1. DICOM and the Pathology Community Experience
Bruce Beckwith, MD
Pathology is in a transition phase, moving from analog to digital images. We are years behind radiology.

2. Whole Slide Imaging (WSI)
Scanner
This file is 110,000 by 75,000 pixels, almost 8 GIGApixels. Uncompressed, it would take about 24 GB of storage, even compressed, it is almost 2GB

3. Pathology Disciplines
Tissue (Surgical) Pathology
Tissue from biopsies, resections & autopsies
Cytology
Individual cells from smears/scrapings or fluids
Clinical laboratory
Blood smears, protein electrophoresis, etc.

4. Surgical Pathology Workflow
Pathology workflow starts with a specimen
Dissection
Chemical processing
Cut thin sections and place on glass slides
Stain with a variety of techniques
Chemical
Immunochemical
in-situ hybridization

5. Why Move To Digital Imaging?
Location independence
Sharing of images with clinicians
Enables new analysis techniques
Computerized screening of pap smears
Image analysis for quantitation of special stains
? Computer aided diagnosis for other specimens
Remote frozen sections, immediate sharing of cases with consultants
Computer aided detection in mammography 5

6. Comparison of Digital Imaging
Radiology
Digital acquisition
Manageable file size
Many clinician interpretable
Cost savings compared to analog
Computer aided detection for mammograms
Pathology
Mainly analog data which is digitized
Very large file size in pathology
Some clinician interpretable
Incremental costs in addition to analog
Computer assisted screening for pap smears

7. Resolution Challenge Pathologists need to be able to zoom in quickly
Also need rapid pan and focus across the entire slide

8. The Image Size Challenge
1 focal plane
24 bit color
40x magnification
15 Gigabytes
10 focal planes
24 bit color
40x magnification
3.75 Terabytes

9. Navigation Challenge
Main challenge is rapid pan, zoom, focus, advancing to next slide
Intuitive “driving” of the slide will help transition
Some equipment is trying to recreate microscope “feel”

10. Slide Navigation Device
Perhaps there will Kinect like input possible in the near future.

11. DICOM Digital Imaging and Communications in Medicine
Voluntary standards organization
Image exchange standard for CLINICAL images
27 working groups
Anyone with a material interest may participate
Version 3 of standard released 1992
Started as partnership of ACR and NEMA
Driven by the radiology community
Most working groups address radiology areas, but dentistry, endoscopy, cardiology, etc are represented by working groups
Has been instrumental in arriving at the current state of radiology where imaging systems are widely interoperable with both pacs and ris. 11

12. DICOM Overview Communication standard High level standard, conceptual
Facilitates interchange, doesn’t mandate internal storage formats within PACS
Image object definitions are central
Widely adopted in radiology
Addresses workflow as well as images
Specimen vs patient
Workflow analogy
Put up comparison side by side 12

13. Pathology in DICOM Visible light supplement approved 1999
Incomplete and rarely used
Doesn’t support the complexity of Pathology practice
Pathology WG needed
Created WG-26 Fall 2005
Has met about 20 times
Representatives from most major pathology imaging vendors
Also pathologists, consultants and researchers
90+ subscribers to listserve
60+ organizations
>10 countries
Make the point pathology/lab driven 13

14. WG-26 Goals Initial goals: Long term goals:
Extend minimal capabilities to describe specimens in DICOM
Create a mechanism to allow exchange and use of whole slide microscopic images within DICOM
Long term goals:
Other imaging modalities, such as multi-spectral images, electron microscopy, flow cytometry, clinical lab images
Pathology presents specific challenges and opportunities to DICOM. In particular:
1. Some pathology-related image formats do not as yet have applicable DICOM Information Object Definitions. Examples include whole slide images, high-order multi-spectral images, flow cytometry, electron microscopy and others. [Note – Flow cytometry applications have much technical overlap with imaging modalities, even though they do not form an image. It will be important to make imaging standards that have cross-application utility between flow cytometry and immuno-histo-chemistry.]
2. Many pathology processes (for example, flow cytometry) provide challenges in distinguishing "objective" image data from "interpretive" image information. These challenges will include specifying standards for pathology specific markup that would reliably distinguish "unprocessed" from "processes" image data, image annotations from primary image data and "constructed" images from "raw" images, etc.
Future Work Items
• Structured Reports and/or Evidence Documents in Pathology involving full demographic information;
• Correlation of radiologic and pathologic images, including image-guided biopsies;
• Coding information based upon existing WHO codification, SNOMED-CT and also ADICAP thesaurus;
• Navigating in a hierarchy of images by means of annotations of images and/or drawings (e.g., gross imaging annotated with blocks’ localization);
• Dealing with Tissue Micro Arrays (one slide for hundreds of patients) with a link to patient information;
• Integration of automated image analysis tools with WSI. 14

15. Supplement 122
Specifies a specimen description model which allows description of:
Type of specimen
Procurement and processing steps
Sampling methods
Physical attributes of slides
Final text approved June 2008
This was a necessary first step to allow full use of pathology images with needed clinical data in DICOM
Attributes will generally come from the LIS if available
These are attributes which are needed to understand and interpret the image
The pathologist’s report is not part of this specimen description 15

