Saturday, September 7, 2013

Share and share alike - CSIDQ

Summary: Image sharing requires the availability (download and transmission) of a complete set of images of diagnostic quality (CSIDQ), even if for a particular task, viewing of a lesser quality subset may be sufficient. The user then needs to be able to decide what they need to view on a case-by-case basis.

Long Version.

The title of this post comes from the legal use of the term "share and share alike", the equal division of a benefit from an estate, trust, or gift.

In the context of image sharing, I mean to say that all potential recipients of images, radiologists, specialists, GPs, patients, family, and yes, even lawyers, need to have the means to access the same thing: a complete set of images of diagnostic quality (CSIDQ). Note the emphasis on "have the means". CSIDQ seems to be a less unwieldy acronym that CSoIoDQ, so that's what I will use for notational convenience.

There are certainly situations in which images of lesser quality (or less than a complete set) might be sufficient, might be expedient, or indeed might even be necessary to enable the use case. A case in point being the need to make an urgent or rapid decision remotely when there is a only slow link available.

For folks defining architectures and standards, and deploying systems to make this happen, it is essential to assure that the CSIDQ is available throughout. In practice, this translates to requiring that
  • the acquisition modality produce a CSIDQ,
  • the means of distribution (typically a departmental or enterprise PACS) in the local environment stores and makes available a CSIDQ,
  • the system of record where the acquired images are stored for archival and evidential purposes contains a CSIDQ
  • any exported CD or DVD contains a CSIDQ,
  • any point-to-point transfer mechanism be capable of supporting transfer of a CSIDQ
  • any "edge server" or "portal" that permits authorized access to the locally stored images is capable of sharing a CSIDQ on request,
  • any "central" archive to which images are stored also retain and be capable of distributing a CSIDQ
  • any "clearinghouse" that acts as an intermediary needs to be capable of transferring a CSIDQ
These requirements apply particularly to the "Download" and "Transmit" parts of the Meaningful Use "View, Download and Transmit" (VDT) approach to defining sharing, as it applies to images and imaging results.

In other words, it is essential that whatever technologies, architectures and standards are used to implement Download and Transmit, that they be capable of supporting a CSIDQ. Otherwise, anything that is lost early in the "chain of custody", if you will, is not recoverable later when it is needed.

From a payload perspective, the appropriate standard for a CSIDQ is obviously DICOM, since that is the only widely (universally) implemented standard that permits the recipient to make full use of the acquired images, including importation, post-processing, measurement, planning, templating, etc. DICOM is the only format whose pixel data and meta data all medical imaging systems can import.

That said, it may be desirable to also provide Download of a subset, or a subset of lesser quality, or in a different format, for one reason or another. In doing so it is vital not to compromise the CSIDQ principle, e.g., by misleading a recipient (such as a patient or a referring physician) into thinking that anything less that a CSIDQ that has been download is sufficient for future use (e.g., subsequent referrals). And it is vital not to discard the DICOM format meta data. EHR and PHR vendors need to be particularly careful about not making expedient implementation decisions in this regard that compromise the CSIDQ principle (and hence may be below the standard of practice, may be misleadingly labelled, may introduce the risk of a bad outcome, and may expose them to product liability or regulatory action).

Viewing is an entirely different matter, however.

Certainly, one can download a CSIDQ and then view it, and in a sense that is what the CD/DVD distribution mechanism is ... a "thick client" viewer is either already installed or executed from the media to display the DICOM (IHE PDI) content. This approach is typically appropriate when one wants to import what has been downloaded (e.g., into the local PACS) so that it can be viewed along with all the other studies for the patient. This is certainly the approach that most referral centers will want to adopt, in order to provide continuity of patient care coupled with familiarity of users with the local viewing tools. It is also equally reasonable to use for an "in office" imaging system, as I have discussed before. It is a natural extension of the current widespread CD importation that takes place, and the only difference is the mode of transport, not the payload.

For sporadic users though, who may have no need to import or retain a local copy of the CSIDQ, many other standard (WADO and XDS-I) and proprietary alternatives exist for viewing. Nowadays web-based image viewing mechanisms, including so-called "zero footprint" viewers, can provide convenient access to an interactively rendered version of that subset of the CSIDQ that the user needs access to, with the appropriate quality, whether using client or server-side rendering, and irrespective of how and in what format the pixel data moves from server to client. Indeed, these same mechanisms may suffice even for the radiologist's viewing interface, as long as the necessary image quality is assured, there is access to the complete set, and the necessary tools are provided.

The moral being that the choice needs to be made by the user, and perhaps on the basis of whatever specific task they need to perform or question they want to answer. For any particular user (or type of user), there may be no single best answer that is generally applicable. For one patient, at one visit, the user might be satisfied with the report. On another occasion they might just want to illustrate something to the patient that requires only modest quality, and on yet another they might have a need to examine the study with the diligence that a radiologist would apply.

In other words, the user needs to be able to make the viewing quality choice dynamically. So, to enable the full spectrum of quality needs, the server needs to have the CSIDQ in the first place.


