Monday, October 30, 2017

What is the future of PACS?

I get this question a lot, i.e. where is PACS headed? It comes from different professionals, from
people who are decision-makers ready to spend another large sum of money for the next generation PACS, or from those who made PACS a career such as PACS administrators who come to my training and want to make sure that their newly acquired skills and/or PACS professional certification will be of use 5 or 10 years from now.

I also get it from the users who are often frustrated by limitations and/or issues with their current system and from start-up companies that are planning to spend a lot of time and energy in developing yet another and better PACS in the already crowded market place. Where do I think PACS is going and is there still a future in this product and market? I don’t have a crystal ball but based on what I have seen in my interactions with PACS professions, here is my assessment and prediction:

When we talk about a PACS (Picture Archiving and Communication System) in the traditional sense as a “product,” instead of as a “functionality,” yes, the PACS product is indeed the equivalent of a gas-powered car that needs a dedicated driver at the steering wheel,  doomed to disappear in favor of electric, self-driving cars using AI technology developed by Google and others. Just as every car manufacturer is scrambling to get on the bandwagon and change their product development to meet the new demands out of fear of going the same route as Kodak. Similarly, If I were be a truck driver who operates a vehicle mostly on interstate highways, I would be worried about my long-time career path.

PACS systems viewed as a “function,” however, will still be around as the need to interpret and manage images and related information will continue. But, many of those functions will become more autonomous using AI. The Wall Street Journal proclaimed recently AI to be the latest Holy Grail for the tech industry, and there is definitely going to be a spillover to the field of healthcare imaging and IT.

Self-learning systems using algorithms developed by Facebook and Amazon that know which friends or product you might want to follow or purchase next will anticipate your steps and tasks and reduce mouse clicks, anticipate information you want to consult and in what form and presentation (think self-learning hanging protocols) that allow you to become more efficient and effective. This will impact the number one complaint that users currently voice about their PACS, i.e. that it does not support their preferred workflow well.

PACS will give up its autonomy regarding the workflow. In several institutions the workflow is starting to shift from being PACS or RIS-driven to now being the EMR-driven workflow. Unlike PACS, the traditional RIS systems are becoming quickly obsolete. Order entry is shifting to CPOE functionality in the EMR and even the modality worklists are starting to become available in the EMR. Not every EMR, however, is quite ready to incorporate the entire process, consequently there are many holes that are covered with interface engines, routers, brokers, workflow managers, etc. from several “middle-ware” vendors who are bridging the gaps and integrating these systems smoothly. If I were to invest in healthcare imaging and IT that is the niche where I would bet my money.

Another major application for AI will be the elimination of the majority of negative findings from screening exams. Early experiences have shown that AI can eliminate perfectly “normal” mammography images and reduce the images that would need to be reviewed by a person to about 20 or 30 percent of the caseload. Computer Aided Diagnosis (CAD) will also become mainstay for not just the current niches in breast imaging but also be available in other types of exams.

Among the periphery, i.e. at the acquisition side, we will also see a shift as new modalities are being introduced and/or existing modalities are being replaced. Mammography screening exams could be replaced by low cost MRI combined with ultrasound and potentially thermography imaging. We can already look inside arteries and veins using IV-OCT (Intravascular Optical Coherence Tomography) using a small catheter, who knows what we will be able to visualize next, maybe the brain?

Note that this transition assumes a “deconstructed PACS,” of which the core is stripped down to an image cache of a few months with diagnostic viewing stations tightly coupled to this core, and using an enterprise VNA image manager/archive which could be from another vendor, which is driven by the EMR, tied together for now by multiple routers and prefetching gateways. Some of the institutions will opt to archive their images in the cloud, which will become very inexpensive as cloud storage rapidly transforms into a commodity with Google, Amazon, Microsoft and others all vying for your business. If nothing else, the cloud will replace the many tape libraries that are still out there. View stations will become super-fast as solid-state memory will be replacing disk drives, so we will finally be able to improve today’s requirement of a “3 second minimum image retrieval” at a workstation, which has been the semi-gold standard for the past 25 years.

Unlimited image sharing is going to be common practice, CD image exchange will go the way of floppy disks, or the large 14-inch optical disks we used to have for image storage. At my last company we used to take these big optical disk platters and make them into wall clocks, I still have one of them in my office. I should save a CD as well to hang next to it. Accessing information across different organizational boundaries will use webservices much like what you see on an Amazon web page right now. On that Amazon page you can purchase a product from Amazon or an external vendor, which is seamlessly linked.

Compare that with the physician portal, he or she can access the local lab results or jump to an outside lab that provides the lab results in a nice graph, while the image access in the local or remote VNA is also just a click away. And of course, access to many educational on-line resources and good practices are all simple apps on that same desktop, or should I say dashboard, which also displays the current wait time in the ER, number of unread reports in the queue and report turn-around time, in addition to the weather forecast and radiology Facebook share page.

So, do I think that PACS is dead as some people are declaring? In don’t think so, especially if you consider PACS as a function. Just as some see the need for fewer radiologists (think truck drivers?) as a doomed career, but I their functional roles will shift to that of a consultant and the job will be less focused on cranking out reports of which many are “normals” as those will be automated, PACS will continue as an important function in clinical-decision making.

Finally, what about the people who support these sophisticated systems, i.e. the PACS administrators? Their role will shift too, many of the mundane jobs will be more automated and they will be able to focus on re-engineering workflows, planning and solving tricky integration problems. So, the future of PACS is bright in my opinion, but is will be a different color of bright, and as always with transitions, there will be people and companies that anticipate and embrace these changes, and others that will have blinders on and be left out.