Michael J. Gray: PACS Technologies

Wed, 27 Feb 2008 19:08:41 GMT

Cost-effective Business Continuity Solutions - So much more than Data Back-up

Most Radiology PACS currently in use have some sort of data back-up in place. At the very least, the Directory database and the Data database are backed up daily to digital tape. In my opinion, digital tape is not reliable and the problem is you don't know what data you have lost until you try and retrieve it. My low opinion of digital tape is supported by a number of reports from the field. I suspect the vendors that continue to insert digital tape back-up solutions in their early round quotes, do so in order to keep the price of the system down, but a much better solution is worth a few dollars more.

The "tape-less" back-up is a much better back-up solution. Instead of digital tape on a shelf or in a mechanical jukebox, a far more reliable and performance-oriented solution is to store the back-up copy of the Directory and the Data on spinning disk. Thanks to today's pricing, a multi-processor, multi-core server coupled with a disk-based storage solution is only slightly more expensive than a digital tape library. I think the reliability is worth the additional investment.

Why stop there?

Instead of just writing a copy of the Directory on the back-up storage solution, why not install a second instance of the Directory application (Oracle, Sybase, DB2, SQL, etc.) on the back-up server? Now you have a reasonably cost-effective Disaster Recovery solution, depending on where you have physically placed that back-up system.

Why stop there?

Why not add a second instance of the PACS application to the back-up server? Now you have a reasonably cost-effective Business Continuity solution. Of course this complicates the PACS application considerably. The optimal software configuration would have the two Servers (Primary and Secondary) functioning in an "Active-Active"mode, and that would mean that the Directories are being automatically synchronized in near-real-time, and the study data is being copied from Primary to Secondary on a fairly regular basis.

Only the newest generation of PACS can support this configuration. Most of the PACS being sold today can support a "tape-less" back-up server, but they do not support a second instance of the Directory application on that back-up server. The few that do support a second Directory do not support a second instance of the PACS application. Fewer still that support a second instance of the Directory and the PACS application have the back-up system operating in a standby mode. The Back-up takes over only when the Primary is off-line for scheduled or unscheduled maintenance. While this version of back-up may not sound so bad, the fact is that the failover and eventual reconstitution processes are often manual and labor intensive.

The point in all of this is, with today's cost of hardware it doesn't make sense to settle for a back-up solution with questionable reliability, when a much more reliable Business Continuity solution is affordable. The problem is most PACS currently being sold are "old" generations of system architecture wrapped in pretty GUI and flashy 3D applications. While GUI and display applications are important, I believe that the system architecture that supports a solid Business Continuity solution is more important, and sooner or later those old generation PACS are going to be upgraded. You can tell a lot about the longevity of a PACS, by investigating the various back-up solutions that it can support. Why start a five year contract with an old PACS? Do you have room for a forklift in your data center?

Tue, 12 Feb 2008 01:29:47 GMT

The Problem with Proprietary Data/Object Formats - their Impact long after Data Migration

This is another take on a long-standing problem with most of today's Radiology PACS: proprietary Data/Object Formats. It has been at least four years since Presentation States and Key Image Notes were included in the DICOM standard, yet the majority of PACS vendors continue to treat these key work products as proprietary objects. The most consistent excuse is "There are many more features on our engineering schedule considered to be more important to our users."

I can almost believe that story, since I have found that most users are not aware of the implications of proprietary data objects. Since almost every PACS supports the creation and display of Presentation States and Key Image Notes, the fact that most PACS treat these as proprietary objects is lost on most buyers and eventual users. Provided that these objects are kept within a given PACS, there is no apparent negative to their being proprietary. The user may not experience a situation where the proprietary nature of these objects presents a problem.

The problem arises when the user of one of these proprietary PACS tries to forward study data to another Facility or Health System that is using a different PACS. Whether that other PACS is DICOM conformant or not, unless it is the same PACS, those presentation States and Key Image Notes cannot be transferred, accessed, or displayed. Physicians using the other PACS will not have the benefit of seeing exactly what the radiologist interpreting the study saw in the images or what he may have typed as a text message. The benefit of these "work products" is lost.

The problem also arises when a user of one of these proprietary PACS tries to copy study data to a CD/DVD. The proprietary work products either cannot be copied, or they cannot be accessed and displayed by another PACS. This is one of the reasons why there is so much consternation over the current CD/DVD copying solutions on the market. The vendors of these proprietary PACS typically have to place a copy of their own viewing software on these CD/DVDs, because their proprietary viewer is the only way to view their proprietary study data.

The real problem will manifest itself only after the user has decided to replace the proprietary PACS with the next PACS. Data migration services will typically migrate the study pixel data to the next PACS, but few of these services currently migrate any proprietary study-related data objects. To do so would require knowing where these objects were stored in the PACS, how to extract them and how to convert them to their DICOM counterparts. This extraction, conversion, migration is not being performed and as a result, those proprietary data objects are lost forever. The images are available for historical comparison in the next PACS, but none of the proprietary work products are available. Now imagine the implication of having to window and level all of these priors again, when they are recalled for viewing with the new images. Imagine not having the spine labels, and not having any other annotation or overlay graphics created when the prior was first interpreted. That's working without benefit of prior information, or a possible expenditure of time redoing all that work.

