, like_shape)
# --- use center of the FOV ---
if not torch.is_tensor(center) and not center:
center = torch.as_tensor(shape0[:dim], dtype=torch.float) * 0.5
# --- convert size/center to voxels ---
size = utils.make_vector(size, dim, dtype=torch.long)
center = utils.make_vector(center, dim, dtype=torch.float)
space_size, space_center = py.make_list(space, 2)
if space_center.lower() == 'ras':
center = spatial.affine_matvec(spatial.affine_inv(aff0), center)
if space_size.lower() == 'ras':
perm = spatial.affine_to_layout(aff0)[:, 0]
size = size[perm.long()]
size = size / spatial.voxel_size(aff0)
# --- compute first/last ---
center = center.float()
size = size.ceil().long()
first = (center - size.float() / 2).round().long()
last = (first + size).tolist()
first = [max(f, 0) for f in first.tolist()]
last = [min(l, s) for l, s in zip(last, shape0[:dim])]
verb = 'Cropping patch ['
verb += ', '.join([f'{f}:{l}' for f, l in zip(first, last)])
verb += f'] from volume with shape {shape0[:dim]}'
print(verb)
slicer = tuple(slice(f, l) for f, l in zip(first, last))
# --- do crop ---
dat = dat[slicer]
aff, _ = spatial.affine_sub(aff0, shape0[:dim], slicer)
shape = dat.shape[:dim]
if output:
if is_file:
output = output.format(dir=dir or '.',
base=base,
ext=ext,
sep=os.path.sep)
io.volumes.save(dat, output, like=fname, affine=aff)
else:
output = output.format(sep=os.path.sep)
io.volumes.save(dat, output, affine=aff)
if transform and not transform_in:
if is_file:
transform = transform.format(dir=dir or '.',
base=base,
ext=ext,
sep=os.path.sep)
else:
transform = transform.format(sep=os.path.sep)
trf = io.transforms.LinearTransformArray(transform, 'w')
trf.set_source_space(aff0, shape0)
trf.set_destination_space(aff, shape)
trf.set_metadata({
'src': {
'filename': fname
},
'dst': {
'filename': output
},
'type': 1
}) # RAS_TO_RAS
trf.set_fdata(torch.eye(4))
trf.save()
if is_file:
return output
else:
return dat, aff
def orient(inp, affine=None, layout=None, voxel_size=None, center=None,
like=None, output=None, output_transform=None):
"""Overwrite the orientation matrix
Parameters
----------
inp : str or (tuple, tensor)
Either a path to a volume file or a tuple `(shape, affine)`, where
the first element contains the volume shape and the second contains
the orientation matrix.
affine : {'self', 'like'} or (4, 4) tensor_like, default='like'
Target affine matrix
layout : {'self', 'like'} or layout-like, default='like'
Target orientation.
voxel_size : {'self', 'like'} or [sequence of] float, default='like'
Target voxel size.
center : {'self', 'like'} or [sequence of] float, default='like'
World coordinate of the center of the field of view.
like : str or (tuple, tensor)
Either a path to a volume file or a tuple `(shape, affine)`, where
the first element contains the volume shape and the second contains
the orientation matrix.
output : str, optional
Output filename.
If the input is not a path, the reoriented data is not written
on disk by default.
If the input is a path, the default output filename is
'{dir}/{base}.{layout}{ext}', where `dir`, `base` and `ext`
are the directory, base name and extension of the input file.
output_transform : str, optional
Filename of output transform.
If the input is not a path, the reoriented data is not written
on disk by default.
If the input is a path, the default output filename is
'{dir}/{base}_to_{layout}.lta', where `dir` and `base`
are the directory and base name of the input file.
Returns
-------
output : str or (tuple, tensor)
If the input is a path, the output path is returned.
Else, the new shape and orientation matrix are returned.