16. Pathology Imaging in DICOM
Base Std
Supp 122
The initial DICOM framework works ok for limited cases such as specimen mammography, but breaks down when more complicated information is needed.
Supp 122 added support for a variety of new concepts and relationships
Disambiguates specimen and container
Container is target of image
Container may have more than one specimen
Specimens have a physical derivation (preparation) from parent specimens
When more than one specimen in an imaged container, each specimen is distinguished (e.g., by position or color-coding) 16

17. Implementation Issues
Supp 122 has the needed data elements, BUT most AP LIS Systems don’t have these data at the SPECIMEN level, if at all
Unique slide ID may not be explicitly present
No ability to identify subregions of a slide/block
Staining and fixation information often co-mingled
Specimen descriptions difficult to parse out from large text blocks
Dictionaries may be poorly implemented
Mention Gilberston study from Case presented at APIII 2007.
Problems with the information in LIS, not really with DICOM model 17

18. Supplement 145 – Whole Slide Images
Need a new DICOM Image Object Definition
Challenges
Vast size
Need for intuitive and fast viewing interface
DICOM specific issues
Image pixel dimensions limited to 64k x 64k
Image size description limited to 4GB
Desirable to be backwards compatible
Efficient sub-region access
Most DICOM services assume entire image transmission
Most DICOM IOD’s are concerned with much smaller images and they assume complete image transfer (e.g. image store, image move). 18

19. Tiling and Multi-frame encoding
Whole Slide Image divided into tiles
Each tile encoded into a frame of multi-frame image object
Per-frame header gives spatial location for each tile: X, Y, and Z (focal plane)
Multi-frame image object
Fixed Header
Per-frame header
Dimension data
Pixel data
H Solomon GE 19

20. typically in 1 DICOM Series
Image Pyramid
Single frame image
Thumbnail
Image
Multi-frame image
(single object)
Intermediate
Image
Multi-frame image
(single object) may include multiple Z-planes, color planes
Facilitates rapid zooming to precompute multiple resolution levels
May add 33% to the size of the file.
Baseline
Image
All image objects
typically in 1 DICOM Series
H Solomon GE

21. Localizer image “flavor”
Thumbnail image (single frame) plus navigation links to each frame at each resolution
Each tile of other resolution images has its corresponding area identified in thumbnail
Full description of target tiles
Object Unique ID and frame number
Resolution
Z-plane, color
Multiple target frames can overlap
Different resolution, Z-plane, color, etc.
Presentation and any interactive behavior is not defined in standard
Mention sparse tiling
Mention that there is significant overhead but the multiresolution pyramid improves performance
H Solomon GE

22. Supplement 145 - Next Steps
Supplement approved August 2010
DICOM is now able to handle most pathology and lab images
Most slide scanner vendors have been involved, along with some PACS vendors
Need to engage LIS vendors and publicize these changes
Pathologists can educate their LIS vendors, talk to hospital PACS administrators and talk to imaging vendors encouraging all of them to support this new digital workflow for pathology.

23. Pathology Imaging Workflow
Case info
Pathology
Order &
Specimen
info
LIS /
APLIS
Slide preparation
Slide prep data
Slide ID
Scanning orders
Pathology
Workstation
Gross
specimen
accessioning
Images
Specimen Images
Whole Slide
Scanner
Surgical or
biopsy
procedure
Our attention up to now has been on the WSI image itself, the output from a whole slide scanner modality. However, that scan process must be seen as just one step in the entire anatomic pathology workflow, which begins with collection of the specimen in a surgical or biopsy procedure, or even earlier with the original screening exam. Detailed discussion of this workflow is not within the scope of this presentation, but I will briefly address specimen management in DICOM as specified in Supplement 122.
Images w/ slide prep data
Images – X-ray, U/S, optical, etc.
PACS
Adapted from H Solomon GE 23

24. Image Sharing Currently some pathologists include snapshots in reports
Tumors, specimen margins, unusual findings, etc
WSI allows ability to review slides remotely with clinicians
The ability to correlate slides with other images would be useful
Gross specimen images
Endoscopy images
Radiology images

25. Challenges to Wider Adoption
Storage and bandwidth
PACS storage is relatively expensive
Don’t want to transfer entire huge files
Pathology systems need to become more image centric (as opposed to report centric)
EMR’s need to be able to accept or connect to images and display correctly
Security, credentialing, optimized viewers, etc

26. Summary
WG-26 has created supplements to incorporate modern digital pathology within DICOM
The collaboration of DICOM, IHE and HL7 has led to a broad based standards effort for digital pathology
The availability of a digital workflow for images will enable major changes in the practice of pathology
DICOM support for radiology, pathology, surgery, and radiation therapy opens the door for true integration of data from these areas
Easy quantitation, collaboration
In the next 5-10 years expect major change to more automated processes in anatomic pathology and widespread adoption of wsi 26 26

27. Acknowledgements Members of DICOM WG-26 Harry Solomon, mentor to WG-26
IHE Anatomic Pathology WG
HL7 Anatomic Pathology WG
DICOM Website: