def _force_consistent_uids(data_dir: Union[str, PathLike],
target_uid: str = None):
"""
Forces all DICOMs inside `data_dir` to have consistent study instance UID
Will recursively crawl sub-directories of `data_dir` to check for DICOMs
Parameters
----------
data_dir : str or pathlib.Path
Path to data directory containg DICOMs that should be overwritten with
new study instance UID.
target_uid : str, optional
Study instance UID to assign to all DICOMs found in `data_dir`. If not
specified the study instance UID from the first DICOM found will be
used instead. Default: None
"""
data_dir = pathlib.Path(data_dir).resolve()
for n, fn in enumerate(data_dir.rglob('*dcm')):
img = dcm.read_file(str(fn))
if target_uid is None and n == 0:
target_uid = str(img[('0020', '000d')].value)
continue
img[('0020', '000d')].value = target_uid
dcm.write_file(str(fn), img)
def create_new_uids(folder, dicom_tag, dry_run=True):
"""Search folder (and subfolders) for DICOM files and create new UIDs for specified DICOM tag
Parameters
----------
folder : Folder to (recursively) search for DICOM objects
dicom_tag : DICOM tag specifying which UID to create new UIDs for (format as "0x0020000D")
dry_run : Dry run, (True)/False
"""
# find dcm files
dcm_files = glob.glob(os.path.join(folder,"**","*.dcm"), recursive=True)
dcm_dict = dict()
# loop over dcm files
for dcm_file in dcm_files:
dcm = pydicom.read_file(dcm_file)
# if dicom tag exists in current dcm file
if int(dicom_tag, 16) in dcm:
# record replaceed uid and set new uid
uid = dcm[dicom_tag].value
if uid not in dcm_dict:
dcm_dict[uid] = pydicom.uid.generate_uid()
dcm[dicom_tag] = DataElement(int(dicom_tag,16), "UI", dcm_dict[uid])
else:
# add dicom tag with new uid
uid = pydicom.uid.generate_uid()
dcm[dicom_tag] = DataElement(int(dicom_tag,16), "UI", uid)
if not dry_run:
# write dcm file to disk
pydicom.write_file(dcm_file, dcm, write_like_original=False)
if dry_run:
print("Uids to update")
print(dcm_file)
def save3dDicom(volume,
info,
path,
newSeriesNumber=None,
newSeriesDescription=None,
newImageComment=None,
startImageNumber=1):
try:
os.mkdir(path)
except:
pass
if newSeriesNumber is None:
newSeriesNumber = info[0].SeriesNumber
newSeriesUID = info[0].SeriesInstanceUID
else:
newSeriesUID = UID.generate_uid()
if newSeriesDescription is None:
newSeriesDescription = info[0].SeriesDescription
if len(info) == 1:
dataArray = volume[:, :, 0]
dicomFileData = copy.deepcopy(info[0]) # create a copy of object
if dataArray.dtype == 'uint16': # check correct format
dicomFileData.PixelData = dataArray.tostring()
else:
dicomFileData.PixelData = dataArray.round().astype(
'uint16').tostring()
dicomFileData.SeriesNumber = newSeriesNumber
dicomFileData.SeriesInstanceUID = newSeriesUID
dicomFileData.SOPInstanceUID = UID.generate_uid()
if newImageComment is not None:
dicomFileData.ImageComment = newImageComment
fName = os.path.join(
path, "image0001.dcm") # if one wants to save a part of a dataset
dicom.write_file(fName, dicomFileData)
else:
bar = Bar('Saving Dicom', max=len(info))
for sl in range(len(info)):
dataArray = volume[..., sl]
dicomFileData = copy.deepcopy(info[sl]) # create a copy of object
if dataArray.dtype == 'uint16': # check correct format
dicomFileData.PixelData = dataArray.tostring()
else:
dicomFileData.PixelData = dataArray.round().astype(
'uint16').tostring()
dicomFileData.SeriesNumber = newSeriesNumber
dicomFileData.SeriesInstanceUID = newSeriesUID
dicomFileData.SOPInstanceUID = UID.generate_uid()
if newImageComment is not None:
dicomFileData.ImageComment = newImageComment
fName = os.path.join(
path, "image%04d.dcm" %
(sl +
startImageNumber)) # if one wants to save a part of a dataset
dicom.write_file(fName, dicomFileData)
bar.next()
bar.finish()
def nii2dcm(dataset, filename):
"""
write dicom filename
filename = 'test.dcm'
"""
# TODO transfer array to dicom directly
pydicom.write_file(filename, dataset)
def write(self, outdir='.', print_filenames=False):
for modality in self.modalityorder:
for sb in self.seriesbuilders[modality]:
print modality, sb
for ds in sb.build():
dicom.write_file(os.path.join(outdir, ds.filename), ds)
if print_filenames:
print ds.filename
def write(self, outdir='.', print_filenames=False):
for modality in self.modalityorder:
for sb in self.seriesbuilders[modality]:
print modality, sb
for ds in sb.build():
dicom.write_file(os.path.join(outdir, ds.filename), ds)
if print_filenames:
print ds.filename
def test_non_fatal_errors(self):
dataset = Dataset()
dataset.SOPClassUID = '1.2.840.10008.5.1.4.1.1.2' # CT Image Storage
file_dataset = FileDataset('test', dataset, file_meta=self.create_metadata())
write_file('test', file_dataset, write_like_original=False)
error_dict = self.validator.validate('test')
self.assertEqual(1, len(error_dict))
errors = error_dict['test']
self.assertNotIn('fatal', errors)
def write_to_file(self, output_file, write_like_original=False):
"""Writes the current IOD to file
Parameters
----------
output_file : Complete path of file to write to
"""
write_file(output_file,
self.dataset,
write_like_original=write_like_original)
def maskTheInfo(inp, opRoot, indexBegin=0):
if not os.path.exists(opRoot):
os.mkdir(opRoot)
patientNames = natsort.natsorted((os.listdir(inp)))
index = indexBegin
menu = []
for patientFileId in patientNames:
patientOutRootPath = os.path.join(opRoot, str(index))
dirCtPath = os.path.join(inp, patientFileId, config.ctInDir)
print(dirCtPath + ':' + str(index))
fileCtNames = os.listdir(dirCtPath)
# patientDirPath = os.path.join(opRoot, str(index))
patientDirPath = os.path.join(opRoot, str(index))
os.mkdir(patientDirPath)
opCtDir = os.path.join(patientOutRootPath, config.ctOutDir)
os.mkdir(opCtDir)
for fileCtName in fileCtNames:
if fileCtName.startswith('CT_'):
meta = pdm.read_file(os.path.join(dirCtPath, fileCtName))
id = meta.PatientID
meta.__setattr__('PatientID', str(index))
meta.__setattr__('Accession Number', str(index)) #这里也包含了患者的id
meta.__setattr__('PatientName', str(index))
pdm.write_file(os.path.join(opCtDir, fileCtName + '.dcm'),
meta, True)
dirPetPath = os.path.join(inp, patientFileId, config.ptInDir)
filePetNames = os.listdir(dirPetPath)
opPetDir = os.path.join(patientDirPath, config.ptOutDir)
os.mkdir(opPetDir)
for filePetName in filePetNames:
if filePetName.startswith('PT_'):
petmeta = pdm.read_file(os.path.join(dirPetPath, filePetName))
pid = petmeta.PatientID
petmeta.__setattr__('PatientID', str(index))
