[ChimeraX DICOM index]

DICOM Viewers

• Viewers
  • Horos
  • 3D Slicer
  • MRIcro
• Acronyms and Terminology
• Feature Lists
• Public Sources of Files and Info

← Viewers

See also: 12 open-source Mac DICOM viewers, the I Do Imaging list of free medical imaging software, especially Mac DICOM display

• Medical Holodeck
  • DICOM viewer for VR (Vive, Oculus Rift, Windows mixed reality HMDs: Samsung Odyssey, Dell Visor, Lenovo Explorer, Asus Windows, Acer Windows)
  • versions: free, basic ($75/mo), pro ($250/mo), cloud ($290/mo); paid versions have 30-day free trial
• Horos
  • free and open source (LGPL-3.0) but no free documentation; Google group forum not that extensive
  • macOS 10.11+ only, not Windows or Linux
  •   more...
    • 2022: download instructions are sent to the e-mail address entered into a registration form (do not use cgl mail, or these instructions will be “eaten” and never appear)
    • at least to novice, Horos and Osirix Lite interfaces seem very similar (organization, layout, options and corresponding icons)
    • File... Import allows selecting a stack of DICOM files or a folder from which multiple stacks found in subdirectories can be opened
    • datsets are saved in a Horos database as either file copies or links, then automatically reopened upon next use unless explicitly deleted from that database
    • to get to volume visualization from the initial database window:
      • click 2D Viewer icon top middle of database window
      • from the resulting 2D viewer, choose menu item 3D Viewer... 3D Volume Rendering (other choices include MPR options, 3D MIP, 3D Surface Rendering)
      • top bar now allows choosing CLUT or 3D Preset, level of detail, shading, parallel (orthographic) vs. perspective, filtering, etc.:
        
      • a 3D preset (see plist files in repository) includes not just a CLUT but other settings like projection mode, background color, shading, and filtering... perhaps more easily understood by clicking Info on the 3D preset browser in Horos
      • show histogram w/ thresholds by double-clicking certain entries in the 3D preset browser (I guess the ones with xy pairs in the plist file, e.g. Airways II, Gold Bone, Soft + Skin, rather than just a reference to a separately defined CLUT, e.g. Glossy, Mid Contrast, Red on White)
      • CLUT plist files also in repository
      • current state can be saved as a 3D preset
    • mouse modes:
      • defaults: middle mouse translates, right mouse zooms, (left) click-drag adjusts windowing (vertical WL, horizontal WW, diagonal both) in 2D viewer, XY-rotates in 3D viewer; scrolling goes through planes in 2D viewer (Mac trackpad click-drag +cmd does translation, +alt windowing, +shift zoom, +ctrl windowing in 2D, Z-rotation in 3D)
      • * values for WL, WW, zoom, etc. reported continuously in text (and color key labels, if color key shown) on window
      • in 2D viewer, cursor position is also reported continuously
      • 3D volume-viewer icons for changing left-mouse function:
        
        ...from left to right, based on experimentation: windowing (WW/WL), XY-translation, zoom, Z-rotation, XY-rotation, rotate around cursor position (? not good because I haven't figured out how to translate in Z), measure distance, add marker, draw circle (maybe it selects contents for later actions but haven't figured out what), hand-draw arbitrary closed shape (again presumably to do something else later), bone removal based on values specified in secondary dialog
      • 3D surface-viewer mouse modes are only zoom, translate, rotate, center of rotation, and mark; threshold level, resolution (step), and smoothing are set in a secondary dialog, and with smoothing it may not be fast enough for a thresholding mouse mode
    • * orientation cube labeled A(nterior), P(osterior), L(eft), R(ight), S(uperior), I(nferior)
    • * the cropping box has a little ball in the center of each side for grabbing/dragging
    • show color key in 2D viewer with 2D Viewer... Color Look Up Table Bar
    • Horos plugins include fairly simple map operations, calculations, and converting other image formats to DICOM
      
    • single-DICOM-file segmentations seem to be ignored (subset of RIDER Lung data from TCIA [zip file]), if I import a directory with both segmentation(s) and a corresponding image stack, I only see the image stack in the database, and even if I explicitly import a single segmentation file, nothing seems to happen.

      1/22/2019 I can create a segmentation using the 2D viewer menu ROI... Grow Region (2D/3D Segmentation) and save it as a .roi (2D) or .rois_series (3D) file (poorly made example that goes with the lung dataset) that can be read back into Horos. If it's a 3D segmentation, then ROI... Compute Volume can be used. It thinks a while and then shows a results panel with options including opacity and “Save as DICOM” but unfortunately that only refers to a single 2D image. Saving lists the image in the database window from where it can be exported to a DICOM file, but again, it's just one 2D image. Can mask data with the 3D segmentation (ROI... Set Pixel Values to). Youtube video not in English but shows some of this stuff.

    • my example time series (subset of mouse brain astrocytoma MRI from TCIA [zip file]) is just intepreted as a really big stack, but maybe some other steps are needed to treat/mark as 4D. Clicking the “4D Viewer” icon says that multiple series taken at different times are needed as input (whereas my example series that is 600 .dcm image files all in one directory seems to be misinterpreted as one big series with multiple mouse heads stacked together). (In Horos, had the same problem with the time series example linked to the Slicer MultiVolume Explorer documentation, see below)
      
      
    • 1/29/2019 Horos displays RTDOSE, RTPLAN (as in the Head-Neck-Cetuximab study from TCIA) as black-and-white stripes in preview/thumbnail, and trying show them in the 2D viewer gives “no files available (readable) in this series”... although thumbnail images are shown for the CT and RTSTRUCT “from rtog conversion,” trying to show them in the 2D viewer gives the same complaint
• OsiriX
  • OsiriX MD commercial, OsiriX HD for iOS, OsiriX Lite free version
  • recent macOS, supports 64-bit computing and multithreading
  • basic tutorials available for free, advanced for $$
• 3D Slicer (see also the wiki and forum)
  • open source “BSD-style” license, NIH-funded, and has documentation
  • 3D Slicer as an image computing platform for the Quantitative Imaging Network. Fedorov A, Beichel R, Kalpathy-Cramer J, Finet J, Fillion-Robin JC, Pujol S, Bauer C, Jennings D, Fennessy F, Sonka M, Buatti J, Aylward S, Miller JV, Pieper S, Kikinis R. Magn Reson Imaging. 2012 Nov;30(9):1323-41.
  •   more...
    • various filtering options
    • interactive and automated segmentation tools
    • rigid and non-rigid registration (superposition)
    • surface-model creation and manipulation
    • users can save scenes (something like sessions?), each potentially containing multiple scene views
    • native formats: Slicer scene file is MRML (medical reality modeling language), saved with data as a MRB (medical reality bundle) file. From using the program, looks like the primary option for saving individual image series and segmentations is NRRD (nearly raw raster data). Several other formats also read, tiff, jpeg, vtk, ...
    • I tried Slicer 4.10.0 on Mac (desktop and laptop)
    • startup is very slow (2022: still true for v4.11.20210226)
    • many tutorials (the PDF slideshows seem nice, the videos vary), including:
      • 3D Visualization of DICOM Data
        • to display DICOM there are two steps: import the data, then load it from the Slicer DICOM database
        • Volume Rendering module has several presets
        • the volume rendering becomes low-resolution during manipulation
        • in this interface, ROI means a cropping box; the little squares are “handles” to drag the box faces or translate the box in any 2D view or the 3D view
        • if you miss the little square in the 2D view, dragging will instead change windowing: WL (vertical) and/or WW (horizontal)
        • then turning volume rendering on by clicking the eye icon only does it within the smaller ROI
      • also viewed Quick Start Guide (mainly just how to install), Welcome, 4-Minute Tutorial for first-time users (looks like more than 4 minutes and mostly showing pre-made VTK surface models, but interesting)
    • module documentation includes:
      • DICOM
      • Models
      • Segmentations “It is important to remember that segmentations are not labelmaps”
      • Segment Editor – manual segmentation; labelmaps are related to, but not (as mentioned above) the same thing as segmentations
      • Volumes
      • Volume Rendering
      • Sequences (time series)
      
