def testGetHyperslab(self):
"""hyperslab should be same as slice from data array"""
inputFile = inputFile_ushort
#inputFile='/export01/data/vfonov/src1/minc2-simple/python/test_icbm.mnc'
v = minc2_file(inputFile)
v.setup_standard_order()
sliceFromData_x = v.data[10, :, :]
sliceFromData_y = v.data[:, 10, :]
sliceFromData_z = v.data[:, :, 10]
v.close()
b = minc2_file(inputFile)
b.setup_standard_order()
hyperslab_x = b.load_hyperslab_t([10, None, None]).squeeze()
hyperslab_y = b.load_hyperslab_t([None, 10, None]).squeeze()
hyperslab_z = b.load_hyperslab_t([None, None, 10]).squeeze()
b.close()
self.assertEqual(
torch.mean((sliceFromData_x - hyperslab_x)**2), 0.0)
self.assertEqual(
torch.mean((sliceFromData_y - hyperslab_y)**2), 0.0)
self.assertEqual(
torch.mean((sliceFromData_z - hyperslab_z)**2), 0.0)
def testSetHyperslabFloat(self):
"""setting hyperslab should change underlying volume (float)"""
# read some data from somwhere
v = minc2_file(inputFile_ushort)
dims = v.store_dims()
v.setup_standard_order()
# use the whole volume
hyperslab_a = v.data[10, :, :]
hyperslab_a_ = v[10, :, :]
v.close()
print("Hyperslab:", hyperslab_a.shape)
print("Hyperslab2:", hyperslab_a_.shape)
print("dims:", dims)
v2 = minc2_file()
v2.define(dims, 'float32', 'float32')
v2.create(outputFilename)
v2.setup_standard_order()
# because we are saving float32 , we don't need slice normalization
v2.save_hyperslab(hyperslab_a, [10, None, None])
v2.close()
v3 = minc2_file(outputFilename)
hyperslab_b = v3.load_hyperslab([10, None, None])
print(N.average((hyperslab_a - hyperslab_b)**2))
self.assertEqual(N.average((hyperslab_a - hyperslab_b)**2), 0.0)
v3.close()
def testSetSliceFloat(self):
"""volume slice setting should change underlying volume (float)"""
# read some data from somwhere
v = minc2_file(inputFile_ushort)
dims = v.store_dims()
v.setup_standard_order()
hyperslab_a = v.data[10, :, :]
v.close()
v2 = minc2_file()
v2.define(dims, 'float32', 'float32')
v2.create(outputFilename)
v2.setup_standard_order()
# because we are saving float32 , we don't need slice normalization
v2[10, :, :] = hyperslab_a
v2.close()
v3 = minc2_file(inputFile_ushort)
v3.setup_standard_order()
hyperslab_b = v3[10, :, :]
v3.close()
self.assertEqual(N.average((hyperslab_a - hyperslab_b)**2), 0.0)
def testSetHyperslabShort(self):
"""setting hyperslab should change underlying volume (short)"""
# read some data from somwhere
v = minc2_file(inputFile_ushort)
dims = v.store_dims()
v.setup_standard_order()
hyperslab_a = v.load_hyperslab_t([10, None, None])
# try with normalization
v2 = minc2_file()
v2.define(dims, 'uint16', 'float32') # , global_scaling=True
v2.create(outputFilename)
v2.set_volume_range(torch.min(hyperslab_a),
torch.max(hyperslab_a))
v2.setup_standard_order()
# have to set slice normalization
v2.save_hyperslab_t(hyperslab_a, [10, None, None])
hyperslab_b = v2.load_hyperslab_t([10, None, None])
# compare results
self.assertAlmostEqual(
torch.mean((hyperslab_a - hyperslab_b)**2).item(), 0.0, 8)
v2.close()
v.close()
def testHyperslabArray(self):
"""hyperslab should be reinsertable into volume"""
if False:
v = minc2_file(inputFile_ushort)
v2 = minc2_file()
v2.create(outputFilename)
v2.close()
v.close()
def save_labels(outfile, reference, data, history=None):
# TODO: add history
ref = minc2_file(reference)
out = minc2_file()
out.define(ref.store_dims(), minc2_file.MINC2_BYTE, minc2_file.MINC2_INT)
out.create(outfile)
out.copy_metadata(ref)
out.setup_standard_order()
out.save_complete_volume(data)
def testVectorRead(self):
"""make sure that a vector file can be read correctly"""
v = minc2_file(inputVector)
v.setup_standard_order()
dims = v.representation_dims()
self.assertEqual(dims[0].id, minc2_file.MINC2_DIM_VEC)
v.close()
def testNonDefaultDirCos3DVFF(self):
"""testing reading the direction cosines of a file with non-standard values (volumeFromFile)"""
v = minc2_file(input3DdirectionCosines)
v.setup_standard_order()
dims = v.representation_dims()
pipe = os.popen(
"mincinfo -attvalue xspace:direction_cosines %s" %
input3DdirectionCosines, "r")
from_file = pipe.read().rstrip().split(" ")
pipe.close()
self.assertAlmostEqual(dims[0].dir_cos[0], float(from_file[0]), 8)
self.assertAlmostEqual(dims[0].dir_cos[1], float(from_file[1]), 8)
self.assertAlmostEqual(dims[0].dir_cos[2], float(from_file[2]), 8)
pipe = os.popen(
"mincinfo -attvalue yspace:direction_cosines %s" %
input3DdirectionCosines, "r")
from_file = pipe.read().rstrip().split(" ")
pipe.close()
self.assertAlmostEqual(dims[1].dir_cos[0], float(from_file[0]), 8)
self.assertAlmostEqual(dims[1].dir_cos[1], float(from_file[1]), 8)
self.assertAlmostEqual(dims[1].dir_cos[2], float(from_file[2]), 8)
pipe = os.popen(
"mincinfo -attvalue zspace:direction_cosines %s" %