"""
dir = ''
base = ''
ext = ''
fname = ''
is_file = isinstance(inp, str)
if is_file:
fname = inp
f = io.volumes.map(inp)
dim = f.affine.shape[-1] - 1
inp = (f.shape[:dim], f.affine)
if output is None:
output = '{dir}{sep}{base}.{layout}{ext}'
if output_transform is None:
output_transform = '{dir}{sep}{base}_to_{layout}.lta'
dir, base, ext = py.fileparts(fname)
like_is_file = isinstance(like, str) and like
if like_is_file:
f = io.volumes.map(like)
dim = f.affine.shape[-1] - 1
like = (f.shape[:dim], f.affine)
shape, aff0 = inp
dim = aff0.shape[-1] - 1
if like:
shape_like, aff_like = like
else:
shape_like, aff_like = (shape, aff0)
if voxel_size in (None, 'like') or len(voxel_size) == 0:
voxel_size = spatial.voxel_size(aff_like)
elif voxel_size == 'self':
voxel_size = spatial.voxel_size(aff0)
elif voxel_size == 'standard':
voxel_size = 1.
voxel_size = utils.make_vector(voxel_size, dim)
if not layout or layout == 'like':
layout = spatial.affine_to_layout(aff_like)
elif layout == 'self':
layout = spatial.affine_to_layout(aff0)
elif layout == 'standard':
layout = 'RAS'
layout = spatial.volume_layout(layout)
if center in (None, 'like') or len(center) == 0:
center = (torch.as_tensor(shape_like, dtype=torch.float) - 1) * 0.5
center = spatial.affine_matvec(aff_like, center)
elif center == 'self':
center = (torch.as_tensor(shape, dtype=torch.float) - 1) * 0.5
center = spatial.affine_matvec(aff0, center)
elif center == 'standard':
center = 0.
center = utils.make_vector(center, dim)
if affine in (None, 'like') or len(affine) == 0:
affine = aff_like
elif affine == 'self':
affine = aff0
elif affine == 'standard':
affine = torch.eye(dim+1, dim+1)
affine = torch.as_tensor(affine, dtype=torch.float)
if affine.numel() == dim*(dim+1):
affine = spatial.affine_make_rect(affine.reshape(dim, dim+1))
elif affine.numel() == (dim+1)**2:
affine = affine.reshape(dim+1, dim+1)
else:
raise ValueError(f'Input affine should have {dim*(dim+1)} or '
f'{(dim+1)**2} element but got {affine.numel()}.')
affine = spatial.affine_modify(affine, shape, voxel_size=voxel_size,
layout=layout, center=center)
affine = affine.double()
if output:
dat = io.volumes.load(fname, numpy=True)
layout = spatial.volume_layout_to_name(layout)
if is_file:
output = output.format(dir=dir or '.', base=base, ext=ext,
sep=os.path.sep, layout=layout)
io.volumes.save(dat, output, like=fname, affine=affine)
else:
output = output.format(sep=os.path.sep, layout=layout)
io.volumes.save(dat, output, affine=affine)
if output_transform:
transform = spatial.affine_rmdiv(affine, aff0)
output_transform = output_transform.format(
dir=dir or '.', base=base, sep=os.path.sep, layout=layout)
io.transforms.savef(transform.cpu(), output_transform, type=2)
if is_file:
return output
else:
return shape, affine
def _main(options):
if isinstance(options.gpu, str):
device = torch.device(options.gpu)
else:
assert isinstance(options.gpu, int)
device = torch.device(f'cuda:{options.gpu}')
if not torch.cuda.is_available():
device = 'cpu'
# prepare options
estatics_opt = ESTATICSOptions()
estatics_opt.likelihood = options.likelihood
estatics_opt.verbose = options.verbose >= 1
estatics_opt.plot = options.verbose >= 2
estatics_opt.recon.space = options.space
if isinstance(options.space, str) and options.space != 'mean':
for c, contrast in enumerate(options.contrast):
if contrast.name == options.space:
estatics_opt.recon.space = c
break
estatics_opt.backend.device = device
estatics_opt.optim.nb_levels = options.levels
estatics_opt.optim.max_iter_rls = options.iter
estatics_opt.optim.tolerance = options.tol
estatics_opt.regularization.norm = options.regularization
estatics_opt.regularization.factor = [*options.lam_intercept, options.lam_decay]