      
    • single-file segmentation DICOM (examples in the RIDER Lung data from TCIA [zip file]) is recognized on import, Modality: SEG in the Slicer DICOM browser; a dialog appeared to suggest I get the Quantitative Reporting module...
      • Extension Manager can be opened by clicking the “E” icon on Slicer window top right.
      • QuantitativeReporting depends on other extensions: PETDICOMExtension, DCMQI, SlicerDevelopmentToolbox, so I clicked to install them all.
      Upon Slicer quit/restart, the segmentations were loaded successfully into the viewer. The example data folder included 2 scans "4" and "8" and a bunch of segmentations, which were aligned to "4" ("8" was offset). In the Volumes module I could delete "8", leaving just the one aligned with the segmentations.

      I tried making my own segmentation in Slicer, as in this video tutorial. There are lots of options, including those analogous to Volume Eraser and Hide Dust. Segmentation save format is only NRRD, and export format choices are only STL and OBJ. Saved NRRD file could be read back in, and was automatically listed as a segmentation. Exported STL file (coordinate system choices LPS or RAS, I chose the default LPS) could be read back in, but was listed as a separate model, not a segmentation, and was not aligned.

      1/19/2019 In the RIDER Lung segmentation DICOM file metadata, manufacturer and software version refer only to Slicer, but the manufacturer model field gives the URL for the Quantitative Reporting module, which sounds like it can save DICOM segmentations. I couldn't use it to save the segmentation opened from previously saved NRRD (error: empty segmentation), but I could save one made in the current session or one previously read from DICOM (RIDER Lung). However, “saving” in this context only means adding it to the internal Slicer DICOM database (SQL), where the RIDER lung segmentation was already. Since the database contents persist through quitting/restarting Slicer, this process at least allows saving my segmentation for later use in Slicer, but doesn't produce a file that can be used in some other program. The Slicer database dialog has an Export button for writing DICOM, but its balloon help says “not yet available” (even in daily build 1/18/2019) and trying to use it anyway gives “Error occurred in exporter” message.

      I see another video tutorial showing a different set of steps for saving segmentations + volume to DICOM. I followed that entire workflow which worked up to clicking the Export button, which did absolutely nothing, as far as I could tell. However, in the video looks like I need a DicomRtImportExport plugin or something like that. I didn't find exactly that plugin name, but I installed one named SlicerRT (and TCIABrowser too since it interested me). Restarted, repeated the entire process, still got error or nothing exported.

      1/20/2019 Got a little farther: if I choose export format RT, I can export a study “to DICOM” using the workflow shown in that video, but only to the Slicer DICOM database and only as modality RTSTRUCT, not SEG. Export type DICOMSegmentation seems to either cause error or do nothing, export type Scalar Volume only exports the images, not the segmentation. Again, no actual DICOM output files, only export to the Slicer DICOM database. The only way I found to get SEG into the database is with the Quantitative Reporting module, as reported above.

    • time series (subset of mouse brain astrocytoma MRI from TCIA [zip file]) was read without error messages; need to investigate further.

      1/22/2019 Now I see it is loaded as a 50-frame “multivolume,” apparently their term for a time series, in agreement with the 50-frame time slider in ChimeraX, and can be played using the MultiVolume Explorer module. Another example time series (prostate) is provided in that module's documentation.

    • 1/29/2019 Slicer displays RTDOSE as colored blobs if associated RTPLAN is also opened, RTSTRUCT as colored overlay if associated CT is also opened

      2/12/2019 Slicer can render RTSTRUCT in 3D as lines or multiple transparent surfaces, in addition to the overlays on the 2D slice views. Settings/options for creating surfaces from contours are described in the Segment Editor documentation.