input3DdirectionCosines, "r")
from_file = pipe.read().rstrip().split(" ")
pipe.close()
self.assertAlmostEqual(dims[2].dir_cos[0], float(from_file[0]), 8)
self.assertAlmostEqual(dims[2].dir_cos[1], float(from_file[1]), 8)
self.assertAlmostEqual(dims[2].dir_cos[2], float(from_file[2]), 8)
def testVectorRead2(self):
"""make sure that volume has four dimensions"""
v = minc2_file(inputVector)
ndims = v.ndim()
self.assertEqual(ndims, 4)
data = v.data
self.assertEqual(len(data.shape), 4)
v.close()
def testWorldToVoxel(self):
"""testing world_to_voxel conversion of a file with non-standard values"""
v = minc2_file(input3DdirectionCosines)
v.setup_standard_order()
xyz = N.array([50.0, -80.0, -30.0])
ijk = v.world_to_voxel(xyz)
self.assertAlmostEqual(ijk[0], -6.362013409627703453, 8)
self.assertAlmostEqual(ijk[1], 6.6280285942264356436, 8)
self.assertAlmostEqual(ijk[2], 75.806692060998855709, 8)
def testFromFileDataDouble(self):
"""ensure that double data is read correct with a precision of 8 decimals on a call to aveage()"""
v = minc2_file(inputFile_double)
a = N.average(v.data)
v.close()
pipe = os.popen("mincstats -mean -quiet %s" % inputFile_double, "r")
output = float(pipe.read())
pipe.close()
self.assertAlmostEqual(a, output, 8)
def testFromFileDataFloat(self):
"""ensure that float data is read correct with a precision of 8 decimals on a call to aveage()"""
v = minc2_file(inputFile_float)
a = N.average(v.load_complete_volume('float64'))
v.close()
pipe = os.popen("mincstats -mean -quiet %s" % inputFile_float, "r")
output = float(pipe.read())
pipe.close()
self.assertAlmostEqual(a, output, 8)
def testVoxelToWorld(self):
"""testing voxel_to_world conversion of a file with non-standard values"""
v = minc2_file(input3DdirectionCosines)
v.setup_standard_order()
ijk = N.array([10, 20, 30])
xyz = v.voxel_to_world(ijk)
self.assertAlmostEqual(xyz[0], -0.32337910608109865507, 8)
self.assertAlmostEqual(xyz[1], -73.698635237250869068, 8)
self.assertAlmostEqual(xyz[2], -29.421450173791534155, 8)
def testFromFileDataInt(self):
"""ensure that int data is read correct with a precision of 8 decimals on a call to average()"""
v = minc2_file(inputFile_int)
a = v.load_complete_volume_tensor(
'torch.DoubleTensor').mean().item()
v.close()
pipe = os.popen("mincstats -mean -quiet %s" % inputFile_int, "r")
output = float(pipe.read())
pipe.close()
self.assertAlmostEqual(a, output, 8)
def testSetHyperslabFloat(self):
"""setting hyperslab should change underlying volume (float)"""
# read some data from somwhere
v = minc2_file(inputFile_ushort)
dims = v.store_dims()
v.setup_standard_order()
hyperslab_a = v.load_hyperslab_t([10, None, None])
v2 = minc2_file()
v2.define(dims, 'float32', 'float32')
v2.create(outputFilename)
v2.setup_standard_order()
# because we are saving float32 , we don't need slice normalization
v2.save_hyperslab_t(hyperslab_a, [10, None, None])
hyperslab_b = v2.load_hyperslab_t([10, None, None])
self.assertEqual(N.average((hyperslab_a - hyperslab_b)**2),
0.0)
v2.close()
v.close()
def merge_segmentations(inputs, output, partition, parameters):
patch_size = parameters.get('patch_size', 1)
border = patch_size * 2
out = None
strip = None
for i in range(len(inputs)):
d = minc2_file(inputs[i]).data
if out is None:
out = np.zeros(d.shape, dtype=np.int32)
strip = d.shape[2] / partition
beg = strip * i
end = strip * (i + 1)
if i == (partition - 1):
end = d.shape[2]
out[:, :, beg:end] = d[:, :, beg:end]
out_i = minc2_file()
out_i.imitate(inputs[0], path=output)
out_i.data = out
def testSlicingGet(self):
"""volume slice should be same as slice from data array"""
inputFile = inputFile_ushort
v = minc2_file(inputFile)
v.setup_standard_order()
sliceFromData_x = v.data[10, :, :]
sliceFromData_y = v.data[:, 10, :]
sliceFromData_z = v.data[:, :, 10]
v.close()
b = minc2_file(inputFile)
b.setup_standard_order()
hyperslab_x = b[10, :, :]
hyperslab_y = b[:, 10, :]
hyperslab_z = b[:, :, 10]
b.close()
self.assertEqual(N.average((sliceFromData_x - hyperslab_x)**2), 0.0)
self.assertEqual(N.average((sliceFromData_y - hyperslab_y)**2), 0.0)
self.assertEqual(N.average((sliceFromData_z - hyperslab_z)**2), 0.0)
def testDims(self):
"""Check data dimensions are correct"""
v = minc2_file(inputFile_double)
v.setup_standard_order()
dims = v.store_dims()
self.assertEqual(len(dims), 3)
# '100', '150', '125'
self.assertEqual(dims[0].id, minc2_file.MINC2_DIM_X) ## X
self.assertEqual(dims[0].length, 125)
self.assertEqual(dims[1].id, minc2_file.MINC2_DIM_Y) ## Y
self.assertEqual(dims[1].length, 150)
self.assertEqual(dims[2].id, minc2_file.MINC2_DIM_Z) ## X
self.assertEqual(dims[2].length, 100)
def load_minc_images(path, winsorize_low=5, winsorize_high=95):
from minc2_simple import minc2_file