A PACS should treat Presentation States, Key Image Notes, .wav files, Technologist Notes, Scanned Documents, even the Radiology Report as DICOM Objects, not only so they can be shared with other systems today, but also so they can easily be migrated and used in the next PACS. DICOM-conformance is always in the user's best interest.

Now if a prospective buyer knew the negatives associated with proprietary data objects, would they choose a proprietary PACS anyway? Logic suggests that they should think twice. At the very least, if an organization goes ahead with the purchase of a PACS that still creates any proprietary data/object formats, that organization should negotiate a "no-cost" data migration clause in their contract that pins the cost of moving these proprietary objects to the next PACS on the vendor who has continued to choose NOT to conform to the standard.

Lack of DICOM conformance is a type of vendor lock. I believe that the PACS vendors still believe that anything that complicates moving to another vendor's PACS may persuade the organization to stay with the incumbent. It's time to make them pay for that strategy.

Thu, 27 Sep 2007 23:13:19 GMT

Take the Archive Out of PACS

Those of you that have been following my recent posts on the subject of PACS-Neutral Archive might find it useful to visit the HIMSS or Emageon web sites to access a webinar delivered today to an audience of 70+ members of HIMSS.

The seminar covered the subject of Tag Morphing and explained how some very common problems faced by Health Systems today can be resolved by deploying a PACS-Neutral Archive; problems such as the sharing of a single archive among multiple dissimilar PACS, and the elimination of future data migrations. The Emageon web site offers the visitor the option of downloading a collection of white papers that describe the concept of PACS-Neutral Archive and Tag Morphing in more detail.

Check it out.

Wed, 26 Sep 2007 18:37:48 GMT

Is new Stark Exemption an Opportunity?

I came across an article in Imaging Economics titled "Surveys Show Paper Legacy Tough to Shake"

What caught my eye was the second paragraph statement "A new Stark exception allows hospitals to donate health information technology in the form of an EMR to private physicians."

I was wondering if the definition of "EMR" could be extended to radiology web viewer? Is this possibly a mechanism for providing the necessary hardware (PC), software and connectivity services to the referring physician office to get them to stop requesting paper and film?

The article is worth reading as it explains why "more than 50% (hospitals) continue to print and distribute paper lab and imaging reports." This does not come as a surprise to me, but it occurs to me that if so many hospitals are still printing paper radiology reports, a similarly large number must also be distributing hardcopy images.

Clearly the success of a Radiology PACS depends on turning off a large percentage of hardcopy and getting the referring physicians to access images and reports from their offices electronically. I have long argued that the cost of providing a suitable PC and basic connectivity services is more than paid for by the value of the hardcopy. Many clients were concerned about the Stark implication. Is this exception an opportunity?

The article goes on to explain that 62% of hospital executives surveyed in February said their organization had no plans to donate technology. "They're waiting to see how the government changes the landscape. How will it affect their nonprofit standing, that kind of thing." Once again, I think this is a shortsighted point of view. The continued printing of hardcopy films is certainly affecting their bottom line. Why not take advantage of this opportunity to legally equip their referring physicians with a much less expensive method to access images and reports?

Tue, 25 Sep 2007 21:39:18 GMT

Separating Storage from the PACS is Good First Step

I was browsing today when I came upon an all too brief article that appeared in HealthCare Informatics in August, 2007. The title of Stacey Kramer's article is A Two-Tier Solution, and the first sentence states that "Memorial Hospital found it takes two vendors to handle imaging properly - one for PACS and one for storage."

According to the article, Memorial Hospital decided to combine a McKesson PACS with "IBM's tiered storage solution". Unfortunately the article provides no real information on the actual configuration, or any explanation as to why this was the ideal combination.

Without any detail, I am left to speculate that this is merely an example of the customer requiring the PACS vendor to substitute the customer's favorite storage solution for the storage solution originally proposed by the PACS vendor. If this is the case, this is hardly a breakthrough.

If in fact, "IBM's tiered storage solution" was their GMAS configuration featuring Bycast's potent Information Lifecycle Management software, that would be a significant upgrade over the typical PACS configuration that features direct attached storage, but once again, hardly a breakthrough.

There is no evidence to suggest that there is a separate data Directory on this separate storage solution, and that the data is in any way being managed independently of the PACS application. Bottom line, the McKesson PACS still controls the study data, and years from now, when Memorial Hospital decides to replace this PACS with another PACS, they will have to migrate all of that study data through the McKesson PACS and through the new PACS, even if this migration is right back to the same IBM storage solution.

Choosing a separate Storage Solution was a good First Step, but the next step would have been to interface the McKesson PACS to a PACS-neutral Archive. There are a number of PACS-neutral Archive software applications that could utilize the same IBM storage solution, but in this case, the study data would be controlled by the PACS-neutral Archive and not the McKesson PACS application. The study data would not have to be migrated downstream, when the McKesson PACS is replaced.

