The DcmMeta extension is a complete but concise representation of the meta data in a series of source DICOM files.
The primary goals are:
The meta data is extracted from each DICOM input into a set of key/value pairs. Each non-private DICOM element uses the standard DICOM keyword as its key. Values are generally left unchanged (except for ‘DS’ and ‘IS’ value representations which are converted from strings to float and integer numbers respectively).
Translators are used to convert private elements into sets of key/value pairs. These are then added to the standard DICOM meta data with the translator name (followed by a dot) prepended to each of the keys it generates.
Private DICOM elements without translators are ignored by default, but this can be overridden. Any element with a value representation of ‘OW’ or ‘OB’ is ignored if it contains non-ASCII characters.
The meta data from individual input files is summarized over the dimensions of the Nifti file. Most of the meta data will be constant across all of the input files. Other meta data will be constant across each time/vector sample, or repeating for the slices in each time/vector sample. We summarize the meta data into one or more dictionaries as follows.
There will always be a dictionary ‘global’ with two nested dictionaries inside, ‘const’ and ‘slices’. The meta data that is constant across all input files get stored under the ‘const’ dictionary. The meta data that varies across all slices will be stored under slices, where each value is a list of values (one for each slice).
If there is a time dimension there will also be ‘time’ dictionary containing two nested dictionaries, ‘samples’ and ‘slices’. Meta data that is constant across a time sample will be stored in the ‘samples’ dictionary with each being value a list of values (one for each time sample). Values that repeat across the slices in a time sample (a single volume) will be stored in the ‘slices’ dictionary with each value being a list of values (one for each slice in a time point).
If there is a vector dimension there will be a ‘vector’ dictionary, handled in the same manner as the ‘time’ dictionary.
The dictionaries of summarized meta data are encoded with JSON. A small amount of “meta meta” data that describes the DcmMeta extension is also included. This includes the affine (‘dcmmeta_affine’), shape (‘dcmmeta_shape’), any reorientation transform (‘dcmmeta_reorient_transform’), and the slice dimension (‘dcmmeta_slice_dim’) of the data described by the meta data. A version number for the DcmMeta extension (‘dcmmeta_version’) is also included.
The affine, shape, and slice dimension are used to determine if varying meta data is still valid. For example, if the image affine no longer matches the meta data affine (i.e. the image has been coregistered) then we cannot directly match the per-slice meta data values to slices of the data array.
The reorientation transform can be used to update directional meta data to match the image orientation. This transform encodes any reordering of the voxel data that occurred during conversion. If the image affine does not match the meta data affine, then an additional transformation needs to be done after applying the reorientation transform (from the meta data space to the image space).
Below is an example DcmMeta extension created from a data set with two slices and three time points (each with a different EchoTime). The meta data has been abridged (the ”...” line) for clarity.
{
"global": {
"const": {
"SpecificCharacterSet": "ISO_IR 100",
"ImageType": [
"ORIGINAL",
"PRIMARY",
"M",
"ND"
],
"StudyTime": 69244.484,
"SeriesTime": 71405.562,
"Modality": "MR",
"Manufacturer": "SIEMENS",
"SeriesDescription": "2D 16Echo qT2",
"ManufacturerModelName": "TrioTim",
"ScanningSequence": "SE",
"SequenceVariant": "SP",
"ScanOptions": "SAT1",
"MRAcquisitionType": "2D",
"SequenceName": "se2d16",
"AngioFlag": "N",
"SliceThickness": 7.0,
"RepetitionTime": 3000.0,
"NumberOfAverages": 1.0,
"ImagingFrequency": 123.250392,
"ImagedNucleus": "1H",
"MagneticFieldStrength": 3.0,
"SpacingBetweenSlices": 10.5,
"NumberOfPhaseEncodingSteps": 96,
"EchoTrainLength": 1,
"PercentSampling": 50.0,
"PercentPhaseFieldOfView": 100.0,
"PixelBandwidth": 420.0,
"SoftwareVersions": "syngo MR B17",
"ProtocolName": "2D 16Echo qT2",
"TransmitCoilName": "TxRx_Head",
"AcquisitionMatrix": [
0,
192,
96,
0
],
"InPlanePhaseEncodingDirection": "ROW",
"FlipAngle": 180.0,
"VariableFlipAngleFlag": "N",
"SAR": 0.11299714843984,
"dBdt": 0.0,
"StudyID": "1",
"SeriesNumber": 3,
"AcquisitionNumber": 1,
"ImageOrientationPatient": [
1.0,
-2.051034e-10,
0.0,
2.051034e-10,
1.0,
1.98754e-11
],
"SamplesPerPixel": 1,
"PhotometricInterpretation": "MONOCHROME2",
"Rows": 192,
"Columns": 192,
"PixelSpacing": [
0.66666668653488,
0.66666668653488
],
"BitsAllocated": 16,
"BitsStored": 12,
"HighBit": 11,
"PixelRepresentation": 0,
"SmallestImagePixelValue": 0,
"WindowCenterWidthExplanation": "Algo1",
"PerformedProcedureStepStartTime": 69244.546,
"CsaImage.EchoLinePosition": 48,
"CsaImage.UsedChannelMask": 1,
"CsaImage.MeasuredFourierLines": 0,
"CsaImage.SequenceMask": 134217728,
"CsaImage.RFSWDDataType": "predicted",
"CsaImage.RealDwellTime": 6200,
"CsaImage.ImaCoilString": "C:HE",
"CsaImage.EchoColumnPosition": 96,
"CsaImage.PhaseEncodingDirectionPositive": 1,
"CsaImage.GSWDDataType": "predicted",
"CsaImage.SliceMeasurementDuration": 286145.0,
"CsaImage.MultistepIndex": 0,
"CsaImage.ImaRelTablePosition": [
0,
0,
0
],
"CsaImage.NonPlanarImage": 0,
"CsaImage.EchoPartitionPosition": 32,
"CsaImage.AcquisitionMatrixText": "96*192s",
"CsaImage.ImaAbsTablePosition": [
0,
0,
-1630
],
"CsaSeries.TalesReferencePower": 334.36266914,
"CsaSeries.Operation_mode_flag": 2,
"CsaSeries.dBdt_thresh": 0.0,
"CsaSeries.ProtocolChangeHistory": 0,
"CsaSeries.GradientDelayTime": [
12.0,
14.0,
10.0
],
"CsaSeries.SARMostCriticalAspect": [
3.2,
1.84627729,
0.0
],
"CsaSeries.B1rms": [
7.07106781,
1.59132133
],
"CsaSeries.RelTablePosition": [
0,
0,
0
],
"CsaSeries.NumberOfPrescans": 0,
"CsaSeries.dBdt_limit": 0.0,
"CsaSeries.Stim_lim": [
45.73709869,
27.64929962,
31.94370079
],
"CsaSeries.PatReinPattern": "1;FFS;45.36;10.87;3;0;2;866892320",
"CsaSeries.B1rmsSupervision": "NO",
"CsaSeries.ReadoutGradientAmplitude": 0.0,
"CsaSeries.MrProtocolVersion": 21710006,
"CsaSeries.RFSWDMostCriticalAspect": "Head",
"CsaSeries.SequenceFileOwner": "SIEMENS",
"CsaSeries.GradientMode": "Fast",
"CsaSeries.SliceArrayConcatenations": 1,
"CsaSeries.FlowCompensation": "No",
"CsaSeries.TransmitterCalibration": 128.29875,
"CsaSeries.Isocentered": 0,
"CsaSeries.AbsTablePosition": -1630,
"CsaSeries.ReadoutOS": 2.0,
"CsaSeries.dBdt_max": 0.0,
"CsaSeries.RFSWDOperationMode": 0,
"CsaSeries.SelectionGradientAmplitude": 0.0,
"CsaSeries.PhaseGradientAmplitude": 0.0,
"CsaSeries.RfWatchdogMask": 0,
"CsaSeries.CoilForGradient2": "AS092",
"CsaSeries.Stim_mon_mode": 2,
"CsaSeries.CoilId": [
255,
196,
238,
238,
238,
238,
238,
238,
238,
238,
238
],
"CsaSeries.Stim_max_ges_norm_online": 0.62600064,
"CsaSeries.CoilString": "C:HE",
"CsaSeries.CoilForGradient": "void",
"CsaSeries.TablePositionOrigin": [
0,
0,
-1630
],
"CsaSeries.MiscSequenceParam": [
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
93,
0,
0,
0,
0,
0,
0
],
"CsaSeries.LongModelName": "NUMARIS/4",
"CsaSeries.Stim_faktor": 1.0,
"CsaSeries.SW_korr_faktor": 1.0,
"CsaSeries.Sed": [
1000000.0,
156.13387238,
156.13387238
],
"CsaSeries.PositivePCSDirections": "+LPH",
"CsaSeries.SliceResolution": 1.0,
"CsaSeries.Stim_max_online": [
0.22781265,
17.30016327,
0.5990392
],
"CsaSeries.t_puls_max": 0.0,
"CsaSeries.MrPhoenixProtocol.ulVersion": 21710006,
"CsaSeries.MrPhoenixProtocol.tSequenceFileName": "%SiemensSeq%\\se_mc",
"CsaSeries.MrPhoenixProtocol.tProtocolName": "2D 16Echo qT2",
...
