Glossary

DICOM = Digital Imaging and Communications in Medicine CTCA = CT coronary angiogram
PACS = Picture Archiving and Communication Storage SR = Structured Reports KOS = Key Object Selection3 SEG = Segmentation RT = Radio Therapy RTSS = Radio Therapy Structure Set RWVM = Real World Value Maps IOD = Information Object Definitions NIfTI = Neuroimaging Informatics Technology Initiative DFWG = Data Format Working Group VTK = Visualization ToolKit 3D Slicer = open source software platform for medical image informatics, image processing, and three-dimensional visualization MPR = Multi Planar Reconstruction CCTA = Coronary Computed Tomography Angiography MIP = Maximum Intensity Projection STL = STereoLithography ITK = Insight Toolkit

Summaries from

• DICOM presentation
• http://dicomiseasy.blogspot.com/2011/10/introduction-to-dicom-chapter-1.html

DICOM

Introduction

What is DICOM:

• Software Integration Standard in Medical Imaging

Annotation

• Annotation = Note/explanation for text/diagram
• Label = Classifying name for person/thing
• Markup = tags indicating relationship between texts
• GLCM = Gray Level Co-occurance Matrix
• TCIA = The Cancer Imaging Archive

Use cases

• Patient/Case
• Imaging Study
• Series/Acquisition
• Image
• Frame (pixel data array)
• Region of Interest ( ROI )
• Single point ( label every voxel/pixel )

Use cases - Qualitative / Quantitative

• Qualitative

  • categorical - shape = round, square

• Ordinal

  • roundness - scale of 1 to 5

• Quantitative

  • morphology
  • size ( length, volume )
  • quantity
  • background staining intensity
  • attenuation coefficient
  • Numerical features
  • GLCM entropy
  • fractal dimension
  • Transformed variants
  • registered, resampled ec
  • Units
  • Absolute / Relative
  • Derivation
  • mean / max
  • Method
  • model, fitting, sampling, binning

Use cases - Concept representation

• Single concept

  • “rectangular”

• Name Value pair

  • “shape” = “rectangular”

• Coded / Text

  • PN:12345 v. name: “Jane Doe”

Use cases - Localization Representation

• Patient /Case - Patient / Specimen ID
• Study / Series - UUID
• Frame - UID + frame offset / number

• ROI

  • contours: image in 2D or space in 3D
  • temporal coordinates = frame #, relative/absolute time
  • segmentations = bitmap / partial occupancy / probability

• Single point

  • parametric maps (bitmap, scaled integer)

• Transformations

  • rigid / affine, non-rigid (spline,…)

How DICOM fits in

DICOM as input

• Every radiology image is initially in DICOM

  • single files, or formatted to single multi-frame DICOM file

• DICOM headers metadata

  • considered “annotations”
  • example: Series Description = “Heart MRI axial”
  • multi-site data DICOM attributes
  • Keep vendor private data elements, might be useful later

DICOM as output

• Sharing annotations as DICOM

• Pros:

  • re-use of existing clinical imaging annotations systems
  • use off-the-shelves DICOM archives : TCIA
  • opened by OTS DICOM tools
  • self-describing using pseudonymous identifiers

• Cons:

  • more efforts to:
  • pseudonymize
  • propagate composite context ( identities & UID )
  • more complex & esoteric compared to creating new annotations ?

DICOM & Annotations

Non-recommended ways

• Burned in graphics and text

  • Screenshots

• Overlay graphics and text

  • in pixel data, header, separate object

• Presentation states

  • no semantics, only graphics and text
  • popular in clinical PACS

Recommended ways

• Standardized in ages

• Structured Reports (SR)

  • code tree, numbers, 2D, 3D & temporal coordinates, references
  • basis for Cardiovascular, Ultrasound, Mammo CAD, radiation dose
  • Key Object Selection (KOS) - key images are flagged with label

• Segmentation (SEG / DSO)

  • rasterized
  • bitmap, probability, occupancy
  • coded property / anatomy
  • ROI, atlases - pixel level categorical annotation
  • surface mesh

• Radiotherapy Structure Sets (RTSS)

  • 3D coordinates, volumes
  • in RT planning and PET

New in DICOM

• Real World Value Maps

  • coded voxel values ( beyond Rescale type )
  • works in current DICOM images
  • makes pixel value means something ( form of annotation )

• Parametric Maps

  • Combination of RWVM with floats / scaled integer pixels

• 2nd-gen Radiotherapy annotations

  • Conceptual Volumes
  • grammar for combining contours, segmentations

DICOM IODs

• Fiducials

  • Markers with shape & location

• Registration

  • rigid, deformable, established reference frames: atlases

Players

iQ-SYSTEM PACS = https://image-systems.biz/products/pacs/