Add derived measures of similarity for ROIs Status Assigned Date of Last Update 2024/01/14 Person Assigned David Clunie Submitter Name IDC Submission Date 2023/12/15 Correction Number CP-2382 Log Summary: Add derived measures of similarity for ROIs Name of Standard PS3.16 Rationale for Correction: Similarity measures like Dice, Jaccard are needed, for example for comparing different segmentations for ROIs. Note that the measurement values of each ROI that contributes to the derived measurement, are specified by reference; i.e., they actually need to exist elsewhere (with the ROI specification via coordinates, segmentations, etc.) elsewhere in a TID 1500 invocation. Also need to add concepts for number of pixels or voxels, since those need to be recorded for each ROI and used as the source for the reference. [Ed.Note: Surprisingly, the count of the number of pixels or voxels is not defined as a distinct measure in IBSI, though often used as a parameter to compute other measures.] [Ed.Note: Do we need a different derived template between two regions (or sets of points) that does not depend on there being source measurements for each, the use of which is a bit contrived and doesn't actually encode the full information used in the computation, e.g., number of points in the intersection. Probably not, because reference from the derived template is to the Measurement Group container, and any source measurements are Type U so can be omitted anyway.] [Ed.Note; Do we need parametrized metrics like Weighted Hamming Similarity, Taversky Index, and if so, how should the weights be communicated? Eelbode T, Bertels J, Berman M, Vandermeulen D, Maes F, Bisschops R, et al. Optimization for Medical Image Segmentation: Theory and Practice When Evaluating With Dice Score or Jaccard Index. IEEE Transactions on Medical Imaging. 2020 Nov;39(11):3679–90. doi:10.1109/TMI.2020.3002417] [Ed.Note: What about geometric measures, e.g., Hausdorff distance, that are measures of similarity but not overlap per se., and what modified forms of them should be included +/- parameterized (e.g., "95%": 1. Huttenlocher DP, Klanderman GA, Rucklidge WJ. Comparing images using the Hausdorff distance. IEEE Trans Pattern Anal Mach Intell. 1993 Sep;15(9):850–63. doi:10.1109/34.232073 2. Dubuisson M-P, Jain AK. A modified Hausdorff distance for object matching. In: Proceedings of 12th International Conference on Pattern Recognition. 1994. p. 566–8 vol.1. Available from: https://ieeexplore.ieee.org/abstract/document/576361/references#references doi:10.1109/ICPR.1994.576361 3. Zhao C, Shi W, Deng Y. A new Hausdorff distance for image matching. Pattern Recognition Letters. 2005;26. Available from: https://doi.org/10.1016/j.patrec.2004.09.022 doi:10.1016/j.patrec.2004.09.022 4. Fedorov A, Billet E, Prastawa M, Gerig G, Radmanesh A, Warfield SK, et al. Evaluation of Brain MRI Alignment with the Robust Hausdorff Distance Measures. In: Bebis G, Boyle R, Parvin B, Koracin D, Remagnino P, Porikli F, et al., editors. International Symposium on Visual Computing - Advances in Visual Computing. Berlin, Heidelberg: Springer; 2008. p. 594–603. (Lecture Notes in Computer Science; vol. 5358). doi:10.1007/978-3-540-89639-5_57 5. Aydin OU, Taha AA, Hilbert A, Khalil AA, Galinovic I, Fiebach JB, et al. On the usage of average Hausdorff distance for segmentation performance assessment: hidden error when used for ranking. Eur Radiol Exp. 2021 Jan 21;5(1):4. doi:10.1186/s41747-020-00200-2 6. Aspert N, Santa-Cruz D, Ebrahimi T. MESH: measuring errors between surfaces using the Hausdorff distance. In: Proceedings IEEE International Conference on Multimedia and Expo. 2002. p. 705–8 vol.1. doi:10.1109/ICME.2002.1035879 7. Zhang J-W, Han G-Q, Wo Y. Image registration based on generalized and mean Hausdorff distances. In: 2005 International Conference on Machine Learning and Cybernetics. Guangzhou, China: IEEE; 2005. p. 5117-5121 Vol. 8. doi:10.1109/ICMLC.2005.1527845 ] Correction Wording: This Template encodes measurements for some metric, e.g., density, flow, or concentration, which are acquired over some defined sampling (e.g., over successive time slots in a dynamic contrast enhanced acquisition), or of some comparison between different ROIs.

Measurements Derived From Multiple ROI Measurements

	NL	Rel with Parent	VT	Concept Name	VM	Req Type	Condition	Value Set Constraint
1			NUM	D	1-n	M
1b	>	HAS CONCEPT MOD	INCLUDE	D	1	U
2	>	R-INFERRED FROM	INCLUDE	D	1-n	MC	XOR Row 3	$Measurement = ...
3	>	R-INFERRED FROM	INCLUDE	D	1-n	MC	XOR Row 2	$Measurement = ...
4	>	HAS PROPERTIES	INCLUDE	D	1	U		$RefAuthority = ...

