def _generate_waveform_sequence(self):
if self._is_12_lead_ecg:
maximum_waveform_sequences = 5 # PS3.3 A.34.3.4.3 Waveform Sequence
maximum_waveform_samples = 16384 # PS3.3 A.34.3.4.5 Number of Waveform Samples
else:
maximum_waveform_sequences = 4 # PS3.3 A.34.4.4.2 Waveform Sequence
maximum_waveform_samples = 2 ** 32 - 1 # NumberOfWaveformSamples's VR is `UL`.
data_file_len = fileutil.file_size(self._file)
pack_size = struct.calcsize(self._format)
data_file_total_samples = data_file_len // pack_size
saved_samples = maximum_waveform_sequences * maximum_waveform_samples
if data_file_total_samples > saved_samples:
warn_msg = (
'File "{}" is too big. File size: {}, pack size: {}, format string: {}, '
"total samples: {}, saved samples: {}, saved size: {}.".format(
fileutil.file_name(self._file),
data_file_len,
pack_size,
self._format,
data_file_total_samples,
saved_samples,
saved_samples * pack_size,
)
)
if not _frozen: # Fix issue #7
warnings.warn(warn_msg)
logger.warn(warn_msg)
data_file_total_samples = saved_samples
waveform_seq = dicom.sequence.Sequence()
data_unpacker = unpacker.unpack_data_from_file(self._file, self._format)
target_fmt = "<{}".format("h" * self._channels)
adjusted_data = map(lambda v: map(self._adjust_callback, v), data_unpacker)
while data_file_total_samples > 0:
seq_item = dicom.dataset.Dataset()
seq_item.WaveformOriginality = "ORIGINAL" # Type 1
seq_item.NumberOfWaveformChannels = self._channels # Type 1.
if self._is_12_lead_ecg:
assert 1 <= seq_item.NumberOfWaveformChannels <= 13 # PS3.3 A.34.3.4.4
else:
assert 1 <= seq_item.NumberOfWaveformChannels <= 24 # PS3.3 A.34.4.4.3 Number of Waveform Channels
if data_file_total_samples >= maximum_waveform_samples:
seq_item.NumberOfWaveformSamples = maximum_waveform_samples # Type 1. UL.
else:
seq_item.NumberOfWaveformSamples = data_file_total_samples # Type 1. UL.
data_file_total_samples -= seq_item.NumberOfWaveformSamples
assert 200 <= self._sampling_frequency <= 1000 # DICOM PS3.3-2015a A.34.3.4.6
seq_item.SamplingFrequency = "{:d}".format(self._sampling_frequency) # Type 1. DS.
seq_item.ChannelDefinitionSequence = self._generate_channel_definition_sequence() # Type 1.
seq_item.WaveformBitsAllocated = (
16
) # Type 1. PS3.3 C.10.9.1.5 Waveform Bits Allocated and Waveform Sample Interpretation
seq_item.WaveformSampleInterpretation = "SS" # Type 1. PS3.3 A.34.3.4.8 Waveform Sample Interpretation
# The VR of `Waveform Padding Value` may be OB or OW, so:
# seq_item.WaveformPaddingValue = b'\x00\x00'
# will not work, instead:
seq_item.add_new((0x5400, 0x100A), "OW", b"\x00\x80") # Type 1C. OB or OW.
data = bytearray()
for i, d in zip(range(seq_item.NumberOfWaveformSamples), adjusted_data):
data.extend(struct.pack(target_fmt, *d))
seq_item.add_new((0x5400, 0x1010), "OW", bytes(data)) # WaveformData. Type 1. OB or OW.
