Virtual Monochromatic Image - DICOM



WG21: Multi-Energy CT Images (Sup XXX)Definitions, Clarifications and Questions for new SupplementMulti-Energy CT Imaging: CT multi-energy imaging techniques (including scanning, reconstruction, processing) where the scanner utilizes multiple energies from the X-Ray beam spectrum, as opposite to the conventional CT imaging, when a single (accumulated) X-Ray spectrum is used.(add IOD strategy column); Parametric Map sup 172 (WG 16/David) overlap of concepts – to have a look at; try to promote Enh. CTTermDefinitionVirtual Monochromatic Image (VMI)HU-based CT image that is equivalent to a monochromatic (of a specific keV value) X-Ray beam. VMI image is seen as standardization of CT images to keV. (may be need a range rather than single value)Material Characterization ImageImages created using material decomposition model, which represents the image data by two or more base materialsMaterial Quantification ImageMaterial Characterization Image that allows quantitative evaluation of the material in a physical scale. Pixel values can be in HU, or equivalent material concentration (e.g., mg/ml), or effective atomic number.Material-Specific ImageMaterial Quantification Image presenting a physical scale of specific materialMaterial-Subtracted ImageMaterial Quantification Image with one or more base materials subtracted, while pixel values may have been corrected for missing fractionMaterial Classification ImageMaterial Characterization Image that allows qualitative evaluation or discrete classification of the materials (Labeling)Material-Highlighted ImageHU-based Material Classification Image where CT values of the target material are artificially highlighted higher than in standard CT imageMaterial-Suppressed ImageHU-based Material Classification Image where CT values of the target material are lower or nulled compared to the conventional CT imageMulti-Material Labeled ImageMaterial Classification Image where each pixel contains values indirectly describing identified material in this pixelBinary LabelingPixel value is a flag whether a certain material is present (mask)Discrete LabelingPixel value is an index corresponding to one or more materials from a list/vector of the known materialsProbability MapPixel value is a number representing the probability that this pixel is classified as one or more of the multiple defined materialsMaterial Visualization ImageMaterial Characterization Image that visualizes material content in a certain user-convenient way (color maps, color overlays, blending, etc.)Questions for WG21:Are these types of images always DERIVED or are they ORIGINAL? ORIGINAL Image: “images whose pixel values are based on original or source data”DERIVED Image: “an image whose pixel values have been derived in some manner from the pixel value of one or more other images”One approach is to distinguish the location/system where such images were created. For instance, ORIGINAL if created within Reconstruction System (Acquisition Workplace); The reason behind this is, that the result images are created without any possibility to change/manipulate the source (regardless if it is a volume, slices, raw, etc. this is source). These images are created the first time with all necessary information and the origin is singular, distinct and unique. DERIVED if created on a Workstation and is based on, e.g., ORIGINAL images with different KV.Another approach is to distinguish the type of input data used to generate such images, regardless of the location/system. For vendors that use image-based multi-energy processing out of published images, it should be DERIVED. For projection-based processing, it is more likely to be DERIVED if VMI is generated as linear combination of the two base material images. However, if the VMI is generated directly from the raw data, it is more likely to be ORIGINAL.The other reason for defining the VMI images as ORIGINAL is that, in theory, such images can be the only clinical images generated by the scanner, e.g. to suppress beam hardening artifact in “normal” image. In this case it would be quite unusual that the study does not contain ORIGINAL images at all.“DERIVED” images can be potentially deleted by the user assuming there are ORIGINAL images as well.Our recommendation: to use ORIGINAL unless there is a specific case where the images are typically created only outside the modality. Some PACS/Apps may refuse to process properly the “DERIVED” images.PRIMARY or SECONDARY?If the image is ORIGINAL, it always comes with PRIMARY. However, if the VMI image is DERIVED (in the cases described above), would it make sense to define them as DERIVED\PRIMARY? This is a quite unusual combination today.Shall we have some Series-level attribute in order to query?E.g., kind of “Multi-Energy Series Type” and thus to require such images to be in a separate series? Is there any clinical/interoperability value in this separation? Need new IOD to have Series-level attribute Type 1 or 2 or informative if type 3.Defining “insignificant” or “don’t care” pixel value?For some types of images this may be useful, e.g., when material characteristics of the pixel cannot be accurately identifiedVirtual Monochromatic ImageDefinition: HU-based CT image that is equivalent to a monochromatic (of a specific keV value) X-Ray beam.Monochromatic