Monday, December 01, 2014

Meeting With The PACS Giants And Other RSNA Tales

There is a touch of melancholy for me here at RSNA 2014 to go with the 20 degree nip in the air. I'm not one to dwell much on the deep meanings of beginings and endings, but while strolling the exhibits today, I realized that I've been attending this monster of a convention on and off since I was a Nuclear Medicine Fellow in 1990. And it occurs to me that since I'm now semi-retired, it is possible that I won't return next year. But we'll see how that goes.

One of the joys of RSNA, and my fame, or at least notariety in the field, is the chance to meet up with those far more promienent in the field than I. Hence the title of this piece. I had the wonderful opportunity to share a cappuchino with two giants of PACS, Mike Cannavo and Dr. David Clunie. Mike I've known for years, but I had only communicated via email and AuntMinnie forums with David. I was very fortunate to get both of them together on the same couch for a few moments today. These two gentlemen have been involved in the business since before anyone could even spell PACS. They both have an amazing level of knowledge, not to mention various documentation, of those early days, and I'm urging them to collaborate on a book.  Maybe I would qualify for a footnote...

I am occasionally accosted, I mean greeted, by some of my loyal readers. In fact, when I stopped at one booth to say hello, a friend who was mentioned in an earlier post and was apparently embarassed by the fact that it proves he's one of my readers spotted me and exclaimed, "It's the Dalai! Shall I kiss your ring? Shall I kiss your ass?" To which I replied, "Not unless someone gets it on camera!"

Today was my informatics day, and I heard some talks about portable platforms, and SOA's, and image sharing. SOA's, aka Service Oriented Architecture, as presented by Dr. Paul Chang  are fascinating constructs with huge potential. Dr. Chang showed an example from U of C wherein the SOA determines if the patient has appropriate labs ordered, and if not, it initiates the order to acquire them. So much for us humans.

I also stopped by a booth or two.  (This is starting to sound like a third-grader's rendition of his trip to Disney World, but for me, that's quite appropriate.)

In my feeble-minded semi-retired state, I've decided not to continue my practice of posting every last little detail about demos and things. You really have to get your own hands on the software (well, the keyboard and mouse, but you know what I mean) to determine if something will work for you or not.  I do pride myself on attempting to wear multiple hats when I evaluate a program, which I think is the key to my success as my group's CTO as well as the premier radiologist PACS blogger (still the only one, but I'll take it).  I'd like to think I can make any commercial PACS client work; that's my ex-engineer hat in action. But I think I'm reasonably good as well at figuring out if something will work for my least-technically-savvy former partner, the one who calls from airplanes wanting to know how to adjust the volume on his laptop. So, in my new, lazier, partially-retired personna, I'm just going to sketch out the very basics and leave the picayune details for another time.

Here we go. I stopped at TeraRecon, and had a look at their latest offering, deconstructed PACS, which basically utilizes TR as a PACS overlay, uniting data from multiple data silos (coming from the Midwest I'm not used to anything other than corn and grain being stored in silos) and adding in the magic of advanced processing for a sort of super enterprise PACS.  From what I could see, there are still a few details to be worked out before the ssytem will work as I would want it to, but the TR folks are on their way. Ultimately, the overlay will require the ability to check for priors in all the silos (which they seem to have almost mastered) and be able to talk back and forth to the underlying PACS to manage workflow, which seems to be on its way. I was most amused and honored to be treated with equal deference to the chairman of a very well-known radiology department who was there at the time. The chairman had actually heard my Laws of PACS talk a while back, and urged me to keep up the good fight. And so I shall.

To be scrupulously fair, Visage has a similar approach to overlying PACS with an advanced imaging platform, but I ran out of time before I could see their latest. Apologies to Sam and Brad. I'll look at it ASAP. ***

I should break off into a separate post, but the following entry will be fairly short.

About three weeks ago, Agfa placed a test version of IMPAX 7 Agility PACS in our reading room, and I was able to have a few hours of playtime with it. I had promised not to report anything until talking with the important people at Agfa, and I usually honor my promises.  As a followup to the home test, I met with some Very important people in a spartan back room of the lavish Agfa booth. You would think that Agfa would not be happy with me, given the rather brutal treatment I've given them over the years. You would be wrong. Agfa has always been gracious in accepting my acerbic criticism and improving where possible. Agility is no exception, save the fact that I didn't really have to criticize as much as usual. Gone (FINALLY) is the tool-toggling I've whined about for years.  Available (FINALLY) is workable user-level hanging protocol creation. And so on. I had some complaints/observations about the way the latter worked, and some of my ideas had already been incorporated between my two recent exposures, and others hopefully will appear soon. There is very tight integration with the "top three" vendors of things like advanced processing and nuclear medicine. For example, my Segami Oasis will come up within the PACS viewport as if it were part of the PACS itself. For better or worse, the port basically reverts to the incoming programing, mouse-controls and all. Could there be a more unified approach to this? I'll have to play some more. Agfa has utilized hot-spots on the image for common controls like window-level, an approach I'm not fond of, for what that's worth.

Agility is considerably different than IMPAX 6 (once code-named Odyssey). It is a worthy successor, and frankly is somewhat more mainstream in operation and appearance than 6. When asked how I would grade it, I said that with the current improvements, I would give it a B+/A-. It has a way to go, but it does represent a significant step in the right direction.

As always, more to come!


I wandered by the Visage booth on my way back to the educational sessions from my $20 mediocre lunch, and I stopped to see my good friend Brad. Given the 10 minutes I could spare, he and the apps folks managed to give me a quick but thorough view of the latest version. It is impressive, all the more so to realize that the system operates with server-side rendering. This allows platform neutrality (it will run on my Mac, iPad, etc.) and really rapid loading of huge datasets since they don't actually go anywhere. Visage has outfitted a Very large healthcare operation with its version of a Deconstructed PACS, operated from a single main server (of course with failover backup) and six rendering servers. Brad tells me this configuration can handle tens of millions of images and hundreds of simultaneous users and still be at only 20% of its capacity. 

Visage has some very nice features such as a lesion marking function that gives volumetric information as well as orthoganal dimensions, nice for RECIST reporting. In its PACS implementation, Visage can dive into silos and match exams, and has a better hanging protocol than a new PACS version I've examined recently. It can handle all modalities, and even can produce MIPS from breast tomosynthesis, something I haven't seen before, altough it won't make me want to start back reading mammograms.  There is of course very powerful advanced imaging as well. 

I do have to point out two deficiencies, which Brad tells me nicely are only problematic for old, senile vacuum-tube loving knuckle draggers like myself (OK, he didn't say it that way, but I don't want you to think I have any sort of elevated opinion of myself). First, the level of automation of things like coronary vessel segmentation is limited. Visage's philosophy is that automated detection is not perfect, and the human eye may better detect a more aberrant vessel path. That's probably true, but I do like the joy of one button operation. (Anyone remember the line from the Lost in Space Movie where Major West launches the Jupiter 2 by saying "...And the monkey flips the switch...")

Secondly, this deconstructed PACS is designed to be driven by an EMR/EHR, speech recognition, RIS, etc. What Visage has declined to provide, and Brad says I don't need, is a worklist! Here we disagree. I come from a PACS driven workflow shop, and I like it that way. Apparently there are third parties who can provide a worklist, but I still think Visage should write their own. I'll be glad to help. 

And to my friends at Merge, please don't worry about the omission of your booth. I walked by a few times and you have all been very busy, hopefully with paying customers. I promise to review the updated PACS and other offerings online with you at a later date. Two days at RSNA is just not enough!!

No comments :