waveform_seq.append(seq_item)
return waveform_seq
if not dest:
dest = fileutil.replace_ext(src, ".dcm")
data_set = DCMECGDataset(
src,
"<{}".format("d" * 12),
500,
12,
("I", "II", "III", "aVR", "aVL", "aVF", "V1", "V2", "V3", "V4", "V5", "V6"),
adjust_callback=lambda v: int(v * 1000 / 6),
)
data_set.save_as(dest)
if __name__ == "__main__":
u = unpacker.unpack_data_from_file(r"E:\data\1\20110607153002.dat", "<d")
print(max(u))
u = unpacker.unpack_data_from_file(r"E:\data\1\20110607153002.dat", "<d")
print(min(u))
ecg_to_dcm(r"E:\data\1\20110607153002.dat")
ecg_to_dcm(r"d:\20120503152310.dat")
# References:
# DICOM 2015a PS3.5 7.4 Data Element Type
# DICOM 2015a PS3.5 8.3 Waveform Data and Related Data Elements
# DICOM 2015a PS3.6 6 Registry of DICOM Data Elements
# [DICOM Waveform Generator](http://libir.tmu.edu.tw/bitstream/987654321/21661/1/B09.pdf)
# [Mandatory Tags for DICOM Instance](http://stackoverflow.com/questions/6608535/mandatory-tags-for-dicom-instance)
# [Questions regarding the DICOM file](http://fixunix.com/dicom/50267-questions-regarding-dicom-file.html)
# [Dicom: What's the point of SOPInstanceUID tag?](http://stackoverflow.com/questions/1434918/dicom-whats-the-point-of-sopinstanceuid-tag)
def main():
prog = os.path.splitext(os.path.basename(sys.argv[0]))[0]
parser = argparse.ArgumentParser(
prog=prog,
description='Extract data from binary file.',
)
parser.add_argument('-i',
'--input-file',
required=True,
help='Input file path')
parser.add_argument('--skip-bytes',
type=int,
default=0,
help='Skip how many bytes at the begin of the file')
parser.add_argument('-f',
'--format',
required=True,
help='Format string to describe the data')
parser.add_argument('--endian',
choices=('little', 'ieee-le', 'le', 'l', 'big',
'ieee-be', 'be', 'b', 'native', 'n'),
default='native',
help='Endianess of binary data')
parser.add_argument('-F', '--factor', help='Factor')
parser.add_argument('-o', '--output-file', help='Output file path')
parser.add_argument('-d',
'--delimiter',
help='Column delimiter of output file')
parser.add_argument('-O',
'--overwrite',
action='store_true',
help='Overwrite existing file')
args = parser.parse_args()
if not os.path.isfile(args.input_file):
sys.stderr.write('Input file \'{0}\' doesn\'t exist!\n'.format(
args.input_file))
sys.exit(1)
if args.skip_bytes < 0:
sys.stderr.write('Invalid skip bytes: {0}.'.format(args.skip_bytes))
sys.exit(1)
# A factor string can be:
# 1. A numeric literal: "0.001"
# 2. A math expression: "1.0/1000.0"
# 3. A math expression with unknown value: "x+300"
# For 1 and 2, we can get the value of factor via `parse_factor`.
# For 3, we first replace `x` with `1.0`, and then we call `math_eval`.
# If it failed, then it means the factor string is not a valid one.
# Otherwise, we evaluate it later.
if args.factor:
factor = parse_factor(args.factor)
if factor is None:
try:
_ = math_eval_v1.math_eval(args.factor, x=1)
fixed_factor = False
except:
sys.stderr.write(
'Invalid factor: \'{0}\' is neither a valid numeric literal nor a math expression.\n'
.format(args.factor))
sys.exit(1)
else:
fixed_factor = True
if args.output_file and os.path.isfile(
args.output_file) and not args.overwrite:
sys.stderr.write('Output file \'{0}\' already exists! '
'Please use a different file path or specify '
'\'--overwrite\' on the command line.\n'.format(
args.output_file))
sys.exit(1)
if not args.delimiter:
if args.output_file:
if is_csv_file(args.output_file):
args.delimiter = ','
elif is_tsv_file(args.output_file):
args.delimiter = '\t'
else:
args.delimiter = ' '
else:
args.delimiter = ' '
if args.output_file:
output_fp = open(args.output_file, 'w', encoding='ascii')
else:
output_fp = sys.stdout
for line in unpacker.unpack_data_from_file(args.input_file, args.format,
args.skip_bytes):
if not args.factor:
output_line = line
elif fixed_factor:
output_line = (col * factor for col in line)
else:
output_line = (math_eval_v1.math_eval(args.factor, x=col)
for col in line)
output_fp.write(args.delimiter.join(
(repr(col) for col in output_line)))
output_fp.write('\n')
if args.output_file:
output_fp.close()
def _generate_waveform_sequence(self):
if self._is_12_lead_ecg:
maximum_waveform_sequences = 5 # PS3.3 A.34.3.4.3 Waveform Sequence
maximum_waveform_samples = 16384 # PS3.3 A.34.3.4.5 Number of Waveform Samples
else:
maximum_waveform_sequences = 4 # PS3.3 A.34.4.4.2 Waveform Sequence
maximum_waveform_samples = 2**32 - 1 # NumberOfWaveformSamples's VR is `UL`.
data_file_len = fileutil.file_size(self._file)
pack_size = struct.calcsize(self._format)
data_file_total_samples = data_file_len // pack_size
saved_samples = maximum_waveform_sequences * maximum_waveform_samples
if data_file_total_samples > saved_samples:
warn_msg = 'File "{}" is too big. File size: {}, pack size: {}, format string: {}, ' \
'total samples: {}, saved samples: {}, saved size: {}.'.format(
fileutil.file_name(self._file), data_file_len, pack_size, self._format,
data_file_total_samples, saved_samples, saved_samples * pack_size)
if not _frozen: # Fix issue #7
warnings.warn(warn_msg)
logger.warn(warn_msg)
data_file_total_samples = saved_samples
waveform_seq = dicom.sequence.Sequence()
data_unpacker = unpacker.unpack_data_from_file(self._file,
self._format)
target_fmt = '<{}'.format('h' * self._channels)
adjusted_data = map(lambda v: map(self._adjust_callback, v),
data_unpacker)
while data_file_total_samples > 0:
seq_item = dicom.dataset.Dataset()
seq_item.WaveformOriginality = 'ORIGINAL' # Type 1
seq_item.NumberOfWaveformChannels = self._channels # Type 1.
if self._is_12_lead_ecg:
assert 1 <= seq_item.NumberOfWaveformChannels <= 13 # PS3.3 A.34.3.4.4
else:
assert 1 <= seq_item.NumberOfWaveformChannels <= 24 # PS3.3 A.34.4.4.3 Number of Waveform Channels
if data_file_total_samples >= maximum_waveform_samples:
seq_item.NumberOfWaveformSamples = maximum_waveform_samples # Type 1. UL.
else:
seq_item.NumberOfWaveformSamples = data_file_total_samples # Type 1. UL.
data_file_total_samples -= seq_item.NumberOfWaveformSamples
assert 200 <= self._sampling_frequency <= 1000 # DICOM PS3.3-2015a A.34.3.4.6
seq_item.SamplingFrequency = '{:d}'.format(
self._sampling_frequency) # Type 1. DS.
seq_item.ChannelDefinitionSequence = self._generate_channel_definition_sequence(
) # Type 1.
seq_item.WaveformBitsAllocated = 16 # Type 1. PS3.3 C.10.9.1.5 Waveform Bits Allocated and Waveform Sample Interpretation
seq_item.WaveformSampleInterpretation = 'SS' # Type 1. PS3.3 A.34.3.4.8 Waveform Sample Interpretation
# The VR of `Waveform Padding Value` may be OB or OW, so:
#seq_item.WaveformPaddingValue = b'\x00\x00'
# will not work, instead:
seq_item.add_new((0x5400, 0x100A), 'OW',
b'\x00\x80') # Type 1C. OB or OW.
data = bytearray()
for i, d in zip(range(seq_item.NumberOfWaveformSamples),
adjusted_data):
data.extend(struct.pack(target_fmt, *d))
seq_item.add_new((0x5400, 0x1010), 'OW',
bytes(data)) # WaveformData. Type 1. OB or OW.
waveform_seq.append(seq_item)
return waveform_seq
# Lead I, II, III, aVR, aVL, aVF, V1, V2, V3, V4, V5, V6, I, II, III, ...
# The unit of signals collected by the cardiac conduction is: 0.4V/(2^15).
if not dest:
dest = fileutil.replace_ext(src, '.dcm')
data_set = DCMECGDataset(src,
'<{}'.format('d' * 12),
500,
12, ('I', 'II', 'III', 'aVR', 'aVL', 'aVF', 'V1',
'V2', 'V3', 'V4', 'V5', 'V6'),
adjust_callback=lambda v: int(v * 1000 / 6))
data_set.save_as(dest)
if __name__ == '__main__':
u = unpacker.unpack_data_from_file(r'E:\data\1\20110607153002.dat', '<d')
print(max(u))
u = unpacker.unpack_data_from_file(r'E:\data\1\20110607153002.dat', '<d')
print(min(u))
ecg_to_dcm(r'E:\data\1\20110607153002.dat')
ecg_to_dcm(r'd:\20120503152310.dat')
# References:
# DICOM 2015a PS3.5 7.4 Data Element Type
# DICOM 2015a PS3.5 8.3 Waveform Data and Related Data Elements
# DICOM 2015a PS3.6 6 Registry of DICOM Data Elements
# [DICOM Waveform Generator](http://libir.tmu.edu.tw/bitstream/987654321/21661/1/B09.pdf)
# [Mandatory Tags for DICOM Instance](http://stackoverflow.com/questions/6608535/mandatory-tags-for-dicom-instance)
# [Questions regarding the DICOM file](http://fixunix.com/dicom/50267-questions-regarding-dicom-file.html)
# [Dicom: What's the point of SOPInstanceUID tag?](http://stackoverflow.com/questions/1434918/dicom-whats-the-point-of-sopinstanceuid-tag)
def _generate_waveform_sequence(self):
if self._is_12_lead_ecg:
maximum_waveform_sequences = 5 # PS3.3 A.34.3.4.3 Waveform Sequence
maximum_waveform_samples = 16384 # PS3.3 A.34.3.4.5 Number of Waveform Samples
else:
maximum_waveform_sequences = 4 # PS3.3 A.34.4.4.2 Waveform Sequence
maximum_waveform_samples = 2 ** 32 - 1 # NumberOfWaveformSamples's VR is `UL`.
data_file_len = fileutil.file_size(self._file)
pack_size = struct.calcsize(self._format)
data_file_total_samples = data_file_len // pack_size
saved_samples = maximum_waveform_sequences * maximum_waveform_samples
if data_file_total_samples > saved_samples:
data_file_total_samples = saved_samples
self._channel_labels = list(self._channel_labels)
channel_cnt = len(self._channel_labels)
waveform_seq = dicom.sequence.Sequence()
data_unpacker = unpacker.unpack_data_from_file(self._file, self._format)
target_fmt = '<{}'.format('h' * channel_cnt)
if self._adjust_callback and hasattr(self._adjust_callback, '__call__'):
adjusted_data = map(lambda v: map(self._adjust_callback, v), data_unpacker)
else:
adjusted_data = data_unpacker
if self._is_12_lead_ecg:
assert 1 <= channel_cnt <= 13 # PS3.3 A.34.3.4.4
else:
assert 1 <= channel_cnt <= 24 # PS3.3 A.34.4.4.3 Number of Waveform Channels
assert 200 <= self._sampling_frequency <= 1000 # DICOM PS3.3-2015a A.34.3.4.6
while data_file_total_samples > 0:
seq_item = dicom.dataset.Dataset()
seq_item.WaveformOriginality = 'ORIGINAL' # Type 1
seq_item.NumberOfWaveformChannels = channel_cnt # Type 1.
if data_file_total_samples >= maximum_waveform_samples:
seq_item.NumberOfWaveformSamples = maximum_waveform_samples # Type 1. UL.
else:
seq_item.NumberOfWaveformSamples = data_file_total_samples # Type 1. UL.
data_file_total_samples -= seq_item.NumberOfWaveformSamples
seq_item.SamplingFrequency = '{:d}'.format(self._sampling_frequency) # Type 1. DS.
seq_item.ChannelDefinitionSequence = self._generate_channel_definition_sequence() # Type 1.
seq_item.WaveformBitsAllocated = 16 # Type 1. PS3.3 C.10.9.1.5 Waveform Bits Allocated and Waveform Sample Interpretation
seq_item.WaveformSampleInterpretation = 'SS' # Type 1. PS3.3 A.34.3.4.8 Waveform Sample Interpretation
# The VR of `Waveform Padding Value` may be OB or OW, so:
#seq_item.WaveformPaddingValue = b'\x00\x00'
# will not work, instead:
seq_item.add_new((0x5400, 0x100A), 'OW', b'\x00\x80') # Type 1C. OB or OW.
data = bytearray()
for i, d in zip(range(seq_item.NumberOfWaveformSamples), adjusted_data):
data.extend(struct.pack(target_fmt, *d))
seq_item.add_new((0x5400, 0x1010), 'OW', bytes(data)) # WaveformData. Type 1. OB or OW.
waveform_seq.append(seq_item)
return waveform_seq