The good news is that it is never to late to build a PACS-neutral Archive, and pro-actively migrate the study data to this archive long before the data migration task gets that much bigger and much more expensive.

I have written several White Papers on the subject of Pro-active Data Migration and PACS-neutral Archives. The papers are too lengthy to publish on this web site, but they are freely available to anyone forwarding an email request.

Fri, 07 Sep 2007 00:29:51 GMT

PACS for the Smaller Organization

Over the last several months a number of posts have shown up on auntminnie.com and pacs-admin@yahoogroups.com asking readers opinions on some of the smaller PACS solutions in the market. I assume that most of these questions are being posed by members of small imaging operations performing less than 40,000 procedures per year who assume that they can only afford the relatively inexpensive PACS solutions offered by the small vendors. In the past, this was probably the case, but that is no longer the case.

Today, several of the biggest vendors in the PACS market, creators of the really big and fully featured PACS, have achieved a scaling feature that allows them to offer effectively the same fully featured PACS at a price point within reach of even the smaller imaging operation.

These vendors have achieved this scaling by reducing the number of servers in the cluster, without eliminating robustness or reliability. They have retained the basic display features, including hanging protocols, but made many of the more advanced display features (like 3D) line item extra-cost options, so they can be added for a modest license fee if needed. They have made many of the professional services that were automatically included in the big system, line item options in the scaled down package. The site that can follow directions and set up their own modality interfaces, complete their RIS interface on their own, and perform their own network testing can save some money.

Perhaps the best feature of this new generation of scaled down PACS is that their upside potential is not artificially limited. If the study volume suddenly jumps by 100%, the small system can be expanded to accommodate growth, without a wholesale exchange of hardware or a whole new tier of software licensing. The user truly pays for only what they need, and only as they need it.

In this scenario, there is no reason for the smaller imaging organization to risk an investment in a fragile company and purchase a bargain-basement PACS with limited features and limited support. It is now possible to afford the economical version of the same PACS being used by the big boys.

Fri, 20 Jul 2007 00:32:48 GMT

GC's Major Guidelines for Picking the Best PACS

Several interesting posts popped up on Auntminnie's PACS Forum today. Two were related to display software for referring physicians, and two others were related to the log-term archive. In my first response , I spoke about the importance of picking a PACS that featured a single display package that allowed system managers to create individual user profiles by assigning display features through user privileges. I also suggested that the Health System should consider providing the display hardware and IT support for their high volume users, because it would be cheaper in the long run than producing and managing film. In a follow-up response, I flat out stated that as much or more attention should be paid to the display software that is going to be used by the referring physicians as is paid to the display software being used by the radiologists. Failure to win over the referring physicians, especially the surgeons, will surely doom a PACS project.

The first of my responses to the archive issue focuses on using a spinning disk solution for the Disaster Recovery subsystem, and the need for some sophisticated Information Lifecycle Management software in the archive that would make it possible to migrate data from media to media and delete data based on information about the study contained in the DICOM header. In the same article, I couldn't help but ask the question why anyone would create an exact duplicate of the original image data, if the PACS utilized any proprietary formats. It seems to me that if you are going to invest good money in a DR solution, the second copy should be 100% DICOM and 100% inter-changeable with another PACS. This would eliminate future data migration costs. In a second response, I suggested once again that the time has come to separate the Archive from the PACS. The PACS vendors insistence on using Private Tags and proprietary encoding is blatant vendor lock. It is expensive (data migrations) and it should be stopped.

So here is my simple Guideline for picking the best PACS

1) Distributed server architecture. Each facility gets its own Directory and Data database servers and there is one shared long-term archive. Each facility is self-sufficient, yet there is one consolidated patient folder. The central shared server "aggregates" all of the information from the facility servers. The user doesn't have to know where to look for any study on any patient in the system.

2) Single master copy of display software, one common GUI, fat client for performance, web-delivered for zero administration. Each user can be granted access to whatever display features and tools they think they need.

3) Software license fee is based on the number of studies under management, NOT the number of users, or the mix of features/tools being assigned.

4) PACS-neutral Archive: guaranteed universal connectivity, ability to morph DICOM Header Tags in order to copy any meta data in Private Tags to Public Tags, no future data migration necessary. If the PACS vendor that ranks the highest in every other category cannot provide this kind of archive, buy the archive from the vendor who can and configure the PACS with a small working cache.

5) Make sure the archive supports a sophisticated ILM strategy, one that migrates data from media to media or deletes data based on information in the DICOM Tags, data transfers have zero impact on the PACS or Archive application server.

There are other important issues and features to be sure, but they pale in significance to these five.

Thu, 12 Jul 2007 20:29:38 GMT

What was the Department of Veteran Affairs Thinking?

I came across a news article today announcing the VA's plan to establish a Disaster Recovery program for all of their Radiology Departments that had already installed the Philips iSite PACS.