meta.__setattr__('Accession Number', str(index)) # 这里也包含了患者的id
petmeta.__setattr__('PatientName', str(index))
pdm.write_file(os.path.join(opPetDir, filePetName + '.dcm'),
petmeta, True)
#保存文档
tryDoc = glob.glob(
os.path.join(inp, patientFileId, '*' + config.docEnd))
docExist = len(tryDoc) != 0
if docExist:
docSrcPath = tryDoc[0]
docDesPath = os.path.join(patientDirPath,
str(index) + 'report' + config.docEnd)
shutil.copy(docSrcPath, docDesPath)
menu.append((index, pid + '_' + patientFileId))
index = index + 1
writeDic(menu)
def fix_file_and_write(dicom_file: str,
dicom_fixed_file: str,
anatomy: tuple,
reference: dict,
dicom_fix_report_file: str = ''):
ds, fix_report = fix_file(dicom_file, anatomy, reference)
pydicom.write_file(dicom_fixed_file, ds)
if dicom_fix_report_file:
ctools.WriteStringToFile(dicom_fix_report_file, fix_report)
return (ds, fix_report)
def generate_dicom_scans(dst, num_scans=10, intercept=0, slope=1):
spacing = (0.4 + 0.4 * np.random.rand(num_scans, 3) +
np.array([1 + 0.5 * np.random.rand(), 0, 0]))
origin = np.random.randint(-200, 200, (num_scans, 3))
for i in range(num_scans):
num_slices = np.random.randint(128, 169)
scan_id = np.random.randint(2**16)
scan_data = np.random.randint(0, 256, (num_slices, 128, 128))
folder = os.path.join(dst,
hex(scan_id).replace('x', '').upper().zfill(8))
if not os.path.exists(folder):
os.makedirs(folder)
for k in range(num_slices):
slice_name = (hex(scan_id + k).replace('x', '').upper().zfill(8))
filename = os.path.join(folder, slice_name)
pixel_array = (scan_data[k, ...] - intercept) / slope
locZ = float(origin[i, 0] + spacing[i, 0] * k)
locY, locX = float(origin[i, 1]), float(origin[i, 2])
file_meta = DicomDataset()
file_meta.MediaStorageSOPClassUID = "Secondary Capture Image Storage"
file_meta.MediaStorateSOPInstanceUID = (hex(scan_id).replace(
'x', '').upper().zfill(8))
file_meta.ImplementationClassUID = slice_name
dataset = DicomFileDataset(filename, {},
file_meta=file_meta,
preamble=b"\0" * 128)
dataset.PixelData = pixel_array.astype(np.uint16).tostring()
dataset.RescaleSlope = slope
dataset.RescaleIntercept = intercept
dataset.ImagePositionPatient = MultiValue(
type_constructor=float, iterable=[locZ, locY, locX])
dataset.PixelSpacing = MultiValue(
type_constructor=float,
iterable=[float(spacing[i, 1]),
float(spacing[i, 2])])
dataset.SliceThickness = float(spacing[i, 0])
dataset.Modality = 'WSD'
dataset.Columns = pixel_array.shape[0]
dataset.Rows = pixel_array.shape[1]
dataset.file_meta.TransferSyntaxUID = pydicom.uid.ImplicitVRLittleEndian
dataset.PixelRepresentation = 1
dataset.BitsAllocated = 16
dataset.BitsStored = 16
dataset.SamplesPerPixel = 1
write_file(filename, dataset)
def compare_dicom_cli(command, original, expected):
pydicom.write_file(ORIGINAL_DICOM_FILENAME, original)
try:
subprocess.check_call(command)
cli_adjusted_ds = pydicom.read_file(ADJUSTED_DICOM_FILENAME, force=True)
assert str(cli_adjusted_ds) == str(expected)
finally:
remove_file(ORIGINAL_DICOM_FILENAME)
remove_file(ADJUSTED_DICOM_FILENAME)
def update_frame_of_ref(root):
finished = os.path.join(root, 'Finished.txt')
folder_dict = {}
image_reader = sitk.ImageFileReader()
dicom_files = [
os.path.join(root, i) for i in os.listdir(root) if i.endswith('.dcm')
]
for file in dicom_files:
image_reader.SetFileName(file)
try:
image_reader.Execute()
except:
continue
rows_cols = '{}_{}'.format(image_reader.GetMetaData("0028|0010"),
image_reader.GetMetaData("0028|0011"))
if rows_cols not in folder_dict:
folder_dict[rows_cols] = {
'files': [],
'uid': None,
'needs_uid': []
}
if image_reader.HasMetaDataKey(
"0020|0052") and image_reader.GetMetaData(
"0020|0052") is not '':
if folder_dict[rows_cols]['uid'] is None:
folder_dict[rows_cols]['uid'] = image_reader.GetMetaData(
"0020|0052")
folder_dict[rows_cols]['needs_uid'].append(False)
else:
folder_dict[rows_cols]['needs_uid'].append(True)
folder_dict[rows_cols]['files'].append(file)
for key in folder_dict:
base_uid = folder_dict[key]['uid']
if base_uid is None:
base_uid = '1.3.12.2.1107.5.1.4.95642.30000017092813491242400000003.{}'.format(
np.random.randint(9999))
file_list = np.asarray(
folder_dict[key]['files'])[folder_dict[key]['needs_uid']]
for file in file_list:
ds = pydicom.read_file(file)
ds.FrameOfReferenceUID = base_uid
pydicom.write_file(filename=file, dataset=ds)
if len(folder_dict) > 1:
os.makedirs(os.path.join(root, key))
for file in folder_dict[key]['files']:
file_name = os.path.split(file)[-1]
os.rename(file, os.path.join(root, key, file_name))
fid = open(os.path.join(root, key, 'Finished.txt'), 'w+')
fid.close()
else:
fid = open(finished, 'w+')
fid.close()
def adjust_RED_cli(args):
adjustment_map = {
key: item
for key, item in zip(
args.adjustment_map[::2], args.adjustment_map[1::2])
}
dicom_dataset = pydicom.read_file(args.input_file, force=True)
new_dicom_dataset = adjust_rel_elec_density(
dicom_dataset, adjustment_map,
ignore_missing_structure=args.ignore_missing_structure)
pydicom.write_file(args.output_file, new_dicom_dataset)
def change_name_and_pat_id(patient_id, root):
print(root)
files = [i for i in os.listdir(root) if i.startswith('IM')]
for file in files:
if not file.endswith('.dcm'):
os.rename(os.path.join(root, file),
os.path.join(root, file + '.dcm'))
file += '.dcm'
ds = pydicom.read_file(os.path.join(root, file))
if ds.PatientID == patient_id: # Already have converted the files in this folder
break
ds.PatientID = patient_id
pydicom.write_file(os.path.join(root, file), ds)
return None
def pseudo_anonymize(dicom_file):
try:
data = read_file(dicom_file)
if hasattr(data, 'PatientID'):
data.PatientID = anonymize(data.PatientID)
if hasattr(data, 'PatientName'):
data.PatientName = ''
if hasattr(data, 'PatientBirthDate'):
data.PatientBirthDate = data.PatientBirthDate[0:4] + "0101"
if hasattr(data, 'PerformingPhysicianName'):
data.PerformingPhysicianName = ''
write_file(dicom_file, data)
except:
pass
def compare_dicom_cli(command, original, expected):
# TODO: Extract CLI definition from subpackages and test separately
if "SUBPACKAGE" not in os.environ:
pydicom.write_file(ORIGINAL_DICOM_FILENAME, original)
try:
subprocess.check_call(command)
cli_adjusted_ds = pydicom.read_file(ADJUSTED_DICOM_FILENAME,
force=True)
assert str(cli_adjusted_ds) == str(expected)
finally:
remove_file(ORIGINAL_DICOM_FILENAME)
remove_file(ADJUSTED_DICOM_FILENAME)
def run_ano(self):
"""
Reads the DICOM file, anonymizes it and write the result.