estatics_opt.distortion.enable = options.meetup
estatics_opt.distortion.bending = options.lam_meetup
estatics_opt.preproc.register = options.register
# prepare files
contrasts = []
distortion = []
for i, c in enumerate(options.contrast):
# read meta-parameters
meta = {}
if c.te:
te, unit = c.te, ''
if isinstance(te[-1], str):
*te, unit = te
if unit:
if unit == 'ms':
te = [t * 1e-3 for t in te]
elif unit not in ('s', 'sec'):
raise ValueError(f'TE unit: {unit}')
if c.echo_spacing:
delta, *unit = c.echo_spacing
unit = unit[0] if unit else ''
if unit == 'ms':
delta = delta * 1e-3
elif unit not in ('s', 'sec'):
raise ValueError(f'echo spacing unit: {unit}')
ne = sum(io.map(f).unsqueeze(-1).shape[3] for f in c.echoes)
te = [te[0] + e*delta for e in range(ne)]
meta['te'] = te
# map volumes
contrasts.append(qio.GradientEchoMulti.from_fname(c.echoes, **meta))
if c.readout:
layout = spatial.affine_to_layout(contrasts[-1].affine)
layout = spatial.volume_layout_to_name(layout)
readout = None
for j, l in enumerate(layout):
if l.lower() in c.readout.lower():
readout = j - 3
contrasts[-1].readout = readout
if c.b0:
bw = c.bandwidth
b0, *unit = c.b0
unit = unit[-1] if unit else 'vx'
fb0 = b0.map(b0)
b0 = fb0.fdata(device=device)
b0 = spatial.reslice(b0, fb0.affine, contrasts[-1][0].affine,
contrasts[-1][0].shape)
if unit.lower() == 'hz':
if not bw:
raise ValueError('Bandwidth required to convert fieldmap'
'from Hz to voxel')
b0 /= bw
b0 = DenseDistortion(b0)
distortion.append(b0)
else:
distortion.append(None)
# run algorithm
[te0, r2s, *b0] = estatics(contrasts, distortion, opt=estatics_opt)
# write results
# --- intercepts ---
odir0 = options.odir
for i, te1 in enumerate(te0):
ifname = contrasts[i].echo(0).volume.fname
odir, obase, oext = py.fileparts(ifname)
odir = odir0 or odir
obase = obase + '_TE0'
ofname = os.path.join(odir, obase + oext)
io.savef(te1.volume, ofname, affine=te1.affine, like=ifname, te=0, dtype='float32')
# --- decay ---
ifname = contrasts[0].echo(0).volume.fname
odir, obase, oext = py.fileparts(ifname)
odir = odir0 or odir
io.savef(r2s.volume, os.path.join(odir, 'R2star' + oext), affine=r2s.affine, dtype='float32')
# --- fieldmap + undistorted ---
if b0:
b0 = b0[0]
for i, b01 in enumerate(b0):
ifname = contrasts[i].echo(0).volume.fname
odir, obase, oext = py.fileparts(ifname)
odir = odir0 or odir
obase = obase + '_B0'
ofname = os.path.join(odir, obase + oext)
io.savef(b01.volume, ofname, affine=b01.affine, like=ifname, te=0, dtype='float32')
for i, (c, b) in enumerate(zip(contrasts, b0)):
readout = c.readout
grid_up, grid_down, jac_up, jac_down = b.exp2(
add_identity=True, jacobian=True)
for j, e in enumerate(c):
blip = e.blip or (2*(j % 2) - 1)
grid_blip = grid_down if blip > 0 else grid_up # inverse of
jac_blip = jac_down if blip > 0 else jac_up # forward model
ifname = e.volume.fname
odir, obase, oext = py.fileparts(ifname)
odir = odir0 or odir
obase = obase + '_unwrapped'
ofname = os.path.join(odir, obase + oext)
d = e.fdata(device=device)
d, _ = pull1d(d, grid_blip, readout)
d *= jac_blip
io.savef(d, ofname, affine=e.affine, like=ifname)
del d
del grid_up, grid_down, jac_up, jac_down
if options.register:
for i, c in enumerate(contrasts):
for j, e in enumerate(c):
ifname = e.volume.fname
odir, obase, oext = py.fileparts(ifname)
odir = odir0 or odir
obase = obase + '_registered'
ofname = os.path.join(odir, obase + oext)
io.save(e.volume, ofname, affine=e.affine)
def _ras_to_layout(x, affine):
"""Guess layout (e.g. "RAS") from an affine matrix"""
layout = spatial.affine_to_layout(affine)
ras_to_layout = layout[..., 0]
return [x[i] for i in ras_to_layout]
def info(inp, meta=None, stat=False):
"""Print information on a volume.