Content Item Descriptions Row 1 Specifies the type of derived measurement reported, e.g., the mean of the individual ROI mean density values, or a similarity measure of the number of pixels or voxels in the ROIs. Note that the units may be different from the units in the ROI measurements Row 1b Describes the algorithm used to produced the derived measurement, as opposed to the referenced measurements from which this measurement was derived. Rows 2, 3 The measurement values of each ROI that contributes to the derived measurement, e.g., the mean density within an ROI, the number of pixels or voxels in an ROI. These are specified by reference, so as to not have to repeat the ROI information when it contributes to multiple derived measurements (e.g., if both mean and SD of ROI mean density values were specified).

Resources: HTML | FHIR JSON | FHIR XML | IHE SVS XML Keyword: MeasurementDerivedFromMultipleROIMeasurements FHIR Keyword: dicom-cid-7465-MeasurementDerivedFromMultipleROIMeasurements Type: Extensible Version: 20121101yyyymmdd UID: 1.2.840.10008.6.1.952

Measurement Derived From Multiple ROI Measurements

Coding Scheme Designator	Code Value	Code Meaning
Include
Include
Include

Resources: HTML | FHIR JSON | FHIR XML | IHE SVS XML Keyword: GenericIntensityAndSizeMeasurement FHIR Keyword: dicom-cid-7469-GenericIntensityAndSizeMeasurement Type: Extensible Version: 20141110yyyymmdd UID: 1.2.840.10008.6.1.1003

Generic Intensity and Size Measurement

Coding Scheme Designator	Code Value	Code Meaning
Include
Include
Include
Include
DCM	ddd003	Number of pixels
DCM	ddd004	Number of voxels

Add new context group to DICOM PS3.16

Resources: HTML | FHIR JSON | FHIR XML | IHE SVS XML Keyword: SimilarityMeasure FHIR Keyword: dicom-cid-ccc1-SimilarityMeasure Type: Extensible Version: yyyymmdd UID: 1.2.840.10008.6.1.uuu1

Similarity Measure

Coding Scheme Designator	Code Value	Code Meaning
DCM	ddd001	Dice Similarity Coefficient
DCM	ddd002	Jaccard Index

Add new concepts to DICOM PS3.16

DICOM Controlled Terminology Definitions (Coding Scheme Designator "DCM" Coding Scheme Version "01")

Code Value	Code Meaning	Definition	Notes
ddd001	Dice Similarity Coefficient	A measure of the pair-wise overlap of two sets defined as two times the intersection of the two samples, divided by the sum of the two samples. Also known as the Sørensen–Dice coefficient. The samples may be points representing pixels or voxels in a geometric region.	1. Dice LR. Measures of the Amount of Ecologic Association Between Species. Ecol. 1945 Jul;26(3):297–302. doi:10.2307/1932409 2. Taha AA, Hanbury A. Metrics for evaluating 3D medical image segmentation: analysis, selection, and tool. BMC Med Imag. 2015 Dec;15(1):29. doi:10.1186/s12880-015-0068-x
ddd002	Jaccard Index	A measure of the pair-wise overlap of two samples defined as the intersection of the two regions, divided by the union of the two samples. Also known as the Intersection Over Union (IoU). The samples may be points representing pixels or voxels in a geometric region.	1. Jaccard P. The Distribution of the Flora of the Alpine Zone. New Phytol. 1912 Feb;11(2):37–50. doi:10.1111/j.1469-8137.1912.tb05611.x 2. Taha AA, Hanbury A. Metrics for evaluating 3D medical image segmentation: analysis, selection, and tool. BMC Med Imag. 2015 Dec;15(1):29. doi:10.1186/s12880-015-0068-x
ddd003	Number of pixels	A count of the number of pixels in an image or a region of an image.
ddd004	Number of voxels	A count of the number of voxels in a volume or a region of a volume.

For reference unchanged DICOM PS3.16

...