"""
if not self.ano_run:
self._dataset.walk(self._anonymize_check)
# Patient Identity Removed
self._dataset.add_new((0x0012, 0x0062), 'CS', 'YES')
# De-identification Method
method = f'CATI DEIDENTIFICATION - {deid.__version__}'
if self.config_profile:
method += f' - {self.config_profile}'
self._dataset.add_new((0x0012, 0x0063), 'LO', method)
pydicom.write_file(self._dicom_fileout, self._dataset)
return 1
def add(self, contours, roiname, roinumber, color):
def newROI():
old = self.findFirstClosedPlanarROI()
new = copy.deepcopy(old)
new.ROINumber = roinumber
new.ROIGenerationAlgorithm = 'PW'
new.ROIName = asASCII(roiname)
return new
self.dataset.StructureSetROIs.append(newROI())
def newContours():
old = self.findFirstClosedPlanarContour()
new = copy.deepcopy(old)
new.RefdROINumber = roinumber
new.ReferencedROINumber = roinumber
new.ROIDisplayColor = color
template = copy.deepcopy(new.Contours[0])
del new.Contours[:]
for c in contours:
c = c.flatten()
numberOfPoints = int(len(c) / 3)
if numberOfPoints > 1:
nc = copy.deepcopy(template)
nc.NumberofContourPoints = numberOfPoints
nc.ContourData = ['%g' % x for x in c]
nc.ContourGeometricType = 'CLOSED_PLANAR'
new.Contours.append(nc)
return new
self.dataset.ROIContours.append(newContours())
def newObservation():
old = self.findLastObservation()
new = copy.deepcopy(old)
new.ObservationNumber = old.ObservationNumber + 1
new.ROIObservationLabel = asASCII(roiname)
new.ReferencedROINumber = roinumber
new.RefdROINumber = roinumber
return new
self.dataset.RTROIObservations.append(newObservation())
# overwrite the dataset
pydicom.write_file(self.dataset.filename, self.dataset)
def handle_and_mv_export_dir(data, file):
data.walk(person_names_callback)
try:
storage = data.MIMETypeOfEncapsulatedDocument
if storage == 'application/pdf':
os.remove(file)
else:
id = get_id(data)
if not os.path.isdir(os.path.join(EXPORT_DIR, id)):
os.mkdir(os.path.join(EXPORT_DIR, id))
write_file(os.path.join(EXPORT_DIR, id, os.path.basename(file)),
data)
os.remove(file)
except Exception:
id = get_id(data)
if not os.path.isdir(os.path.join(EXPORT_DIR, id)):
os.mkdir(os.path.join(EXPORT_DIR, id))
write_file(os.path.join(EXPORT_DIR, id, os.path.basename(file)), data)
os.remove(file)
def add_frame_of_reference_uid(path):
file_reader = sitk.ImageFileReader()
for root, dirs, files in os.walk(path):
has_ref_file = os.path.join(root, 'HasFrameOfReferenceUID.txt')
if os.path.exists(has_ref_file):
continue
dicom_files = [
os.path.join(root, i) for i in files if i.endswith('.dcm')
]
if dicom_files:
base_uid = '1.3.12.2.1107.5.1.4.95642.30000017092813491242400000003.{}'.format(
np.random.randint(9999))
for file in dicom_files:
file_reader.SetFileName(file)
file_reader.Execute()
if not file_reader.HasMetaDataKey("0020|0052"):
ds = pydicom.read_file(file)
ds.FrameOfReferenceUID = base_uid
pydicom.write_file(filename=file, dataset=ds)
fid = open(has_ref_file, 'w+')
fid.close()
return None
def save_as(self, filename, write_like_original=True):
"""Write the dataset to a file.
Parameters
----------
filename : str
Name of file to save new DICOM file to.
write_like_original : boolean
If True (default), preserves the following information from
the dataset:
-preamble -- if no preamble in read file, than not used here
-hasFileMeta -- if writer did not do file meta information,
then don't write here either
-seq.is_undefined_length -- if original had delimiters, write them now too,
instead of the more sensible length characters
- is_undefined_length_sequence_item -- for datasets that belong to a
sequence, write the undefined length delimiters if that is
what the original had.
If False, produces a "nicer" DICOM file for other readers,
where all lengths are explicit.
See Also
--------
pydicom.filewriter.write_file
Write a DICOM file from a FileDataset instance.
Notes
-----
Set dataset.preamble if you want something other than 128 0-bytes.
If the dataset was read from an existing dicom file, then its preamble
was stored at read time. It is up to the user to ensure the preamble is still
correct for its purposes.
If there is no Transfer Syntax tag in the dataset, then set
dataset.is_implicit_VR and dataset.is_little_endian
to determine the transfer syntax used to write the file.
"""
pydicom.write_file(filename, self, write_like_original)
def save_as(self, filename, write_like_original=True):
"""Write the dataset to a file.