"Royal Philips Electronics has announced an agreement with the Department of Veterans Affairs (VA) to provide disaster recovery services that are dedicated to VA users with Philips iSite. Managed and hosted by Philips, the VA disaster recovery services will provide automated backup of all Philips iSite Radiology image data."

It is well known that the Philips iSite PACS stores the image data in a proprietary (iSyntax) format. The Philips PACS is not the first PACS to be deployed by the VA and it probably will not be the last. When the time comes to replace the iSite PACS with something else, all of that study data accumulated over the years will have to be migrated to that next system. That is going to cost both time and money.

A shared Disaster Recovery program is a great idea, but why deploy a DR solution that stores another copy of the study data in a proprietary format? It seems to me that the deployment of a Disaster Recovery solution is an excellent opportunity to create a second copy of the data in a PACS-neutral format. Start copying the historical data already stored in the iSite PACS to a Vendor-neutral Enterprise DR (archive) solution. Call it a "pro-active" data migration. Then continue to store all new study data accumulated by iSite to this Vendor-neutral DR solution.

When the time comes for any of the sites to move on to their next PACS, there would be no need to migrate that site's study data over to the new PACS. A Vendor-neutral archive (server and storage) would be built and loaded with that site's historical data (in a Vendor-neutral format) and then shipped to the site. This local facility server would interface to whatever new PACS is being deployed. The new PACS would not have to be configured with a long-term archive. There would be no need for the time-consuming and expensive data migration.

A Vendor-neutral Enterprise DR solution could also be shared with all those other VA facilities that do not have Philips iSite PACS. What are those sites suppose to do for their DR solution? How many different DR solutions does the VA want to support? Could it be that all VA facilities will be encouraged to upgrade to the iSite PACS? No doubt that's the Philips plan.

Don't misunderstand, I think that iSite is one of the better PACS in the market, but data migration is an inherent problem with changing PACS, in some cases with the next generation PACS of the same vendor (Siemens Magic to Siemens Syngo). It simply doesn't make sense to build a DR strategy that doesn't take into account the high probability that some other PACS will be deployed somewhere downstream, and thus require a sizeable data migration project. A sensible plan would take reasonable steps to avoid that problem.

It should not be a matter of money. Hardware is hardware. Granted, the Philips software license for that second copy of the data is probably less than what the Vendor-neutral Enterprise DR software will cost. But the cost of all those future data migration projects would more than likely cover the premium charged for a Vendor-neutral Enterprise DR solution that could be shared by every VA site today.

I've written a few other posts on this subject that you might find interesting.

PACS-neutral Enterprise Archive - Who will build it?
Looking for a PACS-neutral DICOM Archive?
An Enhanced DICOM Archive would be the ticket!
PACS Vendors think PACS-neutral Archive is crazy idea
SCAR '06 Update

If you would like to have a tool to help you estimate the cost and time associated with your future data migration projects, just email me at graycons@well.com and ask for the Migration Prognosticator.

Mon, 09 Jul 2007 18:02:53 GMT

Pay Close Attention to DICOM Conformance

The typical PACS includes its own long-term archive subsystem. While the all-inclusive package will present few if any data compatibility issues with its own components, there may be serious problems when the time comes to exchange data with other systems. The archive that it totally owned by the PACS application is usually only marginally DICOM-conformant. The assumption of the self-contained PACS is that there is very little sharing of study data with other systems. The PACS may respond to a remote DICOM query with minimal data, i.e. the original image pixel data and little else. Presentation States, Key Image Notes, and other key meta data objects associated with the images and created by the radiologist during interpretation may not be forwarded, because the PACS doesn't treat these as DICOM objects, or it places them in Private Tags, or it uses a proprietary Value Representation (in the tag) to encode the information. Self-contained Radiology PACS are typically very stingy when it comes time to give up their data in a data migration process. The vendors really think of it as their data, and now that the radiology department has decided to leave them for another PACS vendor, the jilted vendor may be reluctant to help in the retrieval and migration of all of the data that really belongs to the health system. This translates to expensive and time consuming data migrations. Study the Archive's DICOM Conformance statement very carefully. Anything less than full conformance for all data objects and SOP Classes should be addressed in the Contract. Build a technically sound, workable exit strategy in advance.

Mon, 18 Jun 2007 22:55:48 GMT

DICOM Tag Morphing - Essential Ingredient in Enterprise PACS Archive

Herman Oosterwijk, principal of Otech Inc., reported his observations of the recent Society of Imaging Informatics in Medicine (SIIM, formerly SCAR) conference that was held in Providence, Rhode Island. In Herman's article, he laments that it was hard to find anything "new". I have heard much of the same, from several colleagues that were also in attendance.

I stopped attending the SCAR meetings several years ago, believing that many of the presentations either focused on updates of old subjects, or introduced concepts that were too impractical for commercial implementation. I decided to break my personal embargo in 2006 and attended the meeting in Austin, Texas. I came away with the same feelings that Herman expressed this year: "nothing new". To me it seems as if there is a major disconnect between the subject matter of the papers in the meeting hall and the subject matter of the vendor booths in the exhibit hall. Perhaps I'm growing jaded, but if an idea is not compelling enough to quickly make its way into a commercial offering, is it worth looking at? At the 2006 meeting, one of the presenters suggested the time was right to replace the mouse with wireless video gamer gloves, sort of creating the diagnostic play station. Well the creators of the Nintendo Wii thought that was a good idea, but...