import numpy as np
input_minc = minc2_file(path)
input_minc.setup_standard_order()
sz = input_minc.shape
input_images = [
input_minc[sz[0] // 2, :, :], input_minc[:, :, sz[2] // 2],
input_minc[:, sz[1] // 2, :]
]
# normalize between 5 and 95th percentile
_all_voxels = np.concatenate(tuple((np.ravel(i) for i in input_images)))
# _all_voxels=input_minc[:,:,:] # this is slower
_min = np.percentile(_all_voxels, winsorize_low)
_max = np.percentile(_all_voxels, winsorize_high)
input_images = [(i - _min) * (1.0 / (_max - _min)) - 0.5
for i in input_images]
# flip, resize and crop
for i in range(3):
#
_scale = min(256.0 / input_images[i].shape[0],
256.0 / input_images[i].shape[1])
# vertical flip and resize
input_images[i] = transform.rescale(input_images[i][::-1, :],
_scale,
mode='constant',
clip=False,
anti_aliasing=False,
multichannel=False)
sz = input_images[i].shape
# pad image
dummy = np.zeros((256, 256), )
dummy[int((256 - sz[0]) / 2):int((256 - sz[0]) / 2) + sz[0],
int((256 - sz[1]) / 2):int((256 - sz[1]) / 2) +
sz[1]] = input_images[i]
# crop
input_images[i] = dummy[16:240, 16:240]
return [torch.from_numpy(i).float().unsqueeze_(0) for i in input_images]
def testWorldToVoxelVec(self):
"""Compare against binary world to voxel"""
v = minc2_file(input3DdirectionCosines)
v.setup_standard_order()
x, y, z = N.meshgrid(N.linspace(-10, 10, 3), N.linspace(0, 20, 3),
N.linspace(-5, 15, 3))
xyz = N.column_stack((N.ravel(x), N.ravel(y), N.ravel(z)))
ijk = v.world_to_voxel(xyz)
for i, x in enumerate(xyz):
pipe = os.popen(
"worldtovoxel {} {} {} {}".format(input3DdirectionCosines,
x[0], x[1], x[2]), "r")
from_file = [float(i) for i in pipe.read().rstrip().split(" ")]
pipe.close()
for k in range(3):
self.assertAlmostEqual(from_file[k], ijk[i, k], 8)
def testDefaultDirCos3DVFF(self):
"""testing reading the direction cosines of a file with standard values (volumeFromFile)"""
v = minc2_file(inputFile_ushort)
#
# This file was created without explicitly setting the direction cosines.
# in that case, the attribute is not set altogether, so we should test
# for it using the known defaults, because libminc does extract the correct
# default values
#
v.setup_standard_order()
dims = v.representation_dims()
self.assertAlmostEqual(dims[0].dir_cos[0], 1.0, 8)
self.assertAlmostEqual(dims[0].dir_cos[1], 0.0, 8)
self.assertAlmostEqual(dims[0].dir_cos[2], 0.0, 8)
self.assertAlmostEqual(dims[1].dir_cos[0], 0.0, 8)
self.assertAlmostEqual(dims[1].dir_cos[1], 1.0, 8)
self.assertAlmostEqual(dims[1].dir_cos[2], 0.0, 8)
self.assertAlmostEqual(dims[2].dir_cos[0], 0.0, 8)
self.assertAlmostEqual(dims[2].dir_cos[1], 0.0, 8)
self.assertAlmostEqual(dims[2].dir_cos[2], 1.0, 8)
def get_minc_metadata(self, files):
for f in files:
# Get all minc headers from the minc2 file.
m = minc2_file(f)
meta = m.metadata()
# We need to convert our minc2-simple dictionary
# to a version of the dictionary that works with
# datalad. That means that the keys need to be strings,
# not bytes, and that we strip out anything that isn't
# hashable or decodeable, to ensure that we restrict
# ourselves to headers that can be serialized to json.
# Note that this only handles headers that are exactly
# 2 deep.
strmeta = {}
for key in meta:
# Convert the key from bytes to string so that datalad doesn't
# die on key.startswith.
kd = key.decode()
strmeta[kd] = {}
for subkey in meta[key]:
# Do the same for the subkeys.
skd = subkey.decode()
if meta[key][subkey].__hash__:
try:
# convert the value to utf-8. If it can't be converted
# to utf-8, it can't be serialized to JSON, so isn't indexable
# by datalad
v = meta[key][subkey]
encodedv = meta[key][subkey].decode('utf-8')
strmeta[kd][skd] = encodedv
except UnicodeDecodeError:
lgr.debug("Skipped %s.%s in %s" % (
key,
subkey,
f,
))
yield f, strmeta
def load_image(infile):
#with minc2_file(infile) as m:
m = minc2_file(infile)
m.setup_standard_order()
data = m.load_complete_volume(minc2_file.MINC2_FLOAT)
return data
from minc2_simple import minc2_xfm
import sys
import numpy as np
if __name__ == "__main__":
if len(sys.argv)<3:
print("Usage: {} input.mnc output.mnc".format(sys.argv[0]))
sys.exit(1)
infile=sys.argv[1]
outfile=sys.argv[2]
m=minc2_file(infile)
o=minc2_file()
# will create file with same dimensions
o.define(m.store_dims(), minc2_file.MINC2_BYTE, minc2_file.MINC2_FLOAT)
print("Will create new volume...")
o.create(outfile)
meta=m.metadata()
print("Metadata:")
print(repr(meta))
print("History:")
print(m.read_attribute("","history"))
def qc(
input,
output,
image_range=None,
mask=None,
mask_range=None,
title=None,
image_cmap='gray',
mask_cmap='red',
samples=5,
mask_bg=None,
use_max=False,
use_over=False,
show_image_bar=False, # TODO:implement this?