Parameters
----------
filename : str
Name of file to save new DICOM file to.
write_like_original : boolean
If True (default), preserves the following information from
the dataset:
-preamble -- if no preamble in read file, than not used here
-hasFileMeta -- if writer did not do file meta information,
then don't write here either
-seq.is_undefined_length -- if original had delimiters, write them now too,
instead of the more sensible length characters
- is_undefined_length_sequence_item -- for datasets that belong to a
sequence, write the undefined length delimiters if that is
what the original had.
If False, produces a "nicer" DICOM file for other readers,
where all lengths are explicit.
See Also
--------
pydicom.filewriter.write_file
Write a DICOM file from a FileDataset instance.
Notes
-----
Set dataset.preamble if you want something other than 128 0-bytes.
If the dataset was read from an existing dicom file, then its preamble
was stored at read time. It is up to the user to ensure the preamble is still
correct for its purposes.
If there is no Transfer Syntax tag in the dataset, then set
dataset.is_implicit_VR and dataset.is_little_endian
to determine the transfer syntax used to write the file.
"""
pydicom.write_file(filename, self, write_like_original)
def Decompress(FileInput, FileOutput): # Descomprime el dicom en si
Out = dc.read_file(FileInput, force=True)
# Out.file_meta.TransferSyntaxUID = dc.uid.ImplicitVRLittleEndian # or whatever is the correct transfer syntax for the file
Out.decompress()
dc.write_file(FileOutput, Out)
def sabr_deid(participant_info, scan_df, raw_dir, deid_outdir):
#Join raw dir with subject name (assumes directory structure is ./rawdir/subj_name/...
subj_main_dir = os.path.join(raw_dir, participant_info['participant_name'])
#Adds BIDS required specifier 'sub' to new id
participant_id = 'sub-' + participant_info['participant_id']
#Get list of sessions within main subj directory, make dir and loop over sessions.
subj_sessions = next(os.walk(subj_main_dir))[1]
subj_sessions.sort()
print('\n***{} has {} session(s)'.format(participant_id,
len(subj_sessions)))
#Create deidentified main (root) directory for subject
subj_deid_main_dir = os.path.join(deid_outdir, participant_id)
try:
os.mkdir(subj_deid_main_dir)
except:
print('\nDirectory {} exists\n'.format(subj_deid_main_dir))
#WARNING! LAZY CODING AHEAD!
if len(subj_sessions) == 0:
raise Exception(
'\n***ERROR! NUMBER OF SESSIONS = 0!***\nPlease check directory structure of {}'
.format(participant_info['participant_name']))
elif len(subj_sessions) == 1:
session = subj_sessions[0]
subj_session_dir = os.path.join(subj_main_dir, session)
for j, scan_type in enumerate(scan_df['scan_type']):
subj_deid_meta_dir = os.path.join(subj_deid_main_dir, scan_type)
try:
os.mkdir(subj_deid_meta_dir)
except:
print('Meta directory {} exists.'.format(scan_type))
#Match common sequence substring with path in os.walk
for root, dr, files in os.walk(subj_session_dir):
match = scan_df.scan_match[j]
match_regex = fnmatch.translate(match)
found = re.search(match_regex, root)
print('\n***{}***\n'.format(found))
#If match, start deid process. If not, move onto next folder.
if found != None:
subj_deid_sequence_dir = os.path.join(
subj_deid_meta_dir, scan_df.scan_filename[j])
print('Making directory {}'.format(subj_deid_sequence_dir))
try:
os.mkdir(
subj_deid_sequence_dir
) #Make "housing" directory to keep dicoms of different sequences but same meta-category separate.
except:
print(
'\n***SEQUENCE DIRECTORY ALREADY EXISTS!***\nSkipping.'
)
continue
#Create list of dicoms in sequence dir rather than use
#files (more control in case any non-dicoms)
anon_files = os.listdir(root)
anon_files = [
x for x in anon_files if 'nii' not in x
] #Remove any previous nii files that may be present < To do - expand to other file types (mgh, analyze, etc)
anon_files.sort()
for anon_file in anon_files:
#Read files in 1 at a time, remove the remove / alter the below tags.
dcm = pydicom.read_file(
os.path.join(root, anon_file), force=True
) #Uses force = True incase dicoms haven't had identifier added to header
#Strip aquisition date information
aqusition_date_list = [[0x0008, 0x0020],
[0x0008, 0x0021],
[0x0008, 0x0022],
[0x0008, 0x0023]]
for tag in aqusition_date_list:
try:
dcm[hex(tag[0]), hex(tag[1])].value = ''
except:
print(
'Tag {} {} does not exist in {}. Moving to next tag'
.format(hex(tag[0]), hex(tag[1]),
scan_df.scan_filename[j]))
#Strip aquisition time information
aqusition_time_list = [[0x0008, 0x0030],
[0x0008, 0x0031],
[0x0008, 0x0032],
[0x0008, 0x0033]]
for tag in aqusition_time_list:
try:
dcm[hex(tag[0]), hex(tag[1])].value = ''
except:
print(
'Tag {} {} does not exist in {}. Moving to next tag'
.format(hex(tag[0]), hex(tag[1]),
scan_df.scan_filename[j]))
#Strip physician information
physician_list = [[0x0008, 0x0090], [0x0008, 0x1050]]
for tag in physician_list:
try:
dcm[hex(tag[0]), hex(tag[1])].value = ''
except:
print(
'Tag {} {} does not exist in {}. Moving to next tag'
.format(hex(tag[0]), hex(tag[1]),
scan_df.scan_filename[j]))
#Strip study description
#dcm[0x0008,0x1030].value = ''
#Strip subject name / patient ID
subj_name_list = [[0x0010, 0x0010], [0x0010, 0x0020]]
#PatientName, PatientID
for tag in subj_name_list:
try:
dcm[hex(tag[0]),
hex(tag[1])].value = participant_id
except:
print(
'Tag {} {} does not exist in {}. Moving to next tag'
.format(hex(tag[0]), hex(tag[1]),
scan_df.scan_filename[j]))
#Strip subject attributes
subj_attrib_list = [[0x0010, 0x0030], [0x0010, 0x1010],
[0x0010, 0x1020], [0x0010, 0x1030]]
#, DoB, Age, PatientHeight, PatientWeight
for tag in subj_attrib_list:
try:
dcm[hex(tag[0]), hex(tag[1])].value = ''
except:
print(
'Tag {} {} does not exist in {}. Moving to next tag'
.format(hex(tag[0]), hex(tag[1]),
scan_df.scan_filename[j]))
#Write anonymised file
pydicom.write_file(
os.path.join(subj_deid_sequence_dir, anon_file),
dcm)
elif len(subj_sessions) > 1:
for sn, session in enumerate(subj_sessions):
#MAKE DIRECTORIES BUT ZERO PAD SESSION
subj_deid_session_dir = os.path.join(
subj_deid_main_dir, 'ses-'
'{:02d}'.format(sn + 1))
try:
os.mkdir(subj_deid_session_dir)
except:
print('\nSession folder {} exists\n'.format(
subj_deid_session_dir))
#Session folder for identifiable subject
subj_session_dir = os.path.join(subj_main_dir, session)
#Loop over scan folder types within scan dataframe (anat, task, etc)
for j, scan_type in enumerate(scan_df['scan_type']):
subj_deid_meta_dir = os.path.join(subj_deid_session_dir,
scan_type)
try:
os.mkdir(subj_deid_meta_dir)
except:
print('Meta directory {} exists.'.format(scan_type))
#Match common sequence substring with path in os.walk
for root, dr, files in os.walk(subj_session_dir):
match = scan_df.scan_match[j]
match_regex = fnmatch.translate(match)
found = re.search(match_regex, root)
#If match, start deid process, not not, move onto next folder.