images are seen as standardization of images to keV. This allows reproducible HU-values for material (e.g. bone) and a reproducible image impression (quality), better iodine representation and better metal artifact reduction.Dicom Image Type: DERIVED\PRIMARY\...Essential information to be included in the images: keV - Monochromatic equivalent of this image in keVImages Based on Material DecompositionThese images provides better distinguishing between different materials based on multi-energy information. Alpha blending technique can be used for displaying such images. (this is about visualization)Essential Information to be captured for these images:KV Specific techniques used for material decomposition, e.g.2 material decomposition3 material decompositionSinogram based or image basedParameters accessible by the user (need examples)There are two categories of images based on material decomposition: Material decomposition Quantification – meaning is a quantitative evaluation with respect to concentration (two material decomposition, bone and soft tissue e.g. bone mineral density; three material decomposition like VNC or VNCa). Image pixel values are either in HU or in equivalent material concentration (e.g., mg/ml)Base for material decomposition images are HU (x mg per ml is the equivalent concentration of the material decomposition). Required information: Name(s) of material(s) shown in the image Names of all base materials Units (e.g. “mg/ml”, “HU”, “mg iodine/ml”) Only for “HU”: Specification of Spectrum (“80kV”, “50keV”, “80/Sn140kV (0.4)”, etc.)Enhanced and suppressed – put here?Material decomposition (each image contains concentration of respective base material) or material subtraction (other base material(s) have been subtracted but voxel values have been corrected for missing volume fraction)Use Cases for Material Quantification:Bone mineral densityVirtual-Non-Contrast or Virtual Non-CalciumIodine MapA question: Does Effective Z belong to this category?Material Labeling (alternative term: “Material Classification”)– meaning is a qualitative evaluation or classification of materials. Basis for enhanced and suppressed images (for distinguishing the different materials). E.g. used in coloring of specific material. Used in material suppressed image or used as material map (material values like VRT ranges). On top of the image different visualization techniques can be applied. Primarily Qualitative (classification of material). Can involve multiple materials in one pixel.Different flavours of characterization need to define a characteristic (binary, percent,… ) and the type of tissue (muscle, …)Subcategories: Material highlighting: image shows CT-values; CT-values of target material are higher than in standard CT-image Material suppression: image shows CT-values; CT-values of target material are lower than in standard CT-imageBinary Material Labeling: image shows numbers corresponding to identified materials (need vector with material names) Probability map: image shows quantitative number that can be interpreted as probability that a voxel contains one material or another (need name of displayed quantity, units [none/%], material name corresponding to high values and material name corresponding to low values) In all cases to record: Name of the target material (highlited or suppressed) Specification of Spectrum (“80kV”, “50keV”, “80/Sn140kV (0.4)”, etc.)Use Cases for Material Labeling:Highlighting of materials that are not well separated from others. E.g.: Increasing the HU value of tendons compared to muscle based on dual energy information. Better VRT visualizationColoring of kidney stones and bone removalColoring of goutMaterial-Enhanced ImageDefinition: a material-specific image created with an intention to have only one specific material present in the image voxels/pixels.Examples: Iodine-Enhanced (aka “Iodine Map”), Calcium-Enhanced, etc. – to make it fit the sub-categories aboveWe stopped hereWhile the above definition does not require only a single pure material to be present, there is a strong assumption that all other materials are suppressed at some level.(The term “Enhanced” can be confused with contrast media enhancement – probably need a better name, such as “Significant”)Such image is usually created using a material decomposition into two (or more) base materials. In theory, such decomposition is accurate if only these two base materials are present/scanned. The equivalent base material concentrations for other materials depend to some extent on the attenuation curve of the material. Various methods of material separation exist, some are in the projection domain, other are in the image domain. The methods to create such images are vendor-specific, thus vendor-specific term can be provided.Material-Enhanced images can be presented either in terms of their partial density (concentration) in each pixel (e.g., mg/cc of iodine, or mg/mm3 calcium, etc.), or in terms of HU, since theoretically the attenuation coefficient (HU) is linear with the density. Shall we support/differentiate these two types? E.g.:2.1 Material-Enhanced Image - Concentration2.2 Material-Enhanced Image - CT or HUWhile one single material is enhanced (intentionally), other materials in various parts of the image can be suppressed at different levels, thus the HU is