show_overlay_bar=False,
dpi=100,
ialpha=0.8,
oalpha=0.2,
format=None):
"""QC image generation, drop-in replacement for minc_qc.pl
Arguments:
input -- input minc file
output -- output QC graphics file
Keyword arguments:
image_range -- (optional) intensity range for image
mask -- (optional) input mask file
mask_range -- (optional) mask file range
title -- (optional) QC title
image_cmap -- (optional) color map name for image,
possibilities: red, green,blue and anything from matplotlib
mask_cmap -- (optional) color map for mask, default red
samples -- number of slices to show , default 5
mask_bg -- (optional) level for mask to treat as background
use_max -- (optional) use 'max' colour mixing
use_over -- (optional) use 'over' colour mixing
show_image_bar -- show color bar for intensity range, default false
show_overlay_bar -- show color bar for mask intensity range, default false
dpi -- graphics file DPI, default 100
ialpha -- alpha channel for colour mixing of main image
oalpha -- alpha channel for colour mixing of mask image
"""
#_img=minc.Image(input)
#_idata=_img.data
_img = minc2_file(input)
_img.setup_standard_order()
_idata = _img.load_complete_volume(minc2_file.MINC2_FLOAT)
_idims = _img.representation_dims()
data_shape = _idata.shape
spacing = [_idims[0].step, _idims[1].step, _idims[2].step]
_ovl = None
_odata = None
omin = 0
omax = 1
if mask is not None:
_ovl = minc2_file(mask)
_ovl.setup_standard_order()
_ovl_data = _ovl.load_complete_volume(minc2_file.MINC2_FLOAT)
if _ovl_data.shape != data_shape:
raise "Overlay shape does not match image!\nOvl={} Image={}".format(
repr(_ovl_data.shape), repr(data_shape))
if mask_range is None:
omin = np.nanmin(_ovl_data)
omax = np.nanmax(_ovl_data)
else:
omin = mask_range[0]
omax = mask_range[1]
_odata = _ovl_data
if mask_bg is not None:
_odata = ma.masked_less(_odata, mask_bg)
slices = []
# setup ranges
vmin = vmax = 0.0
if image_range is not None:
vmin = image_range[0]
vmax = image_range[1]
else:
vmin = np.nanmin(_idata)
vmax = np.nanmax(_idata)
cm = copy.copy(plt.get_cmap(image_cmap))
cmo = copy.copy(plt.get_cmap(mask_cmap))
cmo.set_bad('k', alpha=0.0)
cNorm = colors.Normalize(vmin=vmin, vmax=vmax)
oNorm = colors.Normalize(vmin=omin, vmax=omax)
scalarMap = cmx.ScalarMappable(norm=cNorm, cmap=cm)
oscalarMap = cmx.ScalarMappable(norm=oNorm, cmap=cmo)
aspects = []
# axial slices
for j in range(0, samples):
i = int((data_shape[0] / samples) * j + (data_shape[0] % samples) / 2)
si = scalarMap.to_rgba(_idata[i, :, :])
if _ovl is not None:
so = oscalarMap.to_rgba(_odata[i, :, :])
if use_max: si = max_blend(si, so)
elif use_over: si = over_blend(si, so, ialpha, oalpha)
else: si = alpha_blend(si, so, ialpha, oalpha)
slices.append(si)
aspects.append(spacing[0] / spacing[1])
# coronal slices
for j in range(0, samples):
i = int((data_shape[1] / samples) * j + (data_shape[1] % samples) / 2)
si = scalarMap.to_rgba(_idata[:, i, :])
if _ovl is not None:
so = oscalarMap.to_rgba(_odata[:, i, :])
if use_max: si = max_blend(si, so)
elif use_over: si = over_blend(si, so, ialpha, oalpha)
else: si = alpha_blend(si, so, ialpha, oalpha)
slices.append(si)
aspects.append(spacing[2] / spacing[0])
# sagittal slices
for j in range(0, samples):
i = int((data_shape[2] / samples) * j + (data_shape[2] % samples) / 2)
si = scalarMap.to_rgba(_idata[:, :, i])
if _ovl is not None:
so = oscalarMap.to_rgba(_odata[:, :, i])
if use_max: si = max_blend(si, so)
elif use_over: si = over_blend(si, so, ialpha, oalpha)
else: si = alpha_blend(si, so, ialpha, oalpha)
slices.append(si)
aspects.append(spacing[2] / spacing[1])
w, h = plt.figaspect(3.0 / samples)
fig = plt.figure(figsize=(w, h))
#outer_grid = gridspec.GridSpec((len(slices)+1)/2, 2, wspace=0.0, hspace=0.0)
ax = None
imgplot = None
for i, j in enumerate(slices):
ax = plt.subplot2grid((3, samples), (int(i / samples), i % samples))
imgplot = ax.imshow(j, origin='lower', cmap=cm, aspect=aspects[i])
ax.set_xticks([])
ax.set_yticks([])
ax.title.set_visible(False)
# show for the last plot
if show_image_bar:
cbar = fig.colorbar(imgplot)
if title is not None:
plt.suptitle(title, fontsize=20)
plt.subplots_adjust(wspace=0.0, hspace=0.0)
else:
plt.subplots_adjust(top=1.0,
bottom=0.0,
left=0.0,
right=1.0,
wspace=0.0,
hspace=0.0)
#fig.tight_layout()
#plt.show()
plt.savefig(output, bbox_inches='tight', dpi=dpi, format=format)
plt.close()
plt.close('all')
def qc_field_contour(
input,
output,
image_range=None,
title=None,
image_cmap='gray',
samples=5,
show_image_bar=False, # TODO:implement this?
