Ole Eichhorn, CTO of Aperio and a member of DICOM WG 26 has kindly posted this update following their recent meeting at CAP ’08. This group is charged with addressing the issues related to storage of digital pathology images in DICOM. This group is making significant progress toward this shared goal among numerous vendor colleagues and users alike. His summary below summarizes the challenging and interesting issues related to this, the progress made to date and active proposal which is gaining support for a common standard. The group’s efforts in this endeavor will help to further advance digital pathology technologies for all of us. Look for more positive direction and deliverables to follow and thank you to Ole for the guest post bringing us up to date on their efforts.
"A key challenge in any new field is inter-operability between vendors. At first there are no standards, so each vendor creates their own, perhaps leveraging standards from other fields, but after a while customers demand inter-operability, and the vendors themselves benefit from this due to faster adoption and more flexible response to customer needs. In digital pathology, a key standards area has been storing and retrieving digital slides, also known as whole slide images (to distinguish them from photomicrographs, digital images of a single microscope field.
Most digital pathology vendors have chosen an existing imaging standard like TIFF (tagged image file format) or JFIF (JPEG file interchange format) for storing digital slides, but the need for a single standard is becoming stronger. At the same time, digital pathology is moving from education and research applications into the clinical market, where there is already a strong standard for digital images called DICOM (digital imaging communications in medicine), used for disciplines like radiology and cardiology. Large hospitals and pathology labs inevitably ask “can I store digital slides using DICOM” when they investigate adopting digital pathology. They have substantial investments in large computer systems called PACS (picture archive and communication systems) which implement DICOM.
There are two problems with storing digital slides using DICOM, they’re both interesting. First, digital slides are too big, and second, pathologists access their slide images differently to the way their radiologist colleagues access their images.
As regards size, some math; a typical 20mm x 15mm tissue sample scanned at 20X (or .5 micron/pixel) results in an image of 40,000 x 30,000 pixels, or 3.6GB of data, which is compressed to a file size of about 200MB; a more extreme sample of 25mm x 25mm scanned at 40X (or .25 micron/pixel) results in an image of 100,000 x 100,000 pixels, or 30GB of data, which is compressed to a file size of about 1.5GB. The DICOM medical imaging standard presently does not accommodate such large images; both the pixel dimensions and the object size exceed the maximum values supported.
As you know, pathologists pan through slides, zooming in and out often. A typical diagnosis involves a great deal of inspection at 4X, with occasional zooming to 10X, 20X, or 40X to inspect small regions at higher magnification. In contrast radiologists view an entire X-ray or MRI all at once, at full resolution (and they may have other needs such as rotating the point of view). For this reason present DICOM viewers are not efficient for pathologists to use in diagnosing digital slides. The DICOM standard must provide for sub-regions of the large image to be accessed randomly (to support panning), at varying resolutions (to support zooming). Currently the DICOM standard only allows integral access to an entire image.
The DICOM standard is administered by NEMA (National Electronics Manufacturer’s Association), and NEMA establishes working groups to develop supplements to the standard whenever new needs arise. NEMA Working Group 26, of which I am a member, has the charter to address storage of digital pathology images in DICOM. This working group has already developed a supplement which provides an expanded specimen model for pathology, and is now working on another supplement to store whole slide images in DICOM.
The active proposal has the enthusiastic support of all the digital pathology vendors who are members of Working Group 26, including Aperio, Bacus, Bioimagene, Dako, DMetrix, Hamamatsu, Olympus, Omnyx, and Zeiss, as well as the PACS vendors such as GE and Philips. This proposal encapsulates a digital slide as a conceptual pyramid of images, at various resolutions, each of which may be broken into one or more rectangular tiles. Each tile of each image in the pyramid is stored as a separate image in a DICOM series. The following diagram illustrates:
This works around the key limitations of DICOM quite nicely. Because only individual tiles are stored, the size of the overall image object is no longer material, only that each tile be within the pixel dimension and object size limits of DICOM. And because individual tiles are addressable as separate images, using defined DICOM protocols, random access to sub-regions of the pyramid is supported (for panning). Having multiple levels in the pyramid at several resolutions supports the need to access images at varying resolutions (for zooming).
The proposed supplement is vender-neutral and does provide substantial scope for individual vendors to optimize their implementations, which is important for the early days of a new field. Advances in access mechanisms, image formats, compression types, etc. may well occur, and easily can be incorporated into the scheme. The supplement also supports JPEG2000 image objects and JPIP access to those objects as a proper subset, for vendors who wish to make use of that access mechanism.
The state of this effort is that the proposal has been made and discussed, and approved. The next step is to convert the proposal into the formal language of a DICOM supplement, and then work with NEMA Working Group 6, the standing committee which oversees all supplements to the DICOM standard, to get this supplement approved and incorporated. After that individual digital pathology vendors will begin implementing the standard, fostering inter-operability among themselves, and PACS vendors will adopt the new standard into their systems, and begin deploying it to their large hospital and lab customers."