Parameters
----------
inp : str or (tensor, tensor)
Either a path to a volume file or a tuple `(dat, affine)`, where
the first element contains the volume data and the second contains
the orientation matrix.
meta : sequence of str
List of fields to print.
By default, a list of common fields is used.
stat : bool, default=False
Compute intensity statistics
"""
meta = meta or []
metadata = {}
is_file = isinstance(inp, str)
if is_file:
fname = inp
f = io.volumes.map(inp)
if stat:
inp = (f.fdata(), f.affine)
else:
inp = (f.shape, f.affine)
metadata = f.metadata(meta)
metadata['dtype'] = f.dtype
dat, aff = inp
if not is_file:
metadata['dtype'] = dat.dtype
if torch.is_tensor(dat):
shape = dat.shape
else:
shape = dat
pad = max([0] + [len(m) for m in metadata.keys()])
if not meta:
more_fields = ['shape', 'layout', 'filename']
pad = max(pad, max(len(f) for f in more_fields))
title = lambda tag: ('{tag:' + str(pad) + 's}').format(tag=tag)
if not meta:
if is_file:
print(f'{title("filename")} : {fname}')
print(f'{title("shape")} : {tuple(shape)}')
layout = spatial.affine_to_layout(aff)
layout = spatial.volume_layout_to_name(layout)
print(f'{title("layout")} : {layout}')
center = torch.as_tensor(shape[:3], dtype=torch.float) / 2
center = spatial.affine_matvec(aff, center)
print(f'{title("center")} : {tuple(center.tolist())} mm (RAS)')
if stat and torch.is_tensor(dat):
chandim = list(range(3, dat.ndim))
if not chandim:
vmin = dat.min().tolist()
vmax = dat.max().tolist()
vmean = dat.mean().tolist()
else:
dat1 = dat.reshape([-1, *chandim])
vmin = dat1.min(dim=0).values.tolist()
vmax = dat1.max(dim=0).values.tolist()
vmean = dat1.mean(dim=0).tolist()
print(f'{title("min")} : {vmin}')
print(f'{title("max")} : {vmax}')
print(f'{title("mean")} : {vmean}')
for key, value in metadata.items():
if value is None and not meta:
continue
if torch.is_tensor(value):
value = str(value.numpy())
value = value.split('\n')
value = ('\n' + ' ' * (pad + 3)).join(value)
print(f'{title(key)} : {value}')
def _prepare(data, dist, opt):
# --- options ------------------------------------------------------
# we deepcopy all options so that we can overwrite/simplify them in place
opt = ESTATICSOptions().update(opt).cleanup_()
backend = dict(dtype=opt.backend.dtype, device=opt.backend.device)
# --- be polite ----------------------------------------------------
if len(data) > 1:
pstr = f'Fitting a (shared) exponential decay model with {len(data)} contrasts.'
else:
pstr = f'Fitting an exponential decay model.'
print(pstr)
print('Echo times:')
for i, contrast in enumerate(data):
print(f' - contrast {i:2d}: [' +
', '.join([f'{te*1e3:.1f}' for te in contrast.te]) + '] ms')
# --- estimate noise / register / initialize maps ------------------
data, maps, dist = preproc(data, dist, opt)
nb_contrasts = len(maps) - 1
if opt.distortion.enable:
print('Readout directions:')
for i, contrast in enumerate(data):
layout = spatial.affine_to_layout(contrast.affine)
layout = spatial.volume_layout_to_name(layout)
readout = layout[contrast.readout]
readout = ('left-right' if 'L' in readout or 'R' in readout else
'infra-supra' if 'I' in readout or 'S' in readout else
'antero-posterior'
if 'A' in readout or 'P' in readout else 'unknown')
print(f' - contrast {i:2d}: {readout}')
# --- prepare regularization factor --------------------------------
# 1. Parameter maps regularization
# -> we want lam = [*lam_intercepts, lam_decay]
*lam, lam_decay = opt.regularization.factor
lam = core.py.make_list(lam, nb_contrasts)
lam.append(lam_decay)
if not any(lam):
opt.regularization.norm = ''
opt.regularization.factor = lam
# 2. Distortion fields regularization
lam_dist = dict(factor=opt.distortion.factor,
absolute=opt.distortion.absolute,
membrane=opt.distortion.membrane,
bending=opt.distortion.bending)
opt.distortion.factor = lam_dist
# --- initialize weights (RLS) -------------------------------------
mean_shape = maps.decay.volume.shape
rls = None
if opt.regularization.norm.endswith('tv'):
rls_shape = mean_shape
if opt.regularization.norm == 'tv':
rls_shape = (len(maps), *rls_shape)
rls = ParameterMap(rls_shape, fill=1, **backend).volume
if opt.regularization.norm:
print('Regularization:')
print(f' - type: {opt.regularization.norm.upper()}')
print(f' - log intercepts: [' +
', '.join([f'{i:.3g}' for i in lam[:-1]]) + ']')
print(f' - decay: {lam[-1]:.3g}')
else:
print('Without regularization')
if opt.distortion.enable:
print('Distortion correction:')
print(f' - model: {opt.distortion.model.lower()}')
print(
f' - absolute: {opt.distortion.absolute * opt.distortion.factor["factor"]}'
)
print(
f' - membrane: {opt.distortion.membrane * opt.distortion.factor["factor"]}'
)
print(
f' - bending: {opt.distortion.bending * opt.distortion.factor["factor"]}'
)
else:
print('Without distortion correction')
# --- initialize nb of iterations ----------------------------------
if not opt.regularization.norm.endswith('tv'):
# no reweighting -> do more gauss-newton updates instead
opt.optim.max_iter_prm *= opt.optim.max_iter_rls
opt.optim.max_iter_rls = 1
print('Optimization:')
print(f' - Tolerance: {opt.optim.tolerance}')
if opt.regularization.norm.endswith('tv'):
print(f' - IRLS iterations: {opt.optim.max_iter_rls}')
print(f' - Param iterations: {opt.optim.max_iter_prm}')
if opt.distortion.enable:
print(f' - Dist iterations: {opt.optim.max_iter_dist}')
print(f' - FMG cycles: 2')
print(f' - CG iterations: {opt.optim.max_iter_cg}'
f' (tolerance: {opt.optim.tolerance_cg})')
if opt.optim.nb_levels > 1:
print(f' - Levels: {opt.optim.nb_levels}')
# ------------------------------------------------------------------
# MAIN OPTIMIZATION LOOP
# ------------------------------------------------------------------
if opt.verbose:
pstr = f'{"rls":^3s} | {"gn":^3s} | {"step":^4s} | '
pstr += f'{"fit":^12s} + {"reg":^12s} + {"rls":^12s} '
if opt.distortion.enable:
pstr += f'+ {"dist":^12s} '
pstr += f'= {"crit":^12s}'
if opt.optim.nb_levels > 1:
pstr = f'{"lvl":3s} | ' + pstr
print('\n' + pstr)
print('-' * len(pstr))
return data, maps, dist, opt, rls
def orient(inp,
layout=None,
voxel_size=None,
center=None,
like=None,
output=None):
"""Overwrite the orientation matrix
Parameters
----------
inp : str or (tuple, tensor)
Either a path to a volume file or a tuple `(shape, affine)`, where
the first element contains the volume shape and the second contains
the orientation matrix.