Actually I did mine some nuggets in Austin, but they were found on the other side of the street. I spent enough time in the Exhibitor's Hall to discover several technologies that would prove to be the next new thing; "The PACS-neutral Enterprise Archive".

In Herman's June 15 story, he reported on an interesting paper presented by representatives from the Mayo Clinic. I'd love to give them credit here, but the SIIM papers are not yet available on-line.

Mayo has apparently constructed a massive central image archive that uses "16 servers for image archiving and provides access to more than 100 million images for tens of thousand of users".

According to the paper, the Mayo Clinic uses "a commercial PACS only for short-term archiving (less than 3 months)". Herman points out that this dual-tier storage strategy underscores "how much information goes back and forth between the image manager and long-term archive". Actually it underscores how little data is going back and forth. Sizing the short-term cache on the PACS is all about matching the age of the priors. In this case, the heaviest recall is obviously within three months.

At any rate, the interesting point made in this article is;

"The presenters commented that the lack of standardization between the PACS image manager and the image archive could definitely be improved."

Well, good luck. There are numerous reasons why even the current PACS do not "play well" with standalone DICOM archives, and I don't believe for a minute that those reasons are technology challenges. I don't see the PACS vendors giving up the Archive any time soon. Proprietary data objects or use of Private Tags simply locks the archive (and the data) to the PACS.

Therefore, the more practical strategy for anyone wishing to pair a PACS from one vendor with an Enterprise Archive from another is to expect that the PACS will continue to store some key meta data objects as proprietary objects, or store them as DICOM objects in Private Tags, or use proprietary Value Representations (the text that describes the encoding of the Tag).

The only way to defeat this particular marketing strategy is to deploy an Enterprise Archive that is capable of Tag Morphing. When confronted with a PACS that does not treat all key meta data objects as DICOM objects stored in Public Tags, the vendor can simply program the Archive to extract a copy of the meta data and convert it to a DICOM object and place it in a Public Tag, before committing it to the Archive's long-term storage solution. The Image Header now contains the original PACS meta data and a copy placed by the Archive in a Public DICOM tag. When the originating PACS requests the data back, the original meta data objects (in their original format) are still where the originating PACS expects them to be.

If a different PACS were to request this data from the Enterprise Archive, there are two possibilities. If the second PACS is capable of reading Public DICOM Tags, the vendor of the second PACS can be given the Group and Element numbers of all the Public Tags where the meta data can be found in the Archived version of the image data file. If the second PACS cannot read the Public Tags, the Enterprise Archive can Morph the Image Header once again and transfer the copy of the meta data from its Public Tag to any Public or Private Tag that is utilized by the second PACS to store this meta data.

DICOM Tag Morphing is an intriguing example of "reverse engineering", but it is not new or a stretch of the imagination. It has been perfected over time by those individuals and companies that have been doing the dirty work of DICOM data migrations. The library of their morphing experience is simply being put to use in the Archiving application. Tag Morphing in the Enterprise Archive is the only practical solution today to deal with the inconsistencies in the way meta data is treated by many of the current PACS. You cannot hold your breath waiting for the PACS vendors to truly standardize.

I've found only three vendors that have deployed Enterprise Archives capable of Tag Morphing: Acuo Technologies, DeJarnette Research Systems, Emageon Inc. If there are others, they haven't made this capability apparent.

I've written a few other pieces on this subject that can be found on this web site.

PACS-neutral Enterprise Archive - Who will build it?
Looking for a PACS-neutral DICOM Archive?
An Enhanced DICOM Archive would be the ticket!
PACS Vendors think PACS-neutral Archive is crazy idea
SCAR '06 Update

Wed, 06 Jun 2007 01:29:58 GMT

PACS-neutral Enterprise Archive - Who will build it?

I was rummaging through some search returns on Google today, when I caught an article written back in October 2006 by Nadim Daher, a Medical Imaging Market Analyst for Frost & Sullivan. What caught my eye was his observation about the makeup of a PACS.

"While the front-end clinical applications, already fairly uniform across PACS vendors, continue to evolve slowly to incorporate more features and integrate information from more systems, the back-end infrastructure of PACS is increasingly being shared across the enterprise. This growing 'disconnect' between the front-end clinical application and the back-end infrastructure creates room for an improved layer of middleware to manage enterprise data including, but not limited to PACS."

The article, well worth reading, goes on to discuss the advantages of building a single shared Enterprise Archive that is capable of managing both DICOM and non-DICOM data objects and therefore capable of being the archive for radiology, cardiology, and other medical imaging modalities.

"Aggregating or better yet, integrating data on a patient-centric basis can constitute a sure first-step towards developing a comprehensive EMR."