  MRIcro family of programs
  • NIH-funded; open source “BSD 2-clause” license
  •   MRIcroGL real-time interactive volume rendering, requires modern graphics card (Windows, Mac, Linux)
    • in 2018, these downloads for Mac each included two versions:
      • [github] ...2018 version includes MRIcroGL and MRIcroMTL (uses Apple's Metal instead of OpenGL)
      • [NITRC] ...2018 download included MRIcroGL and “MRIcro for OSX” (I guess with lower graphics requirements), a manual (PDF, Jun 2017), and several sample datasets and templates for overlays
      • MRIcroGL itself was different in those two downloads: the color-scheme menu was in a different place, the github one had fewer import options and shader options, etc.; now, github has version marked 2021-10-07 and NITRC 2021-10-06... the same??
    • developed by Chris Rorden (and lab?), professor at University of South Carolina, co-director of their brain imaging center, winner of the university's Breakthrough Award, etc.
    • MRIcroGL documentation wiki maybe not exhaustive, but see videos...
    • 2022 update: nice tutorial videos for the most recent release at Chris Rorden's YouTube channel:
      • MRIcroGL Intro (~20 mins)
      • MRIcroGL 2021 Release (~30 mins)
      • MRIcro open-source viewer for macOS (~10 mins) (available from App store, lighter-weight but has a lot in it)
    • GLSL volume ray casting rendering
    • ...* surface renderings do look really nice
    • formats: primarily NIfTI (Neuroimaging Informatics Technology Initiative: .nii, .nii.gz, .hdr/.img). Documentation lists other formats.
    • 2018 version did not read DICOM directly, included slightly unfriendly (command-line) DICOM → NIfTI converter ...apparently this conversion is quite complex, the author even wrote a paper about it in 2016
    • tried to import Allen Inst. OME TIFF but got error about incompatible compression type
    • 2 September 2019: latest release supports drag-and-drop to open DICOM folders! Also has expanded TIFF support (now can open my Allen Inst. OME TIFF example) and Python scripting. June 2022: I can still open the Allen Inst. OME TIFF in the latest release, but thought something was wrong... asked Dr. Rorden, and shortly thereafter he gave me an updated version that works perfectly on this dataset; one problem had been different spacing of planes in different dimensions.
    • full colormap control for rendering different tissues differently
    • shaders are text files, users can create their own custom combinations
    • I tried MRIcroGL on 2015 MacBook Pro and 2012 iMac running High Sierra (10.13): nice graphics, smooth response to manipulation, LOTS of rendering options, sliders for individual parameters, many built-in color schemes (some with intensity values for specific tissues in CT), clipping, cutouts (box-shaped), slice views, overlays ...
    • 2022 update: tried the 2021 release on two MacBook Pros, the 2015 one running Catalina (MacOS 10.15.7) and 2021 running Monterey (MacOS 12.4). Most everything works fine on both, except that drag-n-drop DICOM on the main window doesn't work to open it on the newer laptop. The other, longer method works fine, however: drag-n-drop on format conversion dialog (opened from Import menu) to create a NIfTI file that can then be opened the usual way.
    • color schemes not applicable to “visible human,” I guess because that dataset consists of thin-section (color?) photographs rather than monochrome imaging
    • automatically remembers/opens last data, opening something else replaces it
    • can show multiple datasets as overlays in a single program instance, or multiple instances can be run simultaneously and “yoked” to synchronize slice display
    • click-drag rotates, orientation cube shown, scrolling zooms (up to a limit); no panning the 3D view? (Ctrl-drag translates slices)
    • use keys a(nterior), p(osterior), l(eft), r(ight), s(uperior), i(nferior) to move slice plane
    • prefer Chimera(X) interactive clip translation/rotation over the 3 sliders
    • * useful predefined color scheme values for specific tissues (see below)
    • * nice color key automatically shown/updated (yet optional)
    • used the DICOM → NIfTI converter on one of the Belarus chest datasets; got some errors/warnings, some .nii output didn't look right, but the last/largest one appears to have all the data: bones, lungs and other tissues shown nicely with the appropriate color schemes
    • 2022: “MRIcroGL, FSLeyes, Slicer, GE, Mediso and Imeka all use dcm2niix to handle DICOM images. The BSD license would allow your group to use it as well...”
    • 2022: Rorden's repository of DICOM edge cases may also be extremely useful: https://github.com/neurolabusc?tab=repositories&q=dcm_qa&type=&language=&sort=

    
    
• Web-based viewers:
  • NiiVUE WebGL 2.0 medical image viewer (see X3D demo)
  • dicomlibrary.com
  • Cornerstone
  • OHIF (Open Health Imaging Forum)
• many basic Windows-only DICOM viewers that I won't bother to list here