if found != None:
subj_deid_sequence_dir = os.path.join(
subj_deid_meta_dir, scan_df.scan_filename[j])
print('Making directory {}'.format(
subj_deid_sequence_dir))
try:
os.mkdir(
subj_deid_sequence_dir
) #Make "housing" directory to keep dicoms of different sequences but same meta-category separate.
except:
print(
'\n***SEQUENCE DIRECTORY ALREADY EXISTS!***\nSkipping.'
)
continue
#Create list of dicoms in sequence dir rather than use
#files (more control in case any non-dicoms)
anon_files = os.listdir(root)
anon_files = [
x for x in anon_files if 'nii' not in x
] #Remove any previous nii files that may be present < To do - expand to other file types (mgh, analyze, etc)
anon_files.sort()
for anon_file in anon_files:
#Read files in 1 at a time, remove the remove / alter the below tags.
dcm = pydicom.read_file(
os.path.join(root, anon_file), force=True
) #Uses force = True incase dicoms haven't had identifier added to header
#Strip aquisition date information
aqusition_date_list = [[0x0008, 0x0020],
[0x0008, 0x0021],
[0x0008, 0x0022],
[0x0008, 0x0023]]
for tag in aqusition_date_list:
try:
dcm[hex(tag[0]), hex(tag[1])].value = ''
except:
print(
'Tag {} {} does not exist in {}. Moving to next tag'
.format(hex(tag[0]), hex(tag[1]),
scan_df.scan_filename[j]))
#Strip aquisition time information
aqusition_time_list = [[0x0008, 0x0030],
[0x0008, 0x0031],
[0x0008, 0x0032],
[0x0008, 0x0033]]
for tag in aqusition_time_list:
try:
dcm[hex(tag[0]), hex(tag[1])].value = ''
except:
print(
'Tag {} {} does not exist in {}. Moving to next tag'
.format(hex(tag[0]), hex(tag[1]),
scan_df.scan_filename[j]))
#Strip physician information
physician_list = [[0x0008, 0x0090],
[0x0008, 0x1050]]
for tag in physician_list:
try:
dcm[hex(tag[0]), hex(tag[1])].value = ''
except:
print(
'Tag {} {} does not exist in {}. Moving to next tag'
.format(hex(tag[0]), hex(tag[1]),
scan_df.scan_filename[j]))
#Strip study description
#dcm[0x0008,0x1030].value = ''
#Strip subject name / patient ID
subj_name_list = [[0x0010, 0x0010],
[0x0010, 0x0020]]
#PatientName, PatientID
for tag in subj_name_list:
try:
dcm[hex(tag[0]),
hex(tag[1])].value = participant_id
except:
print(
'Tag {} {} does not exist in {}. Moving to next tag'
.format(hex(tag[0]), hex(tag[1]),
scan_df.scan_filename[j]))
#Strip subject attributes
subj_attrib_list = [[0x0010, 0x0030],
[0x0010, 0x1010],
[0x0010, 0x1020],
[0x0010, 0x1030]]
#, DoB, Age, PatientHeight, PatientWeight
for tag in subj_attrib_list:
try:
dcm[hex(tag[0]), hex(tag[1])].value = ''
except:
print(
'Tag {} {} does not exist in {}. Moving to next tag'
.format(hex(tag[0]), hex(tag[1]),
scan_df.scan_filename[j]))
#Write anonymised file
pydicom.write_file(
os.path.join(subj_deid_sequence_dir,
anon_file), dcm)
return (subj_deid_main_dir, subj_sessions, participant_id)
[purposeOfReferenceDataset])
sr_out.ReferencedInstanceSequence = Sequence([referencedSequenceDataset])
sr_out.ValueType = "CONTAINER"
conceptNameDataset = Dataset()
conceptNameDataset.CodeValue = "126000"
conceptNameDataset.CodingSchemeDesignator = "DCM"
conceptNameDataset.CodeMeaning = "Imaging Measurement Report"
sr_out.ConceptCodeSequence = Sequence([conceptNameDataset])
referencedSopSequence = Dataset()
referencedSopSequence.ReferencedSOPClassUID = dcm_in.SOPClassUID
referencedSopSequence.ReferencedSOPInstanceUID = dcm_in.SOPInstanceUID
sr_out.ReferencedSOPSequence = Sequence([referencedSopSequence])
nameDataset = Dataset()
nameDataset.RelationshipType = "CONTAINS"
nameDataset.ValueType = "TEXT"
nameConceptNameDataset = Dataset()
nameConceptNameDataset.CodeValue = "371484003"
nameConceptNameDataset.CodingSchemeDesignator = "SNM3"
nameConceptNameDataset.CodeMeaning = "Patient name (observable entity)"
nameDataset.ConceptNameCodeSequence = Sequence([nameConceptNameDataset])
nameDataset.TextValue = display_name
sr_out.ContentSequence = Sequence([nameDataset])
pydicom.write_file('/envoyai/output/sr.dcm', sr_out)
print("SeriesInstanceUid............:", sr_out.SeriesInstanceUID)
print("SopInstanceUid...............:", sr_out.SOPInstanceUID)
print("MediaStorageSOPInstanceUID...:",
sr_out.file_meta.MediaStorageSOPInstanceUID)
pydicom.write_file('/envoyai/output/out.dcm', sr_out)
exit(0)
def test_missing_sop_class(self):
filename = 'test.dcm'
file_dataset = FileDataset(filename, Dataset(), file_meta=self.create_metadata())
write_file(filename, file_dataset, write_like_original=False)
self.assert_fatal_error(filename, 'Missing SOPClassUID')
def test_unknown_sop_class(self):
dataset = Dataset()
dataset.SOPClassUID = 'Unknown'
file_dataset = FileDataset('test', dataset, file_meta=self.create_metadata())
write_file('test', file_dataset, write_like_original=False)
self.assert_fatal_error('test', 'Unknown SOPClassUID (probably retired): Unknown')
def patchDicomDir(self, inputDirPath, outputDirPath):
"""
Since CTK (rightly) requires certain basic information [1] before it can import
data files that purport to be dicom, this code patches the files in a directory
with some needed fields.