layout : str or layout-like, default=None (= preserve)
Target orientation.
voxel_size : [sequence of] float, default=None (= preserve)
Target voxel size.
center : [sequence of] float, default=None (= preserve)
World coordinate of the center of the field of view.
like : str or (tuple, tensor)
Either a path to a volume file or a tuple `(shape, affine)`, where
the first element contains the volume shape and the second contains
the orientation matrix.
output : str, optional
Output filename.
If the input is not a path, the reoriented data is not written
on disk by default.
If the input is a path, the default output filename is
'{dir}/{base}.{layout}{ext}', where `dir`, `base` and `ext`
are the directory, base name and extension of the input file.
Returns
-------
output : str or (tuple, tensor)
If the input is a path, the output path is returned.
Else, the new shape and orientation matrix are returned.
"""
dir = ''
base = ''
ext = ''
fname = ''
is_file = isinstance(inp, str)
if is_file:
fname = inp
f = io.volumes.map(inp)
dim = f.affine.shape[-1] - 1
inp = (f.shape[:dim], f.affine)
if output is None:
output = '{dir}{sep}{base}.{layout}{ext}'
dir, base, ext = py.fileparts(fname)
like_is_file = isinstance(like, str) and like
if like_is_file:
f = io.volumes.map(like)
dim = f.affine.shape[-1] - 1
like = (f.shape[:dim], f.affine)
shape, aff0 = inp
dim = aff0.shape[-1] - 1
if like:
shape_like, aff_like = like
else:
shape_like, aff_like = (shape, aff0)
if voxel_size in (None, 'like') or len(voxel_size) == 0:
voxel_size = spatial.voxel_size(aff_like)
elif voxel_size == 'self':
voxel_size = spatial.voxel_size(aff0)
voxel_size = utils.make_vector(voxel_size, dim)
if not layout or layout == 'like':
layout = spatial.affine_to_layout(aff_like)
elif layout == 'self':
layout = spatial.affine_to_layout(aff0)
layout = spatial.volume_layout(layout)
if center in (None, 'like') or len(voxel_size) == 0:
center = torch.as_tensor(shape_like, dtype=torch.float) * 0.5
center = spatial.affine_matvec(aff_like, center)
elif center == 'self':
center = torch.as_tensor(shape, dtype=torch.float) * 0.5
center = spatial.affine_matvec(aff0, center)
center = utils.make_vector(center, dim)
aff = spatial.affine_default(shape,
voxel_size=voxel_size,
layout=layout,
center=center,
dtype=torch.double)
if output:
dat = io.volumes.load(fname, numpy=True)
layout = spatial.volume_layout_to_name(layout)
if is_file:
output = output.format(dir=dir or '.',
base=base,
ext=ext,
sep=os.path.sep,
layout=layout)
io.volumes.save(dat, output, like=fname, affine=aff)
else:
output = output.format(sep=os.path.sep, layout=layout)
io.volumes.save(dat, output, affine=aff)
if is_file:
return output
else:
return shape, aff
def ras_to_layout(x, affine):
layout = spatial.affine_to_layout(affine)
ras_to_layout = layout[..., 0]
return [x[i] for i in ras_to_layout]