PS. By the way, do not take any of the foregoing to imply that irreversibly (lossy) compressed images are not of diagnostic quality. It is easy to make the erroneous assumptions that uncompressed images are diagnostic and compressed ones are not, or that DICOM images are uncompressed (when they may be encoded with lossy compression, including JPEG, even right off the modality in some cases), or that JPEG lossy compressed images supplied to a browser are not diagnostic. Sometimes they are and sometimes they are not, depending on the modality, task or question, method and amount of compression, and certainly last but not least, the display and viewing environment.

What "diagnostic quality" means and what constitutes sufficient quality and when, in general, and in the context of "Diagnostically Acceptable Irreversible Compression" (DAIC), are questions for another day. The point of this post is that the safest general solution is to preserve whatever came off the modality. Doing anything less than that might be safe and sufficient, but you need to prove it. Further, regardless of the quality of the pixel data, losing the DICOM "meta data" precludes many downstream use cases, including even simple size measurements.

PPS. This blog post elaborates on a principle that I attempted to convey during my recent testimony to the ONC HIT Standards Committee Clinical Operations Workgroup about standards for image sharing, which you can see, read or listen to if you have the stomach for it. If you are interested in the entire series of meetings at which other folks have testified or the subject has been discussed, here is a short summary, with links (or you can go to the group's homepage and follow the calendar link, to future meetings if you are interested in joining them, or past meetings:

2013-04-19 (initial discussion)
2013-06-14 (RSNA: Chris Carr, David Avrin, Brad Erickson)
2013-06-28 (RSNA: David Mendelson, Keith Dreyer)
2013-07-19 (lifeIMAGE: Hamid Tabatabaie, Mike Baglio)
2013-07-26 (general discussion)
2013-08-09 (general discussion)
2013-08-29 (standards: David Clunie)

Also of interest is the parent HIT Standards Committee:

2013-04-17 (establish goal of image exchange)

And the HIT Policy Committee:

2013-03-14 (prioritize image exchange)

PPPS. The concept of "complete set of images of diagnostic quality" was first espoused by an AMA Safety Panel that met with a group of industry folks (2008/08/27) to try to address the historical "CD problem". The problem was not the existence of the CD transport mechanism, which everyone is now eager to decry in favor of a network-based image sharing solution, but rather the problem of inconsistent formats, content and viewer behavior. The effort was triggered by a group of unhappy neurosurgeons in 2006 (AMA House of Delegates Resolution 539 A-06). They were concerned about potential safety issues caused by inadequate or delayed access or incomplete or inadequately displayed MR images. To cut a long story short, a meeting with industry was proposed (Board of Trustees Report 30 A-07 and House of Delegates Resolution 523 A-08), and that meeting resulted in two outcomes.

One was the statement that we hammered out together in that clinical-industry meeting, which was attended not just by the AMA and MITA (NEMA) folks, but also representatives of multiple professional societies, including the American Association of Neurological Surgeons, Congress of Neurological Surgeons, American Academy of Neurology, American College of Radiology, American Academy of Orthopedic Surgeons, American College of Cardiology, American Academy of Otolaryngology-Head and Neck Surgery, as well as vendors, including Cerner, Toshiba, Philips, General Electric and Accuray, and DICOM/IHE folks like me. You can read a summary of the meeting, but the most important part is the recommendation for a standard of practice, which states in part:

"The American Medical Association Expert Panel on Medical Imaging (Panel) is concerned whether medical imaging data recorded on CD’s/DVD’s is meeting standards of practice relevant to patient care.  

The Panel puts forward the following statement, which embodies the standard the medical imaging community must achieve. 

  • All medical imaging data distributed should be a complete set of images of diagnostic quality in compliance with IHE-PDI.
This standard will engender safe, timely, appropriate, effective, and efficient care; mitigate delayed care and confusion; enhance care coordination and communication across settings of care; decrease waste and costs; and, importantly, improve patient and physician satisfaction with the medical imaging process."

More recently, the recommendation of the panel is incorporated in the AMA's discussion of the implementation of EHRs, in the Board of Trustees Report 24 A-13, which recognizes the need to "disseminate this statement widely".

The other outcome of the AMA-industry meeting was the development of the IHE Basic Image Review (BIR) Profile, intended to standardize the user experience when using any viewer. The original neurosurgeon protagonists contributed actively to the development of this profile, even to the extent of sacrificing entire days of their time to travel to Chicago to sit with us in IHE Radiology Technical Committee meetings. Sadly, adoption of that profile has been much less successful than the now almost universal use of IHE PDI DICOM CDs. Interestingly enough, with a resurgence of interest in web-based viewers, and with many new vendors entering the field, the BIR profile, which is equally applicable to both network and media viewers, could perhaps see renewed uptake, particularly amongst those who have no entrenched "look and feel" user interface conventions to protect.

1 comment:

notthatbbennett said...

Hear Hear!!!

However, this does not seem to be the current trend. The last two sets of "DICOM" CD's with which I have come into contact haven't been (DICOM that is).