def desensitize(ds,
sensitive_tags=None,
img_ds=True,
img_ds_area=(None, None, None, None)):
""" 脱敏函数
:param ds: Dataset 待脱敏Dataset对象
:param sensitive_tags: list 敏感标签
:param img_ds: bool 是否进行图像脱敏
:param img_ds_area: tuple 图像脱敏区域(x_min, x_max, y_min, y_max) None 表示不限
:return: Dataset 已脱敏对象
"""
if not sensitive_tags:
sensitive_tags = [
(0x0008, 0x0080),
(0x0008, 0x0081),
(0x0008, 0x1040),
(0x0010, 0x0010),
(0x0033, 0x1013),
(0x0019, 0x161b),
]
# 标签脱敏
for tag in sensitive_tags:
tag = TupleTag(tag)
logger.debug('Desensitizing %s tag.' % tag)
ds.pop(tag, None)
# 图像脱敏
if ds[0x0008, 0x0060].value in ('CR', 'DR') and img_ds:
logger.debug('Desensitizing pixels data')
pixels = np.fromstring(ds.PixelData, dtype=np.uint16)
pixels = pixels.reshape([ds.Rows, ds.Columns])
pixels = fill_zero_color(pixels, img_ds_area)
ds.PixelData = pixels.tostring()
return ds
def testNestedPrivateSQ(self):
"""Can successfully read a private SQ which contains additional SQ's....."""
# From issue 113. When a private SQ of undefined length is used, the
# sequence is read in and the length of the SQ is determined upon
# identification of the SQ termination sequence. When using nested
# Sequences, the first termination sequence encountered actually
# belongs to the nested Sequence not the parent, therefore the
# remainder of the file is not read in properly
ds = read_file(nested_priv_SQ_name)
# Make sure that the entire dataset was read in
pixel_data_tag = TupleTag((0x7fe0, 0x10))
self.assertTrue(
pixel_data_tag in ds,
"Entire dataset was not parsed properly. PixelData is not present")
# Check that the DataElement is indeed a Sequence
tag = TupleTag((0x01, 0x01))
seq0 = ds[tag]
self.assertEqual(seq0.VR, 'SQ',
"First level sequence not parsed properly")
# Now verify the presence of the nested private SQ
seq1 = seq0[0][tag]
self.assertEqual(seq1.VR, 'SQ',
"Second level sequence not parsed properly")
# Now make sure the values that are parsed are correct
got = seq1[0][tag].value
expected = b'Double Nested SQ'
self.assertEqual(got, expected,
"Expected a value of %s, got %s'" % (expected, got))
got = seq0[0][0x01, 0x02].value
expected = b'Nested SQ'
self.assertEqual(got, expected,
"Expected a value of %s, got %s'" % (expected, got))
def desensitize(ds,
sensitive_tags=None,
img_ds=True,
img_ds_area=(None, None, None, None)):
""" 脱敏函数
:param ds: Dataset 待脱敏Dataset对象
:param sensitive_tags: list 敏感标签
:param img_ds: bool 是否进行图像脱敏
:param img_ds_area: tuple 图像脱敏区域(x_min, x_max, y_min, y_max) None 表示不限
:return: Dataset 已脱敏对象
"""
if not sensitive_tags:
sensitive_tags = SENSITIVE_TAGS
# 标签脱敏
for tag in sensitive_tags:
tag = TupleTag(tag)
logger.debug('Desensitizing %s tag.' % tag)
ds.pop(tag, None)
return ds
def absorb_delimiter_item(fp, is_little_endian, delimiter):
"""Read (and ignore) undefined length sequence or item terminators."""
if is_little_endian:
struct_format = "<HHL"
else:
struct_format = ">HHL"
group, elem, length = unpack(struct_format, fp.read(8))
tag = TupleTag((group, elem))
if tag != delimiter:
msg = "Did not find expected delimiter '%s'" % dictionary_description(
delimiter)
msg += ", instead found %s at file position 0x%x" % (str(tag),
fp.tell() - 8)
logger.warn(msg)
fp.seek(fp.tell() - 8)
return
logger.debug("%04x: Found Delimiter '%s'",
fp.tell() - 8, dictionary_description(delimiter))
if length == 0:
logger.debug("%04x: Read 0 bytes after delimiter", fp.tell() - 4)
else:
logger.debug("%04x: Expected 0x00000000 after delimiter, found 0x%x",
fp.tell() - 4, length)
def convert_tag(byte_string, is_little_endian, offset=0):
if is_little_endian:
struct_format = "<HH"
else:
struct_format = ">HH"
return TupleTag(unpack(struct_format, byte_string[offset:offset + 4]))
def testGoodTuple(self):
"""Tags can be constructed with two-tuple of 2-byte integers."""