not an accurate measurement unit. Shall we leave HU with proper disclaimer, or to define a new unit for Rescale Type (e.g., “Modified HU”)?Dicom Image Type: DERIVED\PRIMARY\...Essential information to be included in the images: Material that is enhancedConcentration/Density Units (via Rescale Type or Real-World Mapping) for “Concentration” typeStandard Decomposition Method name (e.g., “Sinogram-Based”, “Image-Based”)Vendor-specific material decomposition method name/labelParameters used in the material decomposition procedure (at least the ones publicly available and adjustable by users)The other base materials used in the material decompositionMaterial-Suppressed ImageDefinition: an HU-based CT image created with an intention to remove one (or more?) materials from the image voxels/pixels.Examples: Iodine-Suppressed (aka “Virtual Non-Contrast”), Calcium-Suppressed (used to enable seeing only the blood-vessels in CT Angiography), etc.While the above definition does not require that this material is 100% removed, there is a strong assumption that it is significantly suppressed, while all other materials remain essentially intact.Dicom Image Type: DERIVED\PRIMARY\...Essential information to be included in the images: Material(s) that are suppressedStandard Decomposition Method name (e.g., Sinogram-Based, Image-Based)Vendor-specific material decomposition method nameParameters used in the material decomposition procedure (at least the ones publicly available and adjustable by users)The other base materials used in the material decompositionNote: “Poly-Enhanced” or “Poly-Suppressed” images are currently out of scope.Effective Atomic Number ImageDefinition: a CT image where each pixel represents Effective Atomic Number (aka “Effective Z”) of that pixel. Such image is similar to the Material-Enhanced Image (Concentration type), but the unit is Atomic Number (“EAN”?) rather than concentration (mg/cc).There is also a need for color EAN images as it is convenient sometimes to display such images as color maps. This can be done in the following ways:Retaining the original EAN pixel values and using a color lookup table in a rendering pipeline to map from EAN to RGB color. This approach allows measurements but the color display would not be supported by many products Creating color Sec. Capture, but in this case some measurements (such as ROI average or standard deviation) are not possibleOther?Dicom Image Type: DERIVED\PRIMARY\...5. Fused imagesOverlay of material specific image with gray scale HU image6. Blending (mixed) imagesLinear or non-linear blending of the images in each pared to a single energy image? Linear blending covered by DICOM. Non-linear blending currently not covered.. May be blending presentation states can resolve the issue.7. Electron density imagesElectron density can be based on HU units. Relative to water. Used in Radiation therapy for treatment planning. May be a new type of image.Open Questions: In clinical cases the question is often is: which soft tissue density would be seen if a foreign material was not there (e.g., “Virtual Non-Contrast” Image) so the pixel value is corrected for the space that was occupied by the other material. Standard material decomposition would show a decrease in the soft tissue density if there is more of the other base material.? The above definition for Material-Suppressed Image does not cover this case. How can these cases be distinguished? Are there other similar cases?One common use of base material decomposition is for iodine and water. But iodine cannot be produced with variable density and the attenuation of pure iodine. Iodine is too high to be measured reliably (partial volume effects, scattered radiation, etc.). How can iodine base material images be evaluated in a simple way (reproducibility)? For some materials (especially contrast agents) the density of the main atom (iodine) is known, but the exact chemical composition may be unknown (though reproducible).? The resulting density image would show the iodine density under the assumption that the iodine is contained in the contrast agent and diluted e.g. in water. How can this case be clearly defined?The same applies for all other base material with high (effective) atomic number. Seems like it can be (or will be) diluted to some degree, this is not a significant issue – is this correct?How can an unambiguous and reproducible definition/decomposition look like? Are there other procedures possible to standardize such decomposition?There seems to be two variations of Material-Suppressed images. One is when the “suppressed” pixel is replaced by the value that corresponds the HU value as if this material was removed, such as in Virtual Non-Contrast image. Another possibility is to replace the “suppressed” pixel by some “zero” (air?) value. Does this 2nd approach make sense, and, if yes, we need to provide different names for them, e.g., “Material-Suppressed” vs. “MaterailMaterial-Removed”.How do we encode a masking mechanism for MaterialDecompositionImages?Definition of a neutral value (not a significant value, something like 20.000) might be useful. Need feedback from WG 06?ToDo’s:Definition of essential attributes that shall be provided to the user, related to the used image creation technique. ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download