← Acronyms & Terminology

See also: DICOM modality attributes

• ADC – apparent diffusion coefficient
• AIM – Annotation and Image Markup effort to standardize medical image annotation (NIH/NCI)
• CR – computed radiography
• CT – computed tomography
• CLUT – color lookup table
• DCE – dynamic contrast enhancement
• DICOM – file standard for digital imaging and communications in medicine
• DTI – diffusion tensor [magnetic resonance] imaging
• DWI – diffusion-weighted [magnetic resonance] imaging
• Hounsfield scale – normalized units used in medical CT
• LPS – left, posterior, superior anatomical coordinates
• RAS – right, anterior, superior anatomical coordinates
• MIP – maximum intensity projection
• MPR – multiplanar reconstruction or reformatting (generating “slices” in arbitrary directions or even curved)
• MRI or MR – magnetic resonance imaging
• PACS – picture archive and communication system
• PET or PT – positron emission tomography
• QIN – Quantitative Imaging Network (NIH/NCI)
• ROI – region of interest
• RT – radiotherapy
• structure set – set of areas of significance
• SR – structured report
• SUV – standardized uptake values (PET/CT)
• US – ultrasound
• WL – window level
• WW – window width

← Feature Lists

Potential tasks for medical image analysis:

• view header information
• anonymize
• zoom, rotate, flip
• invert dark/light
• adjust contrast and brightness
• color processing, color mask
• adjust windowing/thresholding #1613
  12/17/18 Darrell mentions "color lookup table (CLUT) editor," HOROS has mouse modes to adjust:
  • leveling - moving thresholds in parallel along histogram
  • windowing - narrowing/broadening visible range of data values (moving bounding thresholds in opposite directions)
  12/17/18 text in VR to report the threshold levels #1614
  (Elaine: maybe labeled color key as in MRIcroGL or Horos, WW/WL values as in Horos)
  (related: Phil mentions popup reporting data value at voxel like HOROS)
• built-in color schemes:
  • HOROS 3D presets (he often uses Airways II)
  • HOROS CLUTs
  • 3D Slicer LUTs
• convert data types, formats
• filter (smooth, sharpen, etc.)
• other data editing (copy, paste, erase?), Chimera Volume Eraser
• clipping
• mark points and regions
• select regions, potentially of irregular shape
• detect/mark edges, boundary surfaces; mask and segment
  12/17/18 Beth and Darrell mention segmentation, particularly by hand marking
• annotate (text or geometric objects in image, metadata outside of image)
• geometric measurements (distance, angle, area, volume, etc.)
• data measurements (mean, SD, min, max, plot values along an axis, etc.)
• texture analysis (not sure what this means)
• play time series, save movie files
• compare datasets (with and without contrast agent enhancement; before, during, and after therapy; normal vs. abnormal brain; etc.)
• overlays (annotations, localized brain activity, etc.)
• registration with atlases/templates (anatomical landmarks, reference organs) or between multiple datasets, rigid or non-rigid (with scaling/warping)
• along those lines, medical image fusion ...essentially 3D superposition of multiple datasets or maps
• fetch data from public sources

Other desirables from discussion:

• large data performance, especially in VR
• multi-user VR meeting
• working with multichannel data, e.g. annotations from AI/deep learning may be saved as color channels to be overlaid on the exptl. dataset

← Public Sources of Files and Info

• The Cancer Imaging Archive (TCIA) has many collections of datasets (public collections also listed on the wiki)
  • formerly the National Biomedical Imaging Archive (NBIA)
  • downloading requires NBIA Data Retriever app
  • however, there's also a REST API for programmatic access
  • de-identification procedures
  • general data-usage policies
• Patient Contributed Image Repository (only 10 downloads)
• Medical Image Samples (only 5 multiframe sets)
• open-access medical image repositories – a big list, but images are not necessarily in DICOM format
• NIH DeepLesion dataset – over 32,000 annotated lesions in CT images available via Box, but currently only PNG format

• David Clunie's Medical Image Format Site (some files in Images section near bottom)
• Introduction to DICOM