tag = TupleTag((0xFFFF, 0xFFee))
tag = TupleTag((0x300a, 0x00b0))
self.assertEqual(tag.group, 0x300a, "Expected tag.group 0x300a, got %r" % tag.group)
def data_element_generator(fp, is_implicit_VR, is_little_endian,
stop_when=None, defer_size=None, encoding=default_encoding):
"""Create a generator to efficiently return the raw data elements
Returns (VR, length, raw_bytes, value_tell, is_little_endian),
where:
VR -- None if implicit VR, otherwise the VR read from the file
length -- the length as in the DICOM data element (could be
DICOM "undefined length" 0xffffffffL),
value_bytes -- the raw bytes from the DICOM file
(not parsed into python types)
is_little_endian -- True if transfer syntax is little endian; else False
"""
# Summary of DICOM standard PS3.5-2008 chapter 7:
# If Implicit VR, data element is:
# tag, 4-byte length, value.
# The 4-byte length can be FFFFFFFF (undefined length)*
# If Explicit VR:
# if OB, OW, OF, SQ, UN, or UT:
# tag, VR, 2-bytes reserved (both zero), 4-byte length, value
# For all but UT, the length can be FFFFFFFF (undefined length)*
# else: (any other VR)
# tag, VR, (2 byte length), value
# * for undefined length, a Sequence Delimitation Item marks the end
# of the Value Field.
# Note, except for the special_VRs, both impl and expl VR use 8 bytes;
# the special VRs follow the 8 bytes with a 4-byte length
# With a generator, state is stored, so we can break down
# into the individual cases, and not have to check them again for each
# data element
if is_little_endian:
endian_chr = "<"
else:
endian_chr = ">"
if is_implicit_VR:
element_struct = Struct(endian_chr + "HHL")
else: # Explicit VR
# tag, VR, 2-byte length (or 0 if special VRs)
element_struct = Struct(endian_chr + "HH2sH")
extra_length_struct = Struct(endian_chr + "L") # for special VRs
extra_length_unpack = extra_length_struct.unpack # for lookup speed
# Make local variables so have faster lookup
fp_read = fp.read
fp_tell = fp.tell
logger_debug = logger.debug
debugging = dicom.debugging
element_struct_unpack = element_struct.unpack
while True:
# Read tag, VR, length, get ready to read value
bytes_read = fp_read(8)
if len(bytes_read) < 8:
raise StopIteration # at end of file
if debugging:
debug_msg = "{0:08x}: {1}".format(fp.tell() - 8,
bytes2hex(bytes_read))
if is_implicit_VR:
# must reset VR each time; could have set last iteration (e.g. SQ)
VR = None
group, elem, length = element_struct_unpack(bytes_read)
else: # explicit VR
group, elem, VR, length = element_struct_unpack(bytes_read)
if in_py3:
VR = VR.decode(default_encoding)
if VR in extra_length_VRs:
bytes_read = fp_read(4)
length = extra_length_unpack(bytes_read)[0]
if debugging:
debug_msg += " " + bytes2hex(bytes_read)
if debugging:
debug_msg = "%-47s (%04x, %04x)" % (debug_msg, group, elem)
if not is_implicit_VR:
debug_msg += " %s " % VR
if length != 0xFFFFFFFFL:
debug_msg += "Length: %d" % length
else:
debug_msg += "Length: Undefined length (FFFFFFFF)"
logger_debug(debug_msg)
# Positioned to read the value, but may not want to -- check stop_when
value_tell = fp_tell()
tag = TupleTag((group, elem))
if stop_when is not None:
# XXX VR may be None here!! Should stop_when just take tag?
if stop_when(tag, VR, length):
if debugging:
logger_debug("Reading ended by stop_when callback. "
"Rewinding to start of data element.")
rewind_length = 8
if not is_implicit_VR and VR in extra_length_VRs:
rewind_length += 4
fp.seek(value_tell - rewind_length)
raise StopIteration
# Reading the value
# First case (most common): reading a value with a defined length
if length != 0xFFFFFFFFL:
if defer_size is not None and length > defer_size:
# Flag as deferred by setting value to None, and skip bytes
value = None
logger_debug("Defer size exceeded."
