The annual gathering of healthcare imaging and IT professionals, i.e. SIIM 2019 in Denver kicked off with a moving story by the keynote from a patient, Allison Massari, who survived a life-threatening accident that burned over 50 percent of her body. Her story of the impact healthcare providers had on her recovery set the stage for hundreds of healthcare imaging practitioners, consultants and vendors to exchange their experiences and gave added meaning their professions before talk turned to their products and services, and education of their peers on what is new and what is coming. The meeting had good “vibes” as people were eager to learn and there was excitement about new developments.
Here are my impressions:
1. AI is over its initial hype: the initial fear factor that came with the hype of first AI applications that made radiologists anxious about the potential impact on their jobs has faded and it is becoming obvious that there is still a lot of work to do and a long way to go.
Most AI companies don’t even have FDA approval for their products yet, even though the FDA is stepping up to the plate and is giving special considerations to the fact that many of these products are based on deep learning, whereby the behavior of the software might change over time.
This infographic provides a nice breakdown of FDA approvals over the past several years showing the percentage of radiology algorithms that were approved. AI is finding its way in some of the PACS applications starting with workflow enhancements, there are dose reduction applications for CT screening and some “low hanging fruit” surrounding detection of common diseases.
2. Enterprise imaging is still very challenging: As Jim Whitfill, the current SIIM chair mentioned during his update, enterprise imaging is what most likely saved SIIM from its demise after the 2008 downturn in membership and conference attendance, as IIP professionals were starting to think about how to do enterprise imaging and subsequently publishing about it in the Journal of Digital Imaging.
The VNA or Vendor Neutral Archive became the vehicle to implement enterprise imaging solutions, however, the non-order (aka encounter-based) workflow for those non-radiology or cardiology departments is poorly defined and there are many different options. See my related post in which I identified more than 100 possible implementations. Talking at SIIM with several implementors, I identified three different strategies:
· The "top-down" approach – This model implements a vendor-neutral archive (VNA) for radiology and/or cardiology first, and then starts to expand it with other departments, however, there is no single, uniform workflow for those departments resulting in many different options.
· The "bottom-up" approach – This model, which was used at Stanford University implements a VNA beginning with one department and then adding in other departments using the same workflow (which is DICOM worklist based). After adding many other specialties, they are only now starting to add radiology and eventually cardiology.
· The “hybrid” approach – This method, which was adopted at the Mayo Clinic, is a combination of both approaches, instead of having many different workflows or only one single workflow, they settled for only a handful, in this case five major workflows for the different departments. You can see details of this discussion at this short video clip.
3. Teleradiology workflow is very challenging: there are only a few PACS vendors that do teleradiology well, as a matter of fact, many teleradiology vendors build their own system as the requirements are so different:
a. The turn-around time requirement is very challenging – a typical turn-around time has to be 5-10 minutes for trauma cases. This means that the workflow is super-optimized.
b. AI can make a major impact – Hanging protocols are very hard to define as the source for these studies vary widely, some of the studies group all images in a single series, some in multiple series, and the series descriptions are not uniform, therefore, a simple algorithm determining which is the PA and which is the lateral chest and ordering them consistently saves a few mouse clicks which is time. Prioritizing studies based on certain critical findings is important as well. AI definitely assists in the efficiency and automating of repeated tasks.
c. There is a lack of patient contextual data – There are many challenges to get the prior images for a particular study (see a renewed activity described below) as the use of CD’s for image exchange does not seem to be going away soon. But this workflow is well defined by IHE XDS-I and other profiles, and in many countries other than the US there are successful image exchange implementations based on standards. However, instead of a radiologist logging into an EMR and looking at the images while having the other patient context at their fingertips, a teleradiologist logs into a PACS seeing the image, wanting to have that patient context from potentially many different EMR’s. It is a “reverse” workflow, instead of being EMR-driven pulling multiple imaging studies, it is PACS driven wanting to pull multiple EMR documents. This is a new challenge which is not quite addressed yet; ideally one could maybe pull CDA’s from these EMR’s but those were really defined for a different purpose.
d. The workflow is reversed – the traditional Order-Study-Report workflow looks different for a teleradiology application as in many cases the order comes after the fact, so it would be Study-Report-Order (including “reason for study”)-Report update. Interestingly enough, when talking with teleradiologists, they only have to adjust their report based on the “reason for study” in a few cases. Regardless, this workflow needs to be addressed by their PACS.
e. Many studies, if not all, are “Unverified”– This is particularly true for battlefield and disaster applications. There is often no patient name (“civilian 1”), and definitely no patient ID, and it is not uncommon to have partial studies. A PACS that depends on the traditional order-based workflow will perform very poorly.
4. CD’s are here to stay (for a while): I do have personal experiences (as many do) with image exchanges for me and my family as witnessed by the stack of CD’s I carry to doctors and specialists. Actually, as some of them lack CD readers on their laptops or have their computers locked down by their security departments, I carry a laptop with me with the images preloaded and ready to be viewed. My experience with my veterinarian is completely different. When I asked for a copy of the MRI of our dog on a CD from our neuro-veterinarian, I was told that it is “old-fashioned” but that they would be more than willing to send me a link to view the images in a viewer, or, alternatively allow the images to be downloaded as a zip file for me and my regular veterinarian to review, which I did. How is it that our veterinarians have this all figured out and our physicians don’t? I can come up with many reasons, but one of them was identified by a special ACR/RSNA committee which met during SIIM and that is the lack of a standard governance agreement. Instead of having to get BA’s from all your partners covering the HIPAA requirements, they recommend a standard document as part of the Carequality consortium, in the form of an implementation guide, which is available as a draft for public comment. In the CareQuality framework, 36 million documents are exchanged each month using 16 networks based on IHE XCA standards. If we can exchange documents, there is no reason to not exchange images.
5. Cybersecurity is a hot topic: there is not a day or week that goes by without a report of yet another ransomware attack or security breach exposing literally millions of patient records. There have been reports of CT scans modified to create significant findings using the DICOM header preamble on CD’s to embed viruses on old devices that still run old OS that are not being patched anymore (note that Windows 7 support stops in January 2020).
Key safeguards include upgrading old OS’s, if that is not possible, then isolate them from your network as well as disable the USB’s (which is a problem by itself as several modalities depend on the USB to connect ultrasound, dental, or other wands and detectors), secure networks, and educate your employees on the danger of social engineering is critical. At one facility, the open rate of spam emails dropped from 80% to less than 20% after the IT department started to send out “bogus” spam emails to alert their employees to the danger of social engineering. Another great example of this phenomena was that of a (infected) USB drive that is dropped in the employee parking lot of a hospital with its logo on it so that an unsuspecting employee with good intentions will insert it in a hospital network computer resulting in great harm.
6. New standards are available to provide greater interoperability: DICOM, FHIR and IHE have made several new additions which are covered my SIIM report part 2.
Overall, yet another good year for SIIM and its members. The major differences between SIIM and other mega-meetings such as RSNA is the fact that you can cover the exhibitions without having to walk (and often run) many miles in between different booths, you have much better access to many of the faculty and peers, and last but not least, there are an abundance of hands-on workshops to experiment with new tools and standards.
For example, at the XPert IIP workshop, attendees could learn troubleshooting DICOM headers using DVTK and the DICOM protocol using Wireshark sniffer using pre-loaded laptops provided as part of the training. Sessions covering DICOMWeb and FHIR hands-on experience as well as the IIP sandbox covering Mirth interface engine programming were also very popular. One of the themes this year was empowerment, what better way to empower users than by providing them with the skills and tools to do their job better and more effectively.
Next year will be in Austin, which is closer to the OTech home base (Dallas, TX), I am looking forward to another great meeting next year!