Calling this function with a directory path will make a patched copy of each file.
Importing the old files to CTK should still fail, but the new ones should work.
The directory is assumed to have a set of instances that are all from the
same study of the same patient. Also that each instance (file) is an
independent (multiframe) series.
[1] https://github.com/commontk/CTK/blob/16aa09540dcb59c6eafde4d9a88dfee1f0948edc/Libs/DICOM/Core/ctkDICOMDatabase.cpp#L1283-L1287
"""
import pydicom
self.addLog('DICOM patching started...')
logging.debug('DICOM patch input directory: '+inputDirPath)
logging.debug('DICOM patch output directory: '+outputDirPath)
for rule in self.patchingRules:
rule.logCallback = self.addLog
rule.processStart(inputDirPath, outputDirPath)
for root, subFolders, files in os.walk(inputDirPath):
currentSubDir = os.path.relpath(root, inputDirPath)
rootOutput = os.path.join(outputDirPath, currentSubDir)
# Notify rules that processing of a new subdirectory started
for rule in self.patchingRules:
rule.processDirectory(currentSubDir)
for file in files:
filePath = os.path.join(root,file)
self.addLog('Examining %s...' % os.path.join(currentSubDir,file))
skipFileRequestingRule = None
for rule in self.patchingRules:
if rule.skipFile(currentSubDir):
skipFileRequestingRule = rule
break
if skipFileRequestingRule:
self.addLog(' Rule '+rule.__class__.__name__+' requested to skip this file.')
continue
try:
ds = pydicom.read_file(filePath)
except (OSError, pydicom.filereader.InvalidDicomError):
self.addLog(' Not DICOM file. Skipped.')
continue
self.addLog(' Patching...')
for rule in self.patchingRules:
rule.processDataSet(ds)
patchedFilePath = os.path.abspath(os.path.join(rootOutput,file))
for rule in self.patchingRules:
patchedFilePath = rule.generateOutputFilePath(ds, patchedFilePath)
######################################################
# Write
dirName = os.path.dirname(patchedFilePath)
if not os.path.exists(dirName):
os.makedirs(dirName)
self.addLog(' Writing DICOM...')
pydicom.write_file(patchedFilePath, ds)
self.addLog(' Created DICOM file: %s' % patchedFilePath)
self.addLog(f'DICOM patching completed. Patched files are written to:\n{outputDirPath}')
def Mask_to_Contours(self):
self.RefDs = self.ds
self.shift_list = [[
float(i)
for i in self.reader.GetMetaData(j, "0020|0032").split('\\')
] for j in range(len(self.reader.GetFileNames()))
] #ShiftRows, ShiftCols, ShiftZBase
self.mv = Xx, Xy, Xz, Yx, Yy, Yz = [
float(i)
for i in self.reader.GetMetaData(0, "0020|0037").split('\\')
]
self.ShiftRows = [
i[0] * Xx + i[1] * Xy + i[2] * Xz for i in self.shift_list
]
self.ShiftCols = [
i[0] * Xy + i[1] * Yy + i[2] * Yz for i in self.shift_list
]
self.ShiftZ = [i[2] for i in self.shift_list]
self.mult1 = self.mult2 = 1
self.PixelSize = self.dicom_handle.GetSpacing()
current_names = []
for names in self.RS_struct.StructureSetROISequence:
current_names.append(names.ROIName)
Contour_Key = {}
xxx = 1
for name in self.ROI_Names:
Contour_Key[name] = xxx
xxx += 1
self.all_annotations = self.annotations
base_annotations = copy.deepcopy(self.annotations)
temp_color_list = []
color_list = [[128, 0, 0], [170, 110, 40], [0, 128, 128], [0, 0, 128],
[230, 25, 75], [225, 225, 25], [0, 130, 200],
[145, 30, 180], [255, 255, 255]]
self.struct_index = 0
new_ROINumber = 1000
for Name in self.ROI_Names:
new_ROINumber -= 1
if not temp_color_list:
temp_color_list = copy.deepcopy(color_list)
color_int = np.random.randint(len(temp_color_list))
print('Writing data for ' + Name)
self.annotations = copy.deepcopy(
base_annotations[:, :, :,
int(self.ROI_Names.index(Name) + 1)])
self.annotations = self.annotations.astype('int')
make_new = 1
allow_slip_in = True
if (Name not in current_names
and allow_slip_in) or self.delete_previous_rois:
self.RS_struct.StructureSetROISequence.insert(
0,
copy.deepcopy(self.RS_struct.StructureSetROISequence[0]))
else:
print(
'Prediction ROI {} is already within RT structure'.format(
Name))
continue
self.RS_struct.StructureSetROISequence[
self.struct_index].ROINumber = new_ROINumber
self.RS_struct.StructureSetROISequence[self.struct_index].ReferencedFrameOfReferenceUID = \
self.ds.FrameOfReferenceUID
self.RS_struct.StructureSetROISequence[
self.struct_index].ROIName = Name
self.RS_struct.StructureSetROISequence[
self.struct_index].ROIVolume = 0
self.RS_struct.StructureSetROISequence[
self.struct_index].ROIGenerationAlgorithm = 'SEMIAUTOMATIC'
if make_new == 1:
self.RS_struct.RTROIObservationsSequence.insert(
0,
copy.deepcopy(self.RS_struct.RTROIObservationsSequence[0]))
if 'MaterialID' in self.RS_struct.RTROIObservationsSequence[
self.struct_index]:
del self.RS_struct.RTROIObservationsSequence[
self.struct_index].MaterialID
self.RS_struct.RTROIObservationsSequence[
self.struct_index].ObservationNumber = new_ROINumber
self.RS_struct.RTROIObservationsSequence[
self.struct_index].ReferencedROINumber = new_ROINumber
self.RS_struct.RTROIObservationsSequence[
self.struct_index].ROIObservationLabel = Name
self.RS_struct.RTROIObservationsSequence[
self.struct_index].RTROIInterpretedType = 'ORGAN'
if make_new == 1:
self.RS_struct.ROIContourSequence.insert(
0, copy.deepcopy(self.RS_struct.ROIContourSequence[0]))
self.RS_struct.ROIContourSequence[
self.struct_index].ReferencedROINumber = new_ROINumber
del self.RS_struct.ROIContourSequence[
self.struct_index].ContourSequence[1:]
self.RS_struct.ROIContourSequence[
self.struct_index].ROIDisplayColor = temp_color_list[color_int]
del temp_color_list[color_int]
thread_count = int(cpu_count() * 0.9 - 1)
# thread_count = 1
contour_dict = {}
q = Queue(maxsize=thread_count)
threads = []
kwargs = {
'image_size_rows': self.image_size_rows,
'image_size_cols': self.image_size_cols,
'slice_info': self.slice_info,
'PixelSize': self.PixelSize,
'mult1': self.mult1,
'mult2': self.mult2,
'ShiftZ': self.ShiftZ,
'mv': self.mv,
'shift_list': self.shift_list,
'ShiftRows': self.ShiftRows,
'ShiftCols': self.ShiftCols,
'contour_dict': contour_dict
}
A = [q, kwargs]
for worker in range(thread_count):
t = Thread(target=contour_worker, args=(A, ))
t.start()
threads.append(t)
contour_num = 0
if np.max(self.annotations
) > 0: # If we have an annotation, write it
image_locations = np.max(self.annotations, axis=(1, 2))
indexes = np.where(image_locations > 0)[0]
for index in indexes:
item = [self.annotations[index, ...], index]
q.put(item)
for i in range(thread_count):
q.put(None)
for t in threads:
t.join()
for i in contour_dict.keys():
for output in contour_dict[i]:
if contour_num > 0:
self.RS_struct.ROIContourSequence[
self.struct_index].ContourSequence.append(
copy.