"CsaSeries.MrPhoenixProtocol.sAsl.ulMode": 1,
"CsaSeries.MrPhoenixProtocol.ucAutoAlignInit": 1
},
"slices": {
"InstanceCreationTime": [
71405.671,
71405.562,
71405.671,
71405.578,
71405.671,
71405.578
],
"AcquisitionTime": [
71118.2425,
71116.7375,
71118.2625,
71116.7575,
71118.2825,
71116.7775
],
"ContentTime": [
71405.671,
71405.562,
71405.671,
71405.578,
71405.671,
71405.578
],
"InstanceNumber": [
1,
2,
7,
8,
13,
14
],
"LargestImagePixelValue": [
2772,
2828,
2077,
2085,
1470,
1397
],
"WindowCenter": [
1585.0,
1513.0,
1495.0,
1455.0,
1100.0,
1084.0
],
"WindowWidth": [
3191.0,
3212.0,
2750.0,
2731.0,
2120.0,
2073.0
],
"CsaImage.TimeAfterStart": [
1.505,
0.0,
1.525,
0.02,
1.545,
0.04
],
"CsaImage.ICE_Dims": [
"1_1_1_1_1_1_1_4_1_1_1_1_490",
"1_1_1_1_1_1_1_1_1_1_2_1_490",
"1_2_1_1_1_1_1_4_1_1_1_1_490",
"1_2_1_1_1_1_1_1_1_1_2_1_490",
"1_3_1_1_1_1_1_4_1_1_1_1_490",
"1_3_1_1_1_1_1_1_1_1_2_1_490"
]
}
},
"time": {
"samples": {
"EchoTime": [
20.0,
40.0,
60.0
],
"EchoNumbers": [
1,
2,
3
]
},
"slices": {
"ImagePositionPatient": [
[
-64.000001919919,
-118.13729284881,
-33.707626344045
],
[
-64.000001919919,
-118.13729284881,
-23.207628251394
]
],
"SliceLocation": [
-33.707626341697,
-23.207628249046
],
"CsaImage.ProtocolSliceNumber": [
0,
1
],
"CsaImage.SlicePosition_PCS": [
[
-64.00000192,
-118.13729285,
-33.70762634
],
[
-64.00000192,
-118.13729285,
-23.20762825
]
]
}
},
"dcmmeta_shape": [
192,
192,
2,
3
],
"dcmmeta_affine": [
[
-0.6666666865348816,
1.3673560894655878e-10,
0.0,
64.0
],
[
1.3673560894655878e-10,
0.6666666865348816,
0.0,
-9.196043968200684
],
[
0.0,
-1.325026720289113e-11,
10.499998092651367,
-33.70762634277344
],
[
0.0,
0.0,
0.0,
1.0
]
],
"dcmmeta_reorient_transform": [
[
-0.0,
-1.0,
-0.0,
191.0
],
[
1.0,
0.0,
0.0,
0.0
],
[
0.0,
0.0,
1.0,
0.0
],
[
0.0,
0.0,
0.0,
1.0
]
],
"dcmmeta_slice_dim": 2,
"dcmmeta_version": 0.6
}