"Skipping forward to next data element.")
fp.seek(fp_tell() + length)
else:
value = fp_read(length)
if debugging:
dotdot = " "
if length > 12:
dotdot = "..."
logger_debug("%08x: %-34s %s %r %s" % (value_tell,
bytes2hex(value[:12]), dotdot, value[:12], dotdot))
# If the tag is (0008,0005) Specific Character Set, then store it
if tag == (0x08, 0x05):
from dicom.values import convert_string
encoding = convert_string(value, is_little_endian, encoding=default_encoding)
# Store the encoding value in the generator for use with future elements (SQs)
encoding = convert_encodings(encoding)
yield RawDataElement(tag, VR, length, value, value_tell,
is_implicit_VR, is_little_endian)
# Second case: undefined length - must seek to delimiter,
# unless is SQ type, in which case is easier to parse it, because
# undefined length SQs and items of undefined lengths can be nested
# and it would be error-prone to read to the correct outer delimiter
else:
# Try to look up type to see if is a SQ
# if private tag, won't be able to look it up in dictionary,
# in which case just ignore it and read the bytes unless it is
# identified as a Sequence
if VR is None:
try:
VR = dictionaryVR(tag)
except KeyError:
# Look ahead to see if it consists of items and is thus a SQ
next_tag = TupleTag(unpack(endian_chr + "HH", fp_read(4)))
# Rewind the file
fp.seek(fp_tell() - 4)
if next_tag == ItemTag:
VR = 'SQ'
if VR == 'SQ':
if debugging:
msg = "{0:08x}: Reading/parsing undefined length sequence"
logger_debug(msg.format(fp_tell()))
seq = read_sequence(fp, is_implicit_VR,
is_little_endian, length, encoding)
yield DataElement(tag, VR, seq, value_tell,
is_undefined_length=True)
else:
delimiter = SequenceDelimiterTag
if debugging:
logger_debug("Reading undefined length data element")
value = read_undefined_length_value(fp, is_little_endian,
delimiter, defer_size)
# If the tag is (0008,0005) Specific Character Set, then store it
if tag == (0x08, 0x05):
from dicom.values import convert_string
encoding = convert_string(value, is_little_endian, encoding=default_encoding)
# Store the encoding value in the generator for use with future elements (SQs)
encoding = convert_encodings(encoding)
yield RawDataElement(tag, VR, length, value, value_tell,
is_implicit_VR, is_little_endian)
def data_element_generator(fp, is_implicit_VR, is_little_endian, stop_when=None,
defer_size=None):
"""Create a generator to efficiently return the raw data elements
Specifically, returns (VR, length, raw_bytes, value_tell, is_little_endian),
where:
VR -- None if implicit VR, otherwise the VR read from the file
length -- the length as in the DICOM data element (could be
DICOM "undefined length" 0xffffffffL),
value_bytes -- the raw bytes from the DICOM file (not parsed into python types)
is_little_endian -- True if transfer syntax is little endian; else False
"""