dpi=100,
format=None):
"""show field contours
"""
_img = minc2_file(input)
_img.setup_standard_order()
_idata = _img.load_complete_volume(minc2_file.MINC2_FLOAT)
_idims = _img.representation_dims()
data_shape = _idata.shape
spacing = [_idims[0].step, _idims[1].step, _idims[2].step]
slices = []
# setup ranges
vmin = vmax = 0.0
if image_range is not None:
vmin = image_range[0]
vmax = image_range[1]
else:
vmin = np.nanmin(_idata)
vmax = np.nanmax(_idata)
cm = plt.get_cmap(image_cmap)
cNorm = colors.Normalize(vmin=vmin, vmax=vmax)
scalarMap = cmx.ScalarMappable(norm=cNorm, cmap=cm)
for j in range(0, samples):
i = (data_shape[0] / samples) * j + (data_shape[0] % samples) / 2
si = _idata[i, :, :]
slices.append(si)
for j in range(0, samples):
i = (data_shape[1] / samples) * j + (data_shape[1] % samples) / 2
si = _idata[:, i, :]
slices.append(si)
for j in range(0, samples):
i = (data_shape[2] / samples) * j + (data_shape[2] % samples) / 2
si = _idata[:, :, i]
slices.append(si)
w, h = plt.figaspect(3.0 / samples)
fig = plt.figure(figsize=(w, h))
#outer_grid = gridspec.GridSpec((len(slices)+1)/2, 2, wspace=0.0, hspace=0.0)
ax = None
imgplot = None
for i, j in enumerate(slices):
ax = plt.subplot2grid((3, samples), (i / samples, i % samples))
imgplot = ax.contour(j,
origin='lower',
cmap=cm,
norm=cNorm,
levels=np.linspace(vmin, vmax, 20))
#plt.clabel(imgplot, inline=1, fontsize=8)
ax.set_xticks([])
ax.set_yticks([])
ax.title.set_visible(False)
# show for the last plot
if show_image_bar:
cbar = fig.colorbar(imgplot)
if title is not None:
plt.suptitle(title, fontsize=20)
plt.subplots_adjust(wspace=0.0, hspace=0.0)
else:
plt.subplots_adjust(top=1.0,
bottom=0.0,
left=0.0,
right=1.0,
wspace=0.0,
hspace=0.0)
plt.savefig(output, bbox_inches='tight', dpi=dpi)
plt.close('all')
def errorCorrectionApply(input_images,
output,
input_mask=None,
parameters=None,
debug=False,
history=None,
input_auto=None,
partition=None,
part=None,
multilabel=1,
debug_files=None):
try:
use_coord = parameters.get('use_coord', True)
use_joint = parameters.get('use_joint', True)
patch_size = parameters.get('patch_size', 1)
normalize_input = parameters.get('normalize_input', True)
primary_features = parameters.get('primary_features', 1)
method = parameters.get('method', 'lSVC')
method2 = parameters.get('method2', method)
training = parameters['training']
clf = None
clf2 = None
border = patch_size * 2
if patch_size == 0:
border = 2
if debug:
print(
"Running error-correction, input_image:{} trining:{} partition:{} part:{} output:{} input_auto:{}"
.format(repr(input_images), training, partition, part, output,
input_auto))
if method == 'xgb' and method2 == 'xgb':
# need to convert from Unicode
_training = str(training)
clf = xgb.Booster(model_file=_training)
if multilabel > 1:
clf2 = xgb.Booster(model_file=_training + '_2')
else:
with open(training, 'rb') as f:
c = pickle.load(f)
clf = c[0]
clf2 = c[1]
if debug:
print(clf)
print(clf2)
print("Loading input images...")
input_data = [
minc2_file(k).load_complete_volume('float32') for k in input_images
]
shape = input_data[0].shape
#features = [ extract_part( minc.Image(k, dtype=np.float32).data, partition, part, border) for k in inp[0:-3] ]
#if normalize_input:
#features = [ extract_part( preprocessing.scale( k ), partition, part, border) for k in input_data ]
#else:
features = [
extract_part(k, partition, part, border) for k in input_data
]
coords = None
if use_coord:
c = np.mgrid[0:shape[0], 0:shape[1], 0:shape[2]]
coords = [
extract_part((c[j] - shape[j] / 2.0) / (shape[j] / 2.0),
partition, part, border) for j in range(3)
]
if debug:
print("Features data size:{}".format(len(features)))
mask = None
mask_size = shape[0] * shape[1] * shape[2]
if input_mask is not None:
mask = extract_part(
minc2_file(input_mask).data, partition, part, border)
mask_size = np.sum(mask)
out_cls = None
out_corr = None
test_x = convert_image_list([
prepare_features(features,
coords,
mask=mask,
use_coord=use_coord,
use_joint=use_joint,
patch_size=patch_size,
primary_features=primary_features)
])
if input_auto is not None:
out_corr = np.copy(
extract_part(
minc2_file(input_auto).data, partition, part,
border)) # use input data
out_cls = np.copy(
extract_part(
minc2_file(input_auto).data, partition, part,
border)) # use input data
else:
out_corr = np.zeros(shape, dtype=np.int32)
out_cls = np.zeros(shape, dtype=np.int32)
if mask_size > 0 and not isinstance(clf, dummy.DummyClassifier):
if debug:
print("Running classifier 1 ...")
if method != 'xgb':
pred = np.asarray(clf.predict(test_x), dtype=np.int32)
else:
xg_predict = xgb.DMatrix(test_x)
pred = np.array(clf.predict(xg_predict), dtype=np.int32)
if debug_files is not None:
out_dbg = np.zeros(shape, dtype=np.int32)
if mask is not None:
out_dbg[mask > 0] = pred
else:
out_dbg = pred
out_dbg_m = minc2_file()
out_dbg_m.imitate(input_images[0], path=debug_files[0])
out_dbg_m.data = pad_data(out_dbg, shape, partition, part,
border)
if mask is not None:
out_corr[mask > 0] = pred
else:
out_corr = pred
if multilabel > 1 and clf2 is not None:
if mask is not None:
mask = np.logical_and(mask > 0, out_corr > 0)
else:
mask = (out_corr > 0)
if debug:
print("Running classifier 2 ...")
test_x = convert_image_list([
prepare_features(features,
coords,
mask=mask,
use_coord=use_coord,
use_joint=use_joint,
patch_size=patch_size,
primary_features=primary_features)
])
if method2 != 'xgb':
pred = np.asarray(clf2.predict(test_x), dtype=np.int32)
else:
xg_predict = xgb.DMatrix(test_x)
pred = np.array(clf2.predict(xg_predict), dtype=np.int32)
out_cls[mask > 0] = pred
if debug_files is not None:
out_dbg = np.zeros(shape, dtype=np.int32)
if mask is not None:
out_dbg[mask > 0] = pred
else:
out_dbg = pred
out_dbg_m = minc2_file()
out_dbg_m.imitate(input_images[0], path=debug_files[1])
out_dbg_m.data = pad_data(out_dbg, shape, partition, part,
border)
else:
out_cls = out_corr
else:
pass # nothing to do!
if debug:
print("Saving output...")
out = minc2_file()
out.imitate(input_images[0], path=output)
out.data = pad_data(out_cls, shape, partition, part, border)
except mincError as e:
print("Exception in errorCorrectionApply:{}".format(str(e)))
traceback.print_exc(file=sys.stdout)
raise
except:
print("Exception in errorCorrectionApply:{}".format(sys.exc_info()[0]))
traceback.print_exc(file=sys.stdout)
raise
def errorCorrectionTrain(input_images,
output,
parameters=None,
debug=False,
partition=None,
part=None,
multilabel=1):
try:
use_coord = parameters.get('use_coord', True)
use_joint = parameters.get('use_joint', True)
patch_size = parameters.get('patch_size', 1)
border = patch_size * 2
if patch_size == 0:
border = 2
normalize_input = parameters.get('normalize_input', True)
method = parameters.get('method', 'lSVC')
method2 = parameters.get('method2', method)
method_n = parameters.get('method_n', 15)
method2_n = parameters.get('method2_n', method_n)
method_random = parameters.get('method_random', None)
method_max_features = parameters.get('method_max_features', 'auto')
method_n_jobs = parameters.get('method_n_jobs', 1)
primary_features = parameters.get('primary_features', 1)
training_images = []
training_diff = []
training_images_direct = []
training_direct = []
if debug:
print("errorCorrectionTrain use_coord={} use_joint={} patch_size={} normalize_input={} method={} output={} partition={} part={}".\
format(repr(use_coord),repr(use_joint),repr(patch_size),repr(normalize_input),method,output,partition,part))
coords = None
total_mask_size = 0
total_diff_mask_size = 0
for (i, inp) in enumerate(input_images):
mask = None
diff = None
mask_diff = None
if inp[-2] is not None:
mask = extract_part(
minc2_file(inp[-2]).data, partition, part, border)
ground_data = minc2_file(inp[-1]).data
auto_data = minc2_file(inp[-3]).data
ground_shape = ground_data.shape
ground = extract_part(ground_data, partition, part, border)
auto = extract_part(auto_data, partition, part, border)
shape = ground_shape
if coords is None and use_coord:
c = np.mgrid[0:shape[0], 0:shape[1], 0:shape[2]]
coords = [
extract_part((c[j] - shape[j] / 2.0) / (shape[j] / 2.0),
partition, part, border) for j in range(3)
]
features = [
extract_part(minc2_file(k).data, partition, part, border)
for k in inp[0:-3]
]
mask_size = shape[0] * shape[1] * shape[2]
if debug:
print("Training data size:{}".format(len(features)))
if mask is not None:
mask_size = np.sum(mask)
print("Mask size:{}".format(mask_size))
else:
print("Mask absent")
total_mask_size += mask_size
if multilabel > 1:
diff = (ground != auto)
total_diff_mask_size += np.sum(mask)
if mask is not None:
mask_diff = diff & (mask > 0)
print("Sample {} mask_diff={} diff={}".format(
i, np.sum(mask_diff), np.sum(diff)))
#print(mask_diff)
training_diff.append(diff[mask > 0])
training_direct.append(ground[mask_diff])
else:
mask_diff = diff
training_diff.append(diff)
training_direct.append(ground[diff])
training_images.append(
prepare_features(features,
coords,
mask=mask,
use_coord=use_coord,
use_joint=use_joint,
patch_size=patch_size,
primary_features=primary_features))
training_images_direct.append(
prepare_features(features,
coords,
mask=mask_diff,
use_coord=use_coord,
use_joint=use_joint,
patch_size=patch_size,
primary_features=primary_features))
else:
mask_diff = mask
if mask is not None:
training_diff.append(ground[mask > 0])
else:
training_diff.append(ground)
training_images.append(
prepare_features(features,
coords,
mask=mask,
use_coord=use_coord,
use_joint=use_joint,
patch_size=patch_size,
primary_features=primary_features))
if debug:
print("feature size:{}".format(len(training_images[-1])))
if i == 0 and parameters.get('dump', False):
print("Dumping feature images...")
for (j, k) in enumerate(training_images[-1]):
test = np.zeros_like(images[0])
test[mask > 0] = k
out = minc2_file()
out.imitate(inp[0], path="dump_{}.mnc".format(j))
out.data = test
# calculate normalization coeffecients
if debug: print("Done")
clf = None
clf2 = None
if total_mask_size > 0:
training_X = convert_image_list(training_images)
training_Y = np.ravel(
np.concatenate(tuple(j for j in training_diff)))
if debug: print("Fitting 1st...")
if method == "xgb":
clf = None
elif method == "SVM":
clf = svm.SVC()
elif method == "nuSVM":
clf = svm.NuSVC()
elif method == 'NC':
clf = neighbors.NearestCentroid()
elif method == 'NN':
clf = neighbors.KNeighborsClassifier(method_n)
elif method == 'RanForest':
clf = ensemble.RandomForestClassifier(
n_estimators=method_n,
n_jobs=method_n_jobs,
max_features=method_max_features,
random_state=method_random)
elif method == 'AdaBoost':
clf = ensemble.AdaBoostClassifier(n_estimators=method_n,
random_state=method_random)
elif method == 'AdaBoostPP':
clf = Pipeline(steps=[('normalizer', Normalizer()),
('AdaBoost',
ensemble.AdaBoostClassifier(
n_estimators=method_n,
random_state=method_random))])
elif method == 'tree':
clf = tree.DecisionTreeClassifier(random_state=method_random)
elif method == 'ExtraTrees':
clf = ensemble.ExtraTreesClassifier(
n_estimators=method_n,
max_features=method_max_features,
n_jobs=method_n_jobs,
random_state=method_random)
elif method == 'Bagging':
clf = ensemble.BaggingClassifier(
n_estimators=method_n,
max_features=method_max_features,
n_jobs=method_n_jobs,
random_state=method_random)
elif method == 'dumb':
clf = dummy.DummyClassifier(strategy="constant", constant=0)
else:
clf = svm.LinearSVC()
#scores = cross_validation.cross_val_score(clf, training_X, training_Y)
#print scores
if method == "xgb":
xg_train = xgb.DMatrix(training_X, label=training_Y)
param = {}
num_round = 100
# use softmax multi-class classification
param['objective'] = 'multi:softmax'
# scale weight of positive examples
param['eta'] = 0.1
param['max_depth'] = 8
param['silent'] = 1
param['nthread'] = 4
param['num_class'] = 2
clf = xgb.train(param, xg_train, num_round)
elif method != 'dumb':
clf.fit(training_X, training_Y)
if multilabel > 1 and method != 'dumb':
if debug: print("Fitting direct...")