deepcopy(self.RS_struct.ROIContourSequence[
self.struct_index].ContourSequence[0]))
self.RS_struct.ROIContourSequence[
self.struct_index].ContourSequence[
contour_num].ContourNumber = str(contour_num)
self.RS_struct.ROIContourSequence[
self.struct_index].ContourSequence[
contour_num].ContourImageSequence[
0].ReferencedSOPInstanceUID = self.SOPInstanceUIDs[
i]
self.RS_struct.ROIContourSequence[
self.struct_index].ContourSequence[
contour_num].ContourData = output
self.RS_struct.ROIContourSequence[
self.struct_index].ContourSequence[
contour_num].NumberofContourPoints = round(
len(output) / 3)
contour_num += 1
self.RS_struct.SOPInstanceUID += '.' + str(np.random.randint(999))
if self.template or self.delete_previous_rois:
for i in range(len(self.RS_struct.StructureSetROISequence),
len(self.ROI_Names), -1):
del self.RS_struct.StructureSetROISequence[-1]
for i in range(len(self.RS_struct.RTROIObservationsSequence),
len(self.ROI_Names), -1):
del self.RS_struct.RTROIObservationsSequence[-1]
for i in range(len(self.RS_struct.ROIContourSequence),
len(self.ROI_Names), -1):
del self.RS_struct.ROIContourSequence[-1]
for i in range(len(self.RS_struct.StructureSetROISequence)):
self.RS_struct.StructureSetROISequence[i].ROINumber = i + 1
self.RS_struct.RTROIObservationsSequence[
i].ReferencedROINumber = i + 1
self.RS_struct.ROIContourSequence[
i].ReferencedROINumber = i + 1
if not os.path.exists(self.output_dir):
os.makedirs(self.output_dir)
out_name = os.path.join(
self.output_dir, 'RS_MRN' + self.RS_struct.PatientID + '_' +
self.RS_struct.SeriesInstanceUID + '.dcm')
if os.path.exists(out_name):
out_name = os.path.join(
self.output_dir, 'RS_MRN' + self.RS_struct.PatientID + '_' +
self.RS_struct.SeriesInstanceUID + '1.dcm')
print('Writing out data...')
pydicom.write_file(out_name, self.RS_struct)
fid = open(os.path.join(self.output_dir, 'Completed.txt'), 'w+')
fid.close()
print('Finished!')
return None
def make_uid(self,dicom_path):
print(dicom_path)
if os.path.exists(os.path.join(dicom_path,'UID_val.txt')):
os.remove(os.path.join(dicom_path,'UID_val.txt'))
if os.path.exists(os.path.join(dicom_path,'MRN_val.txt')):
os.remove(os.path.join(dicom_path,'MRN_val.txt'))
#)
file_list = []
dirs = []
for root, dirs, file_list in os.walk(dicom_path):
break
files = [file for file in file_list if file.find('.dcm') != -1]
if not files:
files = file_list
uid_dic = {}
if not self.overlapping_images: # If we know there aren't overlapping image sets, move on
for file in files:
if file.find('{') != -1 or file.find('}') != -1:
os.remove(os.path.join(dicom_path,file))
continue
try:
ds = pydicom.read_file(os.path.join(dicom_path,file))
try:
Acquistion_time = str(ds.SeriesTime)
except:
Acquistion_time = ds.StudyTime
Acquistion_time = Acquistion_time.replace('.', '')
keys = uid_dic.keys()
if ds.Modality.find('RTSTRUCT') == 0:
RT_val = 1
Acquistion_time += str(RT_val)
if Acquistion_time not in keys:
uid_dic[Acquistion_time] = [file]
elif len(files) > self.min_size:
if Acquistion_time not in keys:
uid_dic[Acquistion_time] = [file]
break
except:
continue
else:
data_keys = [(0x08, 0x031), (0x020, 0x012)]
try:
data_all, ds = self.make_uid_and_del_dic(dicom_path,files, data_keys) # Acquisition Time
except:
return None
go_on = False
output = 2
while not go_on:
output -= 1
min_images = 51
while min_images > 0:
for key in data_all.keys():
if len(list(data_all[key][1].keys())) > output and \
len(files)/len(list(data_all[key][1].keys())) >= min_images: #len(list(data_all[key][1].keys())) <= 7 and
output = len(list(data_all[key][1].keys()))
del_dict, uid_dic = data_all[key]
go_on = True
if go_on:
break
min_images -= 10
for key in del_dict.keys():
if len(del_dict[key]) == 1: # If we have one image that wrong orientation, delete it
os.remove(os.path.join(dicom_path,del_dict[key][0]))
if (len(uid_dic.keys()) > 1) or (len(dirs) > 0 and 'query_file' not in dirs): # We have multiple scans.. need to make new folders for each and change the uids
i = -1
for key in uid_dic.keys():
i += 1
files_total = uid_dic[key]
for file_name in files_total:
if os.path.exists(os.path.join(dicom_path,key,file_name)):
os.remove(os.path.join(dicom_path,file_name)) # If it already exists, just move on
continue
if self.overlapping_liver and key.find('RT') != 0 and len(files_total) > self.min_size:
ds = pydicom.read_file(os.path.join(dicom_path,file_name))
ds.SeriesInstanceUID += '.' + str(i)
pydicom.write_file(os.path.join(dicom_path,file_name),ds)
if ds.Modality != 'US': #Raystation can't handle ultrasound
if len(files_total) > self.min_size or key.find('RT') == 0:
if not os.path.exists(os.path.join(dicom_path,key)):
os.mkdir(os.path.join(dicom_path ,key))
os.rename(os.path.join(dicom_path,file_name),os.path.join(dicom_path,key,file_name))
else:
os.remove(os.path.join(dicom_path,file_name))
if ds.Modality != 'US' and (len(files_total) > self.min_size or key.find('RT') == 0):
fid = open(os.path.join(dicom_path,key,'UID_val.txt'), 'w+')
if ds.Modality != 'RTSTRUCT':
fid.write(ds.SeriesInstanceUID)
else:
fid.write(ds.StudyInstanceUID)
fid.close()
fid = open(os.path.join(dicom_path,key,'MRN_val.txt'), 'w+')
fid.write(ds.PatientID)
fid.close()
fid = open(os.path.join(dicom_path,key,'prepped.txt'), 'w+')
fid.close()
elif len(uid_dic.keys()) > 0:
fid = open(os.path.join(dicom_path ,'UID_val.txt'), 'w+')
if ds.Modality != 'RTSTRUCT' and ds.Modality != 'RTPLAN' and ds.Modality != 'RTDOSE':