Planar ROI Measurements and Qualitative Evaluations

	NL	Rel with Parent	VT	Concept Name	VM	Req Type	Condition	Value Set Constraint
1			CONTAINER	EV (125007, DCM, "Measurement Group")	1	M
...
5	>	CONTAINS	SCOORD	EV (111030, DCM, "Image Region")	1	MC	XOR Rows 7, 7b, 8b	GRAPHIC TYPE = not {MULTIPOINT}
6	>>	SELECTED FROM	IMAGE		1	M
7	>	CONTAINS	IMAGE	EV (121214, DCM, "Referenced Segmentation Frame")	1	MC	XOR Rows 5, 7b	Reference shall be to a Segmentation Image, with a single value specified in Referenced Segment Number (0062,000B). For references to tiled Segmentation Images, one or more values shall be specified in Referenced Frame Number (0008,1160), unless all frames in the referenced Segmentation Image are selected and there is only a single Segment, in which case Referenced Frame Number (0008,1160) will be absent. The referenced tiles shall all be in the same plane. For references to non-tiled Segmentation Images, a single value shall be specified in Referenced Frame Number (0008,1160), unless there is only one frame in the referenced Segmentation Image, in which case Referenced Frame Number (0008,1160) will be absent.
7b	>	CONTAINS	SCOORD3D	EV (111030, DCM, "Image Region")	1	MC	XOR Rows 5, 7, 8b	GRAPHIC TYPE = not {MULTIPOINT, POLYLINE or ELLIPSOID}
8	>	CONTAINS	IMAGE	EV (121233, DCM, "Source image for segmentation")	1	MC	IFF Row 7
8b	>	CONTAINS	COMPOSITE	EV (130488, DCM, "Region in Space")	1	MC	XOR Rows 5, 7, 7b	Reference shall be to an Instance of the RT Structure Set Storage SOP Class.
8c	>>	HAS PROPERTIES	TEXT	EV (130489, DCM, "Referenced Region of Interest Identifier")	1	M		Shall be the value of ROI Number (3006,0022) within the single referenced Item of Structure Set ROI Sequence (3006,0020) of the referenced Instance of the RT Structure Set Storage SOP Class.
...
11	>	CONTAINS	INCLUDE	D	1	U		$Measurement = $Measurement ...
...

...

Volumetric ROI Measurements and Qualitative Evaluations

	NL	Rel with Parent	VT	Concept Name	VM	Req Type	Condition	Value Set Constraint
1			CONTAINER	EV (125007, DCM, "Measurement Group")	1	M
...
5	>	CONTAINS	SCOORD	EV (111030, DCM, "Image Region")	1-n	MC	XOR Rows 7, 10, 12b	GRAPHIC TYPE = not {MULTIPOINT}
6	>>	SELECTED FROM	IMAGE		1	M
7	>	CONTAINS	IMAGE	EV (121191, DCM, "Referenced Segment")	1	MC	XOR Rows 5, 10, 12b	Reference shall be to a Segmentation Image or Surface Segmentation object, with a single value specified in Referenced Segment Number
10	>	CONTAINS	SCOORD3D	EV (121231, DCM, "Volume Surface")	1-n	MC	XOR Rows 5, 7, 12b	If one item, GRAPHIC TYPE = {ELLIPSOID or POINT} If more than one item, GRAPHIC TYPE = {POLYGON or ELLIPSE}
11	>	CONTAINS	IMAGE	EV (121233, DCM, "Source image for segmentation")	1-n	MC	XOR Row 12 and IFF (Row 7 or Row 10)
12	>	CONTAINS	UIDREF	EV (121232, DCM, "Source series for segmentation")	1	MC	XOR Row 11 and IFF (Row 7 or Row 10)
12b	>	CONTAINS	COMPOSITE	EV (130488, DCM, "Region in Space")	1	MC	XOR Rows 5, 7, 10	Reference shall be to an Instance of the RT Structure Set Storage SOP Class.
12c	>>	HAS PROPERTIES	TEXT	EV (130489, DCM, "Referenced Region of Interest Identifier")	1	M		Shall be the value of ROI Number (3006,0022) within the single referenced Item of Structure Set ROI Sequence (3006,0020) of the referenced Instance of the RT Structure Set Storage SOP Class.
...
15	>	CONTAINS	INCLUDE	D	1	U		$Measurement = $Measurement ...
...					1

ROI Measurements

	NL	Rel with Parent	VT	Concept Name	VM	Req Type	Condition	Value Set Constraint
...
5			NUM	$Measurement	1-n	M		UNITS = $Units
...

...

Measurement Report

	NL	Rel with Parent	VT	Concept Name	VM	Req Type	Condition	Value Set Constraint
...
7	>>	CONTAINS	INCLUDE	D	1-n	U		$Measurement = B ...
8	>>	CONTAINS	INCLUDE	D	1-n	U		$Measurement = B ...
9	>>	CONTAINS	INCLUDE	D	1-n	U		$Measurement = B ...
10	>	CONTAINS	CONTAINER	EV (126011, DCM, "Derived Imaging Measurements")	1	MC	IF Row 6 and 12 are absent
...

...

...

Population Statistical Descriptor

Coding Scheme Designator	Code Value	Code Meaning	SNOMED-RT ID	UMLS Concept Unique ID
SCT	371889001	95th Percentile Value of population	R-00337	C1299358
...

Sample Statistical Descriptor

Coding Scheme Designator	Code Value	Code Meaning
DCM	121415	Percentile Ranking of measurement
DCM	121416	Z-Score of measurement

Resources: HTML | FHIR JSON | FHIR XML | IHE SVS XML Keyword: QuantitativeImageFeature FHIR Keyword: dicom-cid-218-QuantitativeImageFeature Type: Extensible Version: 20200920 UID: 1.2.840.10008.6.1.1269

Quantitative Image Feature

Coding Scheme Designator	Code Value	Code Meaning
Include
Include
Include
Include