def _is_bval_type_a(grouped_dicoms):
"""
Check if the bvals are stored in the first of 2 currently known ways for single frame dti
"""
bval_tag = Tag(0x2001, 0x1003)
bvec_x_tag = Tag(0x2005, 0x10b0)
bvec_y_tag = Tag(0x2005, 0x10b1)
bvec_z_tag = Tag(0x2005, 0x10b2)
for group in grouped_dicoms:
if bvec_x_tag in group[0] and _is_float(common.get_fl_value(group[0][bvec_x_tag])) and \
bvec_y_tag in group[0] and _is_float(common.get_fl_value(group[0][bvec_y_tag])) and \
bvec_z_tag in group[0] and _is_float(common.get_fl_value(group[0][bvec_z_tag])) and \
bval_tag in group[0] and _is_float(common.get_fl_value(group[0][bval_tag])) and \
common.get_fl_value(group[0][bval_tag]) != 0:
return True
return False
def _is_bval_type_a(grouped_dicoms):
"""
Check if the bvals are stored in the first of 2 currently known ways for single frame dti
"""
bval_tag = Tag(0x2001, 0x1003)
bvec_x_tag = Tag(0x2005, 0x10b0)
bvec_y_tag = Tag(0x2005, 0x10b1)
bvec_z_tag = Tag(0x2005, 0x10b2)
for group in grouped_dicoms:
if bvec_x_tag in group[0] and _is_float(common.get_fl_value(group[0][bvec_x_tag])) and \
bvec_y_tag in group[0] and _is_float(common.get_fl_value(group[0][bvec_y_tag])) and \
bvec_z_tag in group[0] and _is_float(common.get_fl_value(group[0][bvec_z_tag])) and \
bval_tag in group[0] and _is_float(common.get_fl_value(group[0][bval_tag])) and \
common.get_fl_value(group[0][bval_tag]) != 0:
return True
return False
def _create_singleframe_bvals_bvecs(grouped_dicoms, bval_file, bvec_file,
nifti, nifti_file):
"""
Write the bvals from the sorted dicom files to a bval file
"""
# create the empty arrays
bvals = numpy.zeros([len(grouped_dicoms)], dtype=numpy.int32)
bvecs = numpy.zeros([len(grouped_dicoms), 3])
# loop over all timepoints and create a list with all bvals and bvecs
if _is_bval_type_a(grouped_dicoms):
bval_tag = Tag(0x2001, 0x1003)
bvec_x_tag = Tag(0x2005, 0x10b0)
bvec_y_tag = Tag(0x2005, 0x10b1)
bvec_z_tag = Tag(0x2005, 0x10b2)
for stack_index in range(0, len(grouped_dicoms)):
bvals[stack_index] = common.get_fl_value(
grouped_dicoms[stack_index][0][bval_tag])
bvecs[stack_index, :] = [
common.get_fl_value(
grouped_dicoms[stack_index][0][bvec_x_tag]),
common.get_fl_value(
grouped_dicoms[stack_index][0][bvec_y_tag]),
common.get_fl_value(grouped_dicoms[stack_index][0][bvec_z_tag])
]
elif _is_bval_type_b(grouped_dicoms):
bval_tag = Tag(0x0018, 0x9087)
bvec_tag = Tag(0x0018, 0x9089)
for stack_index in range(0, len(grouped_dicoms)):
bvals[stack_index] = common.get_fd_value(
grouped_dicoms[stack_index][0][bval_tag])
bvecs[stack_index, :] = common.get_fd_array_value(
grouped_dicoms[stack_index][0][bvec_tag], 3)
# truncate nifti if needed
nifti, bvals, bvecs = _fix_diffusion_images(bvals, bvecs, nifti,
nifti_file)
# save the found bvecs to the file
if numpy.count_nonzero(bvals) > 0 or numpy.count_nonzero(bvecs) > 0:
common.write_bval_file(bvals, bval_file)
common.write_bvec_file(bvecs, bvec_file)
else:
bval_file = None
bvec_file = None
return bval_file, bvec_file
def _create_singleframe_bvals_bvecs(grouped_dicoms, bval_file, bvec_file, nifti, nifti_file):
"""
Write the bvals from the sorted dicom files to a bval file
"""
# create the empty arrays
bvals = numpy.zeros([len(grouped_dicoms)], dtype=numpy.int32)
bvecs = numpy.zeros([len(grouped_dicoms), 3])
# loop over all timepoints and create a list with all bvals and bvecs
if _is_bval_type_a(grouped_dicoms):
bval_tag = Tag(0x2001, 0x1003)
bvec_x_tag = Tag(0x2005, 0x10b0)
bvec_y_tag = Tag(0x2005, 0x10b1)
bvec_z_tag = Tag(0x2005, 0x10b2)
for stack_index in range(0, len(grouped_dicoms)):
bvals[stack_index] = common.get_fl_value(grouped_dicoms[stack_index][0][bval_tag])
bvecs[stack_index, :] = [common.get_fl_value(grouped_dicoms[stack_index][0][bvec_x_tag]),
common.get_fl_value(grouped_dicoms[stack_index][0][bvec_y_tag]),
common.get_fl_value(grouped_dicoms[stack_index][0][bvec_z_tag])]
elif _is_bval_type_b(grouped_dicoms):
bval_tag = Tag(0x0018, 0x9087)
bvec_tag = Tag(0x0018, 0x9089)
for stack_index in range(0, len(grouped_dicoms)):
bvals[stack_index] = common.get_fd_value(grouped_dicoms[stack_index][0][bval_tag])
bvecs[stack_index, :] = common.get_fd_array_value(grouped_dicoms[stack_index][0][bvec_tag], 3)
# truncate nifti if needed
nifti, bvals, bvecs = _fix_diffusion_images(bvals, bvecs, nifti, nifti_file)
# save the found bvecs to the file
if numpy.count_nonzero(bvals) > 0 or numpy.count_nonzero(bvecs) > 0:
common.write_bval_file(bvals, bval_file)
common.write_bvec_file(bvecs, bvec_file)
else:
bval_file = None
bvec_file = None
bvals = None
bvecs = None
return nifti, bvals, bvecs, bval_file, bvec_file
def _multiframe_to_block(multiframe_dicom):
"""
Generate a full datablock containing all stacks
"""
# Calculate the amount of stacks and slices in the stack
number_of_stack_slices = int(
common.get_ss_value(multiframe_dicom[Tag(0x2001, 0x105f)][0][Tag(
0x2001, 0x102d)]))
number_of_stacks = int(
int(multiframe_dicom.NumberOfFrames) / number_of_stack_slices)
# We create a numpy array
size_x = multiframe_dicom.pixel_array.shape[2]
size_y = multiframe_dicom.pixel_array.shape[1]
size_z = number_of_stack_slices
size_t = number_of_stacks
# get the format
format_string = common.get_numpy_type(multiframe_dicom)
# get header info needed for ordering
frame_info = multiframe_dicom[0x5200, 0x9230]
data_4d = numpy.zeros((size_z, size_y, size_x, size_t),
dtype=format_string)
# loop over each slice and insert in datablock
t_location_index = _get_t_position_index(multiframe_dicom)
for slice_index in range(0, size_t * size_z):
z_location = frame_info[slice_index].FrameContentSequence[
0].InStackPositionNumber - 1
if t_location_index is None:
t_location = frame_info[slice_index].FrameContentSequence[
0].TemporalPositionIndex - 1
else:
t_location = frame_info[slice_index].FrameContentSequence[
0].DimensionIndexValues[t_location_index] - 1
block_data = multiframe_dicom.pixel_array[slice_index, :, :]
# apply scaling
rescale_intercept = frame_info[
slice_index].PixelValueTransformationSequence[0].RescaleIntercept
rescale_slope = frame_info[
slice_index].PixelValueTransformationSequence[0].RescaleSlope
private_scale_slope = 1.0
private_scale_intercept = 0.0
private_sequence_tag = Tag(0x2005, 0x140f)
private_scale_intercept_tag = Tag(0x2005, 0x100d)
private_scale_slope_tag = Tag(0x2005, 0x100e)
if private_sequence_tag in frame_info[slice_index]:
if private_scale_intercept_tag in frame_info[slice_index][
private_sequence_tag][0]:
private_scale_intercept = common.get_fl_value(
frame_info[slice_index][private_sequence_tag][0]
[private_scale_intercept_tag])
if private_scale_slope_tag in frame_info[slice_index][
private_sequence_tag][0]:
private_scale_slope = common.get_fl_value(
frame_info[slice_index][private_sequence_tag][0]
[private_scale_slope_tag])
block_data = common.do_scaling(block_data, rescale_slope,
rescale_intercept, private_scale_slope,
private_scale_intercept)
# switch to float if needed
if block_data.dtype != data_4d.dtype:
data_4d = data_4d.astype(block_data.dtype)
data_4d[z_location, :, :, t_location] = block_data
full_block = numpy.zeros((size_x, size_y, size_z, size_t),
dtype=data_4d.dtype)
# loop over each stack and reorganize the data
for t_index in range(0, size_t):
# transpose the block so the directions are correct
data_3d = numpy.transpose(data_4d[:, :, :, t_index], (2, 1, 0))
# add the block the the full data
full_block[:, :, :, t_index] = data_3d
return full_block
