News & Announcements

NEMA Working Group 6 Discusses Whole Slide Imaging for Pathology

Meeting notes prepared by Ole Eichhorn, Chief Technology Officer, Aperio

Just wanted to report on my attendance at a NEMA Working Group 6 meeting in Washington DC on Thursday, 10/29. WG 6 is the “meta” working group that coordinates all the revisions to the DICOM specification made by all the other working groups, including WG 26 which is focused on incorporating Pathology into DICOM.

At this meeting WG 6 reviewed Supplement 145, Whole Slide Imaging for Pathology, which was proposed for adoption by WG 26. Aperio has contributed significantly to this supplement, and recently it has been revised and “DICOMified” by Harry Solomon, a GE employee who was a member of WG 6 and is their liaison to WG 26. Harry and I were accompanied by Bruce Beckwith, who serves as the chairman of WG 26 on behalf of the CAP.

Linked from here are two documents, Supplement 145, Draft 9, and a PowerPoint Presentation giving an overview of the supplement and various issues prepared by Harry for the meeting.

The bottom line is that the proposal is moving. It could be approved as early as the WG 6 meeting scheduled for June 2010 in Barcelona. This process has been under way for four years now but the end is in sight.

The working style at WG 6 meetings is that whatever supplement is under discussion is projected, and edited in realtime with everyone wordsmithing. The members are pretty technical and [of course] familiar with DICOM and imaging.

Notes from the review and discussion:

Overall Approach
  • The essential approach of the proposed supplement is to divide up a WSI into a virtual pyramid with tiled levels, and then to map each of the tiles into images within a DICOM series.
  • There is a new mechanism within DICOM, recently adopted, called “multiframe images”, which enable multiple individual images to be grouped. This type of compound image requires less storage for metadata since the data only have to be stored once and apply to all the frames. Each of the frames are individually addressable and retrievable, providing the essential capability needed for WSI. Harry adapted the original proposal to use multiframe images, which is current style within DICOM. I expressed a concern that this would raise the bar for implementation, since a PACS vendor would have to implement support for multiframe images before implementing support for WSI, but apparently all PACS vendors are already implementing multiframe images because they are being used by a wide variety of modalities
Tiling, Z-planes, and viewing
  • There was some discussion about whether tiles have to adjoin each other or can overlap. This had been discussed at length in WG 26 meetings since there are some vendors who create overlapped tiles. WG 6 would much prefer NOT to have overlap, as it creates ambiguity in which image data to use. The supplement was modified to require a regular non-overlapped tiling. It is acceptable for the tiling to be sparse. It is also acceptable to have multiple layers at a given resolution, to accommodate multiple signals. This will be perfect for us to store fused images
  • There was some discussion about storage of Z-planes, and whether each Z-plane is a separate pyramid, or whether they can share lower resolution levels. Again, this duplicated a long discussion previously had in WG 26, at which the flexibility for either was preserved. This is good for us as we only store Z for the highest resolution images, but our architecture can handle entire pyramids for each Z-level.
  • WG 6 asked about whether WSI vendors planned to hook viewers directly to PACS systems, or whether they would install a dedicated server adjacent to a PACS, retrieve objects from the PACS, and then serve image data to the viewer in a proprietary manner. Apparently this is the approach some radiology vendors are taking for performance, as it accommodates incremental transmission which DICOM does not. It was more a point of interest than anything bearing on the supplement.
  • There was a discussion about “presentation states”, a DICOM concept which defines how an image is mapped to be displayed on an actual device. Examples would include false coloring or other filters. New presentation states are not treated in the supplement and it was agreed none are needed.
Ancillary images and image details
  • The supplement provides for one or more ancillary images such as slide labels, macro images, pictures of the patient (!), gross images, etc. DICOM has a PHI indicator which will have to reflect whether any of the ancillary images contain PHI.
  • A question was raised whether images acquired via multiple image passes actually constitute separate images. It was agreed that this is up to the vendor to decide. The image metadata can indicate how the acquisition was performed.
  • There was discussion about specification of the optical path. There is an older supplement to DICOM for microscopy images which provides for specifics like light source, condenser, objective, eyepiece, etc. Identification of the imager seems preferred to description of the implementation inside the imager.
  • The support and application of ICC profiles was discussed; DICOM is moving in the direction of requiring ICC profiles for all acquired images. A vendor can always specify a generic profile such as sRGB if there’s no device-specific profile available. This only applies to RGB images. No specific changes to this support are required in the WSI supplement.
  • DICOM includes extensive capabilities for annotations, and the annotation “module” has been enhanced to extend annotations over all the frames in a multiframe object. There was discussion about how this applies to WSI. It was concluded that nothing had to be changed, whew. The annotation model is somewhat different to ours; it does support vectors, but not regions, and there are no data attributes directly associated with annotations. Overall we’ll have a bit of work to do to adapt storage of our annotations into DICOM.
Worklists
  • There was discussion about “modality worklists”, essentially instructions to the imager to acquire an image. Various attributes of an acquisition can be specified, for example Z-levels, resolution, frequency band, and subregions. Various attributes of the image as stored can also be specified, such as synthesized resolution levels, tile sizes, etc. I’m not sure how useful this will be in our workflow but eventually we’ll probably have to support this; it would allow a “generic” PACS to specify acquisition characteristics for individual slides. There would likely have to be some extensions made to the current DICOM spec specifically for WSI imagers, e.g. specification of a particular slide within the autoloader, or at least a mapping from current attributes.

That’s the news; the next WG 6 meeting is back in Washington, in January, but this supplement probably won’t make the already-overloaded agenda and it won’t be enough time for public comment anyway. The next WG 26 meeting is at USCAP, in March, in Boston.

PS NEMA headquarters is in downtown Arlington, right next to the Potomac river. A pretty nice view of the Capital Mall at middle left, and the Pentagon at right. Yeah, it was rainy. The projected WSI supplement is reflected in the window at left :)