While I'm all for looking ahead to the data repository for the EMR, forgive me if I am presently more concerned about building an Enterprise Archive that can exchange data between different PACS. The ability to dynamically translate DICOM header tags in order to accommodate dissimilarities between PACS, would eliminate the need for costly and time-consuming data migrations required every time a PACS is replaced.

There are a number of Enterprise Archives currently in the market that can share their data between different PACS, but most of these systems can only share DICOM data objects, and they assume that each of the PACS are capable of interpreting all of the essential meta data that is stored in the DICOM header. Unfortunately, too many PACS in operation today treat one or more key meta data objects associated with the study data as proprietary objects, or they store these objects in private tags. Few PACS systems can actually interpret all of the data transferred from another PACS, and no matter how DICOM-conformant the archive may be, if it simply accepts, stores, and returns whatever DICOM objects it is given, that archive can not solve the interpretation problem.

Nadim talks about "enterprise archive management middleware", that special software that would "fill in this new-found gap in the structure of the modern-day information system". In my view that middleware has to go beyond DICOM object management and Information Lifecycle Management strategies triggered by patient and study specific information stored in the header. What must be included in this middleware is the tag morphing software that would dynamically translate header information is such a way as to make the data created in PACS A become completely understandable to PACS B, and vice versa. Not only would this make it possible for disparate PACS to simultaneously share a common enterprise archive, it would also eliminate the need for future data migrations.

I have come to believe that what we now think of as a Picture Archiving and Communications System is going to bifurcate into two distinct subsystems, each evolving as they become more specialized. The front-end or clinical part of the PACS would focus on work list management, display functions, and clinical information distribution, becoming the Picture Communications System (PCS). In this new paradigm, the "A" or Archive in PACS is removed and upgraded to become the Enterprise Archive subsystem.

There is considerable speculation on which companies will be drawn to which path. It is perhaps easy to see the bigger established PACS vendors, especially those with modality product lines, continuing to develop the clinical subsystem. They have the army of software engineers required to implement and support the clinical applications. But which companies will be drawn to the evolving Enterprise Archive path? Open systems is rarely seen as an admirable quality to either the big PACS vendors or the big Storage Hardware vendors, and the concept of an Enterprise Archive absolutely demands an open system approach. There are a few specialized medical archive software vendors out there, and I think the time has come for them to step forward. There are early adopters ready to take the right steps towards building the PACS-neutral Enterprise Archive.

Tue, 29 May 2007 22:40:39 GMT

Looking for a PACS-neutral DICOM Archive?

In a news piece released today on auntminnie.com, it appears that PACS vendor Carestream Health and storage vendor Hitachi Data Systems have entered into a collaborative agreement to create a clinical information archive. The collaboration will combine Carestream's Versatile Intelligent Patient Archive (VIParchive) and the newly enhanced Hitachi Content Archive Platform into a single footprint.

This means that three of the top storage solution vendors (Hitachi, HP, and IBM) now have "partnerships" that add the all-important DICOM layer to an otherwise basic storage solution. This latest Carestream/Hitachi storage solution will enable multiple application servers (Radiology PACS, Cardiology PACS) from different vendors to share a common storage solution.

Unlike the DeJarnette/HP and Acuo Technologies/IBM combo systems, this new Carestream/Hitachi entry does not appear capable of modifying the DICOM file formats to accommodate differences between the vendor's use of the DICOM header. In this case the archive stores exactly what it is given and returns the same. If the requesting application server does not know how to interpret the DICOM header created by another vendor's server, too bad! Both the DeJarnette and Acuo software packages are capable of DICOM Tag Morphing, which effectively translates one vendor's DICOM header format into the other's.

The PACS vendor Emageon is also marketing an Enterprise Content Manager (archive) that is capable of Tag Morphing to accommodate any differences that may exist between the way two vendors store metadata in the DICOM header. A plus for the Emageon approach is the fact that their ECM software is hardware agnostic and can be hosted by a variety of platforms and is validated with a variety of storage solutions.

Is the Carestream/Hitachi agreement an indication that the storage wars are taking a new turn?

Tue, 22 May 2007 17:56:29 GMT

An Enhanced DICOM Archive would be the ticket!

I felt compelled to create a response to a post on auntminnie.com regarding DICOM Archives. The author of the original post thought it impossible to combine a DICOM Archive with a PACS and make it work.

"You can't just take a DICOM archive and slam a new front end on it and expect it to work. Each of the PACS has many, many, many database and workstation features beyond the DICOM storage of images that make them work."

I agree, if the term "DICOM Archive" simply means an archive that stores what it is given and returns the same when it is requested.

But an Enhanced DICOM Archive, one that is additionally equipped to perform customized Tag Morphing, would not only work with any given PACS, it would also be able to exchange study data between different PACS. That alone would make such an archive very useful to a Health System that owns multiple PACS from different vendors. If that archive were also capable of triggering a pre-fetch of relevant priors based on the new Order, and an auto-route of the priors to the PACS, such an archive would be invaluable at the time any of those PACS is replaced, because it would not be necessary to migrate the study data. Now we're talking!