# Summary of DICOM standard PS3.5-2008 chapter 7:
# If Implicit VR, data element is:
# tag, 4-byte length, value.
# The 4-byte length can be FFFFFFFF (undefined length)*
# If Explicit VR:
# if OB, OW, OF, SQ, UN, or UT:
# tag, VR, 2-bytes reserved (both zero), 4-byte length, value
# For all but UT, the length can be FFFFFFFF (undefined length)*
# else: (any other VR)
# tag, VR, (2 byte length), value
# * for undefined length, a Sequence Delimitation Item marks the end
# of the Value Field.
# Note, except for the special_VRs, both impl and expl VR use 8 bytes;
# the special VRs follow the 8 bytes with a 4-byte length
# With a generator, state is stored, so we can break down
# into the individual cases, and not have to check them again for each
# data element
# Make local variables so have faster lookup
fp_read = fp.read
fp_tell = fp.tell
logger_debug = logger.debug
debugging = dicom.debugging
if is_little_endian:
endian_chr = "<"
else:
endian_chr = ">"
if is_implicit_VR:
unpack_format = endian_chr + "HHL" # XXX in python >=2.5, can do struct.Struct to save time
else: # Explicit VR
unpack_format = endian_chr + "HH2sH" # tag, VR, 2-byte length (or 0 if special VRs)
extra_length_format = endian_chr + "L" # for special VRs
while True:
# Read tag, VR, length, get ready to read value
bytes_read = fp_read(8)
if len(bytes_read) < 8:
raise StopIteration # at end of file
if debugging: debug_msg = "%08x: %s" % (fp.tell()-8, bytes2hex(bytes_read))
if is_implicit_VR:
VR = None # must reset each time -- may have looked up on last iteration (e.g. SQ)
group, elem, length = unpack(unpack_format, bytes_read)
else: # explicit VR
group, elem, VR, length = unpack(unpack_format, bytes_read)
if VR in ('OB','OW','OF','SQ','UN', 'UT'):
bytes_read = fp_read(4)
length = unpack(extra_length_format, bytes_read)[0]
if debugging: debug_msg += " " + bytes2hex(bytes_read)
if debugging:
debug_msg = "%-47s (%04x, %04x)" % (debug_msg, group, elem)
if not is_implicit_VR: debug_msg += " %s " % VR
if length != 0xFFFFFFFFL:
debug_msg += "Length: %d" % length
else:
debug_msg += "Length: Undefined length (FFFFFFFF)"
logger_debug(debug_msg)
# Now are positioned to read the value, but may not want to -- check stop_when
value_tell = fp_tell()
# logger.debug("%08x: start of value of length %d" % (value_tell, length))
tag = TupleTag((group, elem))
if stop_when is not None:
if stop_when(tag, VR, length): # XXX VR may be None here!! Should stop_when just take tag?
if debugging:
logger_debug("Reading ended by stop_when callback. Rewinding to start of data element.")
rewind_length = 8
if not is_implicit_VR and VR in ('OB','OW','OF','SQ','UN', 'UT'):
rewind_length += 4
fp.seek(value_tell-rewind_length)
raise StopIteration
# Reading the value
# First case (most common): reading a value with a defined length
if length != 0xFFFFFFFFL:
if defer_size is not None and length > defer_size:
# Flag as deferred read by setting value to None, and skip bytes
value = None
logger_debug("Defer size exceeded. Skipping forward to next data element.")
fp.seek(fp_tell()+length)
else:
value = fp_read(length)
if debugging:
dotdot = " "
if length > 12:
dotdot = "..."
logger_debug("%08x: %-34s %s %r %s" % (value_tell,
bytes2hex(value[:12]), dotdot, value[:12], dotdot))
yield RawDataElement(tag, VR, length, value, value_tell,
is_implicit_VR, is_little_endian)
# Second case: undefined length - must seek to delimiter,
# ... unless is SQ type, in which case is easier to parse it, because
# undefined length SQs and items of undefined lengths can be nested
# and it would be error-prone to read to the correct outer delimiter
else:
# Try to look up type to see if is a SQ
# if private tag, won't be able to look it up in dictionary,
# in which case just ignore it and read the bytes
if VR is None:
try:
VR = dictionaryVR(tag)
except KeyError:
pass
if VR == 'SQ':
if debugging:
logger_debug("%08x: Reading and parsing undefined length sequence"
% fp_tell())
seq = read_sequence(fp, is_implicit_VR, is_little_endian, length)
yield DataElement(tag, VR, seq, value_tell, is_undefined_length=True)
else:
delimiter = SequenceDelimiterTag
if debugging:
logger_debug("Reading undefined length data element")
value = read_undefined_length_value(fp, is_little_endian, delimiter,
defer_size)
yield RawDataElement(tag, VR, length, value, value_tell,
is_implicit_VR, is_little_endian)