training_X = convert_image_list(training_images_direct)
training_Y = np.ravel(
np.concatenate(tuple(j for j in training_direct)))
if method2 == "xgb":
clf2 = None
if method2 == "SVM":
clf2 = svm.SVC()
elif method2 == "nuSVM":
clf2 = svm.NuSVC()
elif method2 == 'NC':
clf2 = neighbors.NearestCentroid()
elif method2 == 'NN':
clf2 = neighbors.KNeighborsClassifier(method_n)
elif method2 == 'RanForest':
clf2 = ensemble.RandomForestClassifier(
n_estimators=method_n,
n_jobs=method_n_jobs,
max_features=method_max_features,
random_state=method_random)
elif method2 == 'AdaBoost':
clf2 = ensemble.AdaBoostClassifier(
n_estimators=method_n, random_state=method_random)
elif method2 == 'AdaBoostPP':
clf2 = Pipeline(steps=[('normalizer', Normalizer()),
('AdaBoost',
ensemble.AdaBoostClassifier(
n_estimators=method_n,
random_state=method_random))])
elif method2 == 'tree':
clf2 = tree.DecisionTreeClassifier(
random_state=method_random)
elif method2 == 'ExtraTrees':
clf2 = ensemble.ExtraTreesClassifier(
n_estimators=method_n,
max_features=method_max_features,
n_jobs=method_n_jobs,
random_state=method_random)
elif method2 == 'Bagging':
clf2 = ensemble.BaggingClassifier(
n_estimators=method_n,
max_features=method_max_features,
n_jobs=method_n_jobs,
random_state=method_random)
elif method2 == 'dumb':
clf2 = dummy.DummyClassifier(strategy="constant",
constant=0)
else:
clf2 = svm.LinearSVC()
if method2 == "xgb":
xg_train = xgb.DMatrix(training_X, label=training_Y)
param = {}
num_round = 100
# use softmax multi-class classification
param['objective'] = 'multi:softmax'
# scale weight of positive examples
param['eta'] = 0.1
param['max_depth'] = 8
param['silent'] = 1
param['nthread'] = 4
param['num_class'] = multilabel
clf2 = xgb.train(param, xg_train, num_round)
elif method != 'dumb':
clf2.fit(training_X, training_Y)
#print(clf.score(training_X,training_Y))
if debug:
print(clf)
print(clf2)
else:
print("Warning : zero total mask size!, using null classifier")
clf = dummy.DummyClassifier(strategy="constant", constant=0)
if method == 'xgb' and method2 == 'xgb':
#save
clf.save_model(output)
clf2.save_model(output + '_2')
else:
with open(output, 'wb') as f:
pickle.dump([clf, clf2], f, -1)
except mincError as e:
print("Exception in linear_registration:{}".format(str(e)))
traceback.print_exc(file=sys.stdout)
raise
except:
print("Exception in linear_registration:{}".format(sys.exc_info()[0]))
traceback.print_exc(file=sys.stdout)
raise
c = np.mgrid[0:img.shape[0], 0:img.shape[1], 0:img.shape[2]]
seg = np.zeros_like(img, dtype=np.int32)
seg=( c[2]>center[0] )*1+\
( c[1]>center[1] )*2+\
( c[0]>center[2] )*4+ 1
seg[img < 50] = 0
return np.asarray(seg, dtype=np.int32)
if __name__ == "__main__":
options = parse_options()
print(repr(options))
input = minc2_file(options.input)
input.setup_standard_order()
data = input.load_complete_volume(minc2_file.MINC2_FLOAT)
#seg=np.zeros_like( input.data, dtype=np.int32 )
center_vox = [(options.center[i] - input.representation_dims()[i].start) /
input.representation_dims()[i].step for i in xrange(3)]
print(repr(center_vox))
seg = dumb_segment(data, center_vox)
save_labels(options.output, input, seg)
# kate: space-indent on; indent-width 4; indent-mode python;replace-tabs on;word-wrap-column 80;show-tabs on
def qc(
input,
output,
image_range=None,
mask=None,
mask_range=None,
title=None,
image_cmap='gray',
mask_cmap='red',
samples=6,
mask_bg=None,
use_max=False,
use_over=False,
show_image_bar=False, # TODO:implement this?