fid.write(ds.SeriesInstanceUID)
else:
fid.write(ds.StudyInstanceUID)
fid.close()
try:
MRN_val = ds.PatientID
except:
MRN_val = 0
fid = open(os.path.join(dicom_path, 'MRN_val.txt'), 'w+')
fid.write(MRN_val)
fid.close()
fid = open(os.path.join(dicom_path, 'prepped.txt'), 'w+')
fid.close()
return None
def adjust_machine_name_cli(args):
dicom_dataset = pydicom.read_file(args.input_file, force=True)
new_dicom_dataset = adjust_machine_name(
dicom_dataset, args.new_machine_name)
pydicom.write_file(args.output_file, new_dicom_dataset)
def adjust_RED_by_structure_name_cli(args):
dicom_dataset = pydicom.read_file(args.input_file, force=True)
new_dicom_dataset = adjust_RED_by_structure_name(dicom_dataset)
pydicom.write_file(args.output_file, new_dicom_dataset)
def save_as(self, filename, write_like_original=True):
"""Write the Dataset to `filename`.
Saving a Dataset requires that the Dataset.is_implicit_VR and
Dataset.is_little_endian attributes exist and are set appropriately. If
Dataset.file_meta.TransferSyntaxUID is present then it should be set to
a consistent value to ensure conformance.
Conformance with DICOM File Format
----------------------------------
If `write_like_original` is False, the Dataset will be stored in the
DICOM File Format in accordance with DICOM Standard Part 10 Section 7.
To do so requires that the `Dataset.file_meta` attribute exists and
contains a Dataset with the required (Type 1) File Meta Information
Group elements (see pydicom.filewriter.write_file and
pydicom.filewriter.write_file_meta_info for more information).
If `write_like_original` is True then the Dataset will be written as is
(after minimal validation checking) and may or may not contain all or
parts of the File Meta Information (and hence may or may not be
conformant with the DICOM File Format).
Parameters
----------
filename : str or file-like
Name of file or the file-like to write the new DICOM file to.
write_like_original : bool
If True (default), preserves the following information from
the Dataset (and may result in a non-conformant file):
- preamble -- if the original file has no preamble then none will
be written.
- file_meta -- if the original file was missing any required File
Meta Information Group elements then they will not be added or
written.
If (0002,0000) 'File Meta Information Group Length' is present
then it may have its value updated.
- seq.is_undefined_length -- if original had delimiters, write them
now too, instead of the more sensible length characters
- is_undefined_length_sequence_item -- for datasets that belong to
a sequence, write the undefined length delimiters if that is
what the original had.
If False, produces a file conformant with the DICOM File Format,
with explicit lengths for all elements.
See Also
--------
pydicom.filewriter.write_dataset
Write a DICOM Dataset to a file.
pydicom.filewriter.write_file_meta_info
Write the DICOM File Meta Information Group elements to a file.
pydicom.filewriter.write_file
Write a DICOM file from a FileDataset instance.
"""
# Ensure is_little_endian and is_implicit_VR exist
if not (hasattr(self, 'is_little_endian')
and hasattr(self, 'is_implicit_VR')):
raise AttributeError("'{0}.is_little_endian' and "
"'{0}.is_implicit_VR' must exist and be "
"set appropriately before "
"saving.".format(self.__class__.__name__))
pydicom.write_file(filename, self, write_like_original)
def dump_dicom(data, folder, spacing=(1, 1, 1),
origin=(0, 0, 0), intercept=0, slope=1):
""" Dump 3D scan in dicom format.
Parameters
----------
data : ndarray
3D numpy array containing ct scan's data.
folder : str
folder where dicom files will be dumped.
spacing : ArrayLike
ndarray of shape (3,) that contains spacing along z, y, x axes.
origin : ArrayLike
ndarray of shape (3,) that contains origin for z, y, x axes.
interception : float
interception value. Default is 0.
slope : float
slope value. Default is 1.
"""
spacing = np.array(spacing).reshape(-1)
origin = np.array(origin).reshape(-1)
if not os.path.exists(folder):
os.makedirs(folder)
num_slices = data.shape[0]
scan_id = np.random.randint(2 ** 16)
for i in range(num_slices):
slice_name = (
hex(scan_id + i)
.replace('x', '')
.upper()
.zfill(8)
)
filename = os.path.join(folder, slice_name)
pixel_array = (data[i, ...] - intercept) / slope
locZ, locY, locX = (float(origin[0] + spacing[0] * i),
float(origin[1]), float(origin[2]))
file_meta = Dataset()
file_meta.MediaStorageSOPClassUID = 'Secondary Capture Image Storage'
file_meta.MediaStorageSOPInstanceUID = (
hex(scan_id)
.replace('x', '')
.upper()
.zfill(8)
)
file_meta.ImplementationClassUID = slice_name
dataset = FileDataset(filename, {},
file_meta=file_meta,
preamble=b"\0"*128)
dataset.PixelData = pixel_array.astype(np.uint16).tostring()
dataset.RescaleSlope = slope
dataset.RescaleIntercept = intercept
dataset.ImagePositionPatient = MultiValue(type_constructor=float,
iterable=[locZ, locY, locX])
dataset.PixelSpacing = MultiValue(type_constructor=float,
iterable=[float(spacing[1]),
float(spacing[2])])
dataset.SliceThickness = float(spacing[0])
dataset.Modality = 'WSD'
dataset.Columns = pixel_array.shape[0]
dataset.Rows = pixel_array.shape[1]
dataset.file_meta.TransferSyntaxUID = pydicom.uid.ImplicitVRLittleEndian
dataset.PixelRepresentation = 1
dataset.BitsAllocated = 16
dataset.BitsStored = 16
dataset.SamplesPerPixel = 1
write_file(filename, dataset)