Tue, 15 May 2007 19:39:45 GMT

PACS Vendors think PACS-neutral Archive is crazy idea

Tim Gee of Medical Connectivity wrote an interesting blog the other day that picks up on my suggestion to make the Disaster Recovery copy a DICOM copy. He mentioned that few if any PACS vendors will want to talk about such a crazy idea. Of course not! Vendor lock-in is a popular marketing strategy.

I think it's time to start talking to any vendor that is positioning their company as a DICOM Archive provider. Can they assist with the data migration project and guarantee connectivity to both the current and future PACS subsystems? Can they price their software license well under a dollar per study? Building an exact duplicate back-up storage solution for the existing PACS is just a matter of hardware. Building a DICOM back-up storage solution that elimintes the need for future data migrations is certainly worth something on top of the cost of the hardware, but what exactly should that premium be? Something well under a buck a study would make the concept of the DICOM Disaster Recovery solution very attractive.

Thu, 10 May 2007 05:51:10 GMT

What is your Disaster Recovery Solution?

Exactly what disaster do you want to recover from? The most common concern is the loss of a RAID cache. Most medical imaging departments have a plan in place to recover lost image data. At the very least, they have written a copy of the data to a Digital Tape cartridge and stashed that copy off-site in a relatively secure location. Taken one step further, they are managing the second copy on a separate remote storage solution; near-line tape, MOD, CD/DVD; or even on-line RAID. That's great, they have the means to replace their primary copy of the image data, should their primary system suffer a "disaster".

But that second copy of their data is most likely an exact duplicate of the same file format stored on their primary PACS archive. They have two copies of the exact same file. At first glance that seems like a solid plan. But is it?

Most current PACS, radiology or cardiology, are less than 100% pure DICOM-conformant. There is always something special, unique, proprietary about how they create or store the data objects. Just enough to make it difficult to change vendors. In effect, this is the essence of vendor lock.

Somewhere in the future, a Health System is going to want to replace an incumbent PACS with a different PACS. That is going to require an expensive and time-consuming data migration. Even if you know this, and actually have the wherewithal to plan for this and accurately budget for this, this could be considered something of a Disaster.

So here is something to consider. Given that you are mandated by HIPAA to create a reasonable Disaster Recovery solution, why make that second copy a duplicate of the semi, not really 100% DICOM-compliant version of the study data created by your existing PACS? Why not avoid the "disaster" of data migration and make that second copy a 100% DICOM-conformant copy of your study data? What I am suggesting is that you redirect your Disaster Recovery dollars into an second, standalone "archive" solution, most likely from a different vendor, who can demonstrate/guarantee that this second copy is accessible to both the current PACS and the next PACS, and the one after that. There would be no need to migrate the data, when you change PACS!

Make your Disaster Recovery solution a PACS-neutral solution, and avoid the very predictable disaster of data migration.

Fri, 27 Apr 2007 00:30:32 GMT

PACS-Neutral DICOM Archive?

A recent thread on the Auntminnie.com PACS forum was begun with a request for information about a "vendor-neutral PACS storage solution". The thread quickly bifurcated into two distinctly different paths.

One path provided information about a storage solution for a PACS (Picture Archiving and Communications System) that would be vendor-neutral with respect to the actual storage media being used. This is the concept of building a storage solution based on storage media from a variety of vendors. For example a NAS Head (or gateway server) is connected to the PACS Archive Server and the actual storage media (RAID) could be purchased over time from multiple vendors. The advantage of this storage media neutral solution is that the buyer would be able to purchase the storage media on an "as needed" basis and shop the market each time for the best price.

The Second path provided information about DICOM Archive solutions that would be vendor-neutral with respect to the variety of PACS that could be interfaced to it. This is the concept of a neutral, DICOM-conformant Archive that would manage medical image study data over the full course of its lifetime. Of necessity this would mean that the Archive would have a practical lifetime of 10 to 20 years. The hardware (servers and storage media) would be replaced at the appropriate intervals, and the software would undergo routine upgrades, but the study data itself would be managed for as long as legally required or deemed useful. By design this Archive would survive multiple PACS (systems), which are typically replaced every 5 to 7 years.

While the advantage of a media-neutral storage solution is significant, the advantages of a PACS-neutral Archive are compelling. A PACS-neutral Archive would by definition mean the end of Data Migration. Migrating data from the old PACS to the new PACS is nearly always required because there are still significant differences in the way PACS vendors store data objects and preserve key study-related information in the DICOM header. A PACS-neutral Archive would accommodate these differences by translating the study data format of PACS A into the study data format of PACS B, and visa versa. This effectively eliminates the need to migrate the data from PACS A to PACS B, because the PACS-neutral DICOM Archive is the long-term archive managing the true DICOM-conformant version of the study data.