show_overlay_bar=False,
dpi=100,
ialpha=0.8,
oalpha=0.2,
format=None,
bg_color=None,
fg_color=None
):
"""QC image generation, drop-in replacement for minc_qc.pl
Arguments:
input -- input minc file
output -- output QC graphics file
Keyword arguments:
image_range -- (optional) intensity range for image
mask -- (optional) input mask file
mask_range -- (optional) mask file range
title -- (optional) QC title
image_cmap -- (optional) color map name for image,
possibilities: red, green,blue and anything from matplotlib
mask_cmap -- (optional) color map for mask, default red
samples -- number of slices per dimension to show, default 6
should be even number, becaus it will be split across two rows
mask_bg -- (optional) level for mask to treat as background
use_max -- (optional) use 'max' colour mixing
use_over -- (optional) use 'over' colour mixing
show_image_bar -- show color bar for intensity range, default false
show_overlay_bar -- show color bar for mask intensity range, default false
dpi -- graphics file DPI, default 100
ialpha -- alpha channel for colour mixing of main image
oalpha -- alpha channel for colour mixing of mask image
"""
_img=minc2_file(input)
_img.setup_standard_order()
_idata=_img.load_complete_volume(minc2_file.MINC2_FLOAT)
_idims=_img.representation_dims()
data_shape=_idata.shape
# order of dimensions in representation dimension is reversed
spacing=[_idims[2].step, _idims[1].step, _idims[0].step]
_ovl=None
_odata=None
omin=0
omax=1
# setup view
columns=samples//2
rows=(samples//columns)*3
if mask is not None:
_ovl=minc2_file(mask)
_ovl.setup_standard_order()
_ovl_data=_ovl.load_complete_volume(minc2_file.MINC2_FLOAT)
if _ovl_data.shape != data_shape:
raise "Overlay shape does not match image!\nOvl={} Image={}".format(repr(_ovl_data.shape),repr(data_shape))
if mask_range is None:
omin=np.nanmin(_ovl_data)
omax=np.nanmax(_ovl_data)
else:
omin=mask_range[0]
omax=mask_range[1]
_odata=_ovl_data
if mask_bg is not None:
_odata=ma.masked_less(_odata, mask_bg)
slices=[]
# setup ranges
vmin=vmax=0.0
if image_range is not None:
vmin=image_range[0]
vmax=image_range[1]
else:
vmin=np.nanmin(_idata)
vmax=np.nanmax(_idata)
cm = copy.copy(plt.get_cmap(image_cmap))
cmo= copy.copy(plt.get_cmap(mask_cmap))
cm.set_bad('k', alpha = 1.0)
cmo.set_bad('k',alpha = 0.0)
cNorm = colors.Normalize(vmin=vmin, vmax=vmax)
oNorm = colors.Normalize(vmin=omin, vmax=omax)
scalarMap = cmx.ScalarMappable(norm=cNorm, cmap=cm)
oscalarMap = cmx.ScalarMappable(norm=oNorm, cmap=cmo)
aspects = []
# axial slices
for j in range(0, samples):
i=int(10+(150.0-10.0)*j/(samples-1))
i=int(data_shape[0]*i/181.0)
si=scalarMap.to_rgba(_idata[i , : ,:])
if _ovl is not None:
so = oscalarMap.to_rgba(_odata[i , : ,:])
if use_max: si=max_blend(si, so)
elif use_over: si=over_blend(si, so, ialpha, oalpha)
else: si=alpha_blend(si, so, ialpha, oalpha)
slices.append( si )
aspects.append( spacing[1] / spacing[2] )
# sagittal slices
for j in range(0, samples//2):
i=int(28.0+(166.0-28.0)*j/(samples-1))
i=int(data_shape[2]*i/193.0)
si=scalarMap.to_rgba(_idata[: , : , i])
if _ovl is not None:
so = oscalarMap.to_rgba(_odata[: , : , i])
if use_max: si=max_blend(si,so)
elif use_over: si=over_blend(si,so, ialpha, oalpha)
else: si=alpha_blend(si, so, ialpha, oalpha)
slices.append( si )
aspects.append( spacing[0] / spacing[1] )
for j in range(samples-1, samples//2-1,-1):
i=int(28.0+(166.0-28.0)*j/(samples-1))
i=int(data_shape[2]*i/193.0)
si=scalarMap.to_rgba(_idata[: , : , i])
if _ovl is not None:
so = oscalarMap.to_rgba(_odata[: , : , i])
if use_max: si=max_blend(si,so)
elif use_over: si=over_blend(si,so, ialpha, oalpha)
else: si=alpha_blend(si, so, ialpha, oalpha)
slices.append( si )
aspects.append( spacing[0] / spacing[1] )
# coronal slices
for j in range(0,samples):
i=int(25+(195.0-25.0)*j/(samples-1))
i=int(data_shape[1]*i/217.0)
si=scalarMap.to_rgba(_idata[: , i ,:])
if _ovl is not None:
so = oscalarMap.to_rgba(_odata[: , i ,:])
if use_max: si=max_blend(si,so)
elif use_over: si=over_blend(si,so, ialpha, oalpha)
else: si=alpha_blend(si, so, ialpha, oalpha)
slices.append( si )
aspects.append( spacing[0]/spacing[2] )
rc={'interactive': False}
rc['figure.frameon']=False
#rc['aa']=True
if bg_color is not None:
rc['figure.edgecolor']=bg_color
rc['figure.facecolor']=bg_color
rc['grid.color']=bg_color
if fg_color is not None:
rc['text.color']=fg_color
#axes.labelcolor
#axes.titlecolor
with matplotlib.rc_context(rc):
#with plt.style.context('dark_background'):
w, h = plt.figaspect(rows/columns)
fig = plt.figure(figsize=(w,h))
# if bg_color is not None:
# fig.set_facecolor(bg_color)
# fig.set_edgecolor(bg_color)
# fig.set_frameon(False)
#outer_grid = gridspec.GridSpec((len(slices)+1)/2, 2, wspace=0.0, hspace=0.0)
ax=None
imgplot=None
print(f"rows:{rows} columns:{columns}")
for i,j in enumerate(slices):
ax = plt.subplot2grid( (rows, columns), ( i//columns, i%columns) )
# if bg_color is not None:
# ax.set_facecolor(bg_color)
imgplot = ax.imshow(j, origin='lower', cmap=cm, aspect=aspects[i])
ax.set_xticks([])
ax.set_yticks([])
ax.title.set_visible(False)
# show for the last plot
if show_image_bar:
cbar = fig.colorbar(imgplot)
if title is not None:
plt.suptitle(title,fontsize=20)
plt.subplots_adjust(wspace = 0.0 ,hspace=0.0)
else:
plt.subplots_adjust(top=1.0,bottom=0.0,left=0.0,right=1.0,wspace = 0.0 ,hspace=0.0)
# , facecolor=fig.get_facecolor(), edgecolor='none')
plt.savefig(output, bbox_inches='tight', dpi=dpi, format=format)
plt.close()
plt.close('all')