You can read more about this new PACS Paradigm from my postings to this thread, and much more detail is available in the form of a white paper on the subject titled "Building a Better Medical Image Archive". The white paper can be obtained by forwarding your request to Michael Gray at graycons@well.com

Fri, 19 Aug 2005 18:31:18 GMT

The latest buzz in the PACS world, especially in a few Discussion Threads on AuntMinnie.com, is the "web-enabled" PACS. Two of the more significant features of a PACS that would qualify it as web-enabled are communications between client displays and server based on http:// or http(s):// and the image objects are all assigned their own URL. The merits of a web-enabled PACS aside, I have a concern about the assignment of a URL to all of the image objects.

The implication is that all study data objects would be treated as URL objects. This is in fact the way Fuji treats the study objects in Synapse. While I would agree that there are some distinct advantages to "organizing" a database using URL, the problem is when the use of URL supercedes the use of DICOM.

Web tools like http:// and URL are effectively standards, but there is only one universally recognized standard in medical imaging communications and that is DICOM. A lot of effort continues to be poured into defining DICOM objects and DICOM communication SOP Classes. The vendors are being pushed hard to keep up with the advancements in the DICOM standard and nowhere is this more evident than in the area of IHE-inspired DICOM objects and SOP classes like Presentation States and Key Image Notes. To this day there are a number of major PACS systems that still treat key data objects associated with the study as proprietary objects. It's very difficult to extract and migrate non-DICOM data objects to another DICOM-conformant PACS.

My fear is that the adoption of URL as the primary method of managing study related data will de-emphasize the importance of first making all study-related data DICOM objects. There is no way to predict how many PACS systems will be able to exchange URL data objects five years from now. However there is a very good chance that most PACS systems will be able to exchange DICOM objects five years from now.

There may be a number of good arguments for tagging all study related data objects with a URL, but let's keep the pressure on the PACS vendors to make all of those data objects DICOM objects first.

Test Server Shortfalls

Mon, 08 Aug 2005 17:45:45 GMT

I started reading a discussion thread last week on AuntMinnie.com dealing with downtime during PACS software upgrades. While reading comment after comment, I'm thinking about how test servers are intended to minimize that downtime. Then it occurs to me that there are very few PACS systems out there with Test Servers. First of all they are expensive if they are configured to test more than a few modalities and more than a few display stations, Secondly, there are some PACS vendors that do not even offer a Test Server.

A properly configured Test Server will be configured to accept all of the modality inputs and have a small Directory database and sufficient local storage to accommodate the studies that are acquired during the testing. You don't want to add test studies to the primary PACS database. It will have two network interfaces, so it can be used when the primary network in undergoing periodic scheduled maintenance. It will include a HIS/RIS interface component, so that functionality can be tested. It will include all of the display applications. In short, the real Test Server is as close to being a mirror of the production server's interfaces and applications as you can make it. That's the point of it being a Test Server. On the other hand, the vendors should realize that a Test Server is just that, a Test Server. Placing it in service once a year to test a new release does not warrant pricing the software anywhere near that of the production server.

Even though a true Test Server is configured as close to a Production Server as possible, that does not necessarily make it a back-up server. The Test Server is typically configured on a much smaller, less robust server platform than the Production Server. A Test Server configured with sufficient horsepower to be a Business Continuity Sever would be considerably more expensive. A true Test Server and a Business Continuity Server differ primarily in the server hardware. While the concept of a Business Continuity Server is very valid in parts of the country that are at greater risk for natural disasters, most facilities would consider it a luxury. The true Test Server however, should be a requisite in every system.

I'll manage my own storage

Thu, 04 Aug 2005 21:06:47 GMT

With the advent of the FDA clarification that the storage infrastructure of a PACS system is in no sense a clinical device requiring FDA certification (and hence is independent of the PACS vendorís FDA certifications) there is increasing interest in hospitals, faced with a PACS deployment (or redeployment), to investigate, evaluate, and implement a separate storage infrastructure that would be utilized not just for a radiology PACS system, or a cardio-PACS system but as a resource for server and storage consolidation across the entire healthcare enterprise.

Up until recently, if a user wanted to deploy their PACS system on enterprise storage devices (i.e. a storage area network (SAN)) the PACS vendor would need to have certified the storage vendor's equipment by testing their PACS system on the specific equipment to be used. When a user wanted to introduce additional (or new) storage arrays or devices, these too would have to be validated by the PACS vendor. This process, in effect, gives the PACS vendor veto power over what storage technologies a user could employ. Additionally, PACS vendors would typically charge a support fee to provide testing and support beyond those vendors which the PACS vendor chose to remarket with their PACS solution, creating a strong disincentive for a user to consider an independent storage strategy.

With the advent of open-storage NAS gateways, these vendor-imposed hurdles have been jumped. An open-storage NAS gateway provides a virtually infinite NFS/CIFS file space that can be used by various PACS systems (radiology, cardiology, pathology, etc) deployed across the enterprise on virtually any vendor's storage without fear of incompatibility with the PACS vendor's system. In fact, the physical characteristics of the storage technologies are masked from the PACS application. So new storage vendors, new arrays, new technologies can be deployed without disrupting users and transparent to the application (and your PACS vendor).

I think it's time to separate storage and study data from the PACS vendor's proprietary grip.