Whatever happened to Enterprise Imaging (EI?)

A few years back, Enterprise Imaging (EI) was the “hot topic” at the annual conferences, which was before AI came into the main spotlight. You could ask yourself what happened with EI, did the problem get solved, are we there yet, or is there still work to be done? Let’s start to investigate.

Enterprise imaging is complex; it is hard to define, witnessed by the fact that the HIMMS-SIIM enterprise imaging Workgroup needed several white papers trying to break down its complexity. So, if it takes a group of very smart people many papers to come up with a definition, you can imagine that it is hard to come up with a simple definition, but I’ll try anyway:

Enterprise Imaging provides access to all medical imaging and related information in a digital format from anywhere to anyone in an enterprise.?

An important question is “What is an enterprise”, which could include a single hospital, a related group of institutions such as the 100+ hospitals of the US Department of Veterans Affairs (VA), or big HMOs such as Kaiser Permanente, which includes all of their thousands of affiliated physicians and patients. It can also include a geographic entity such as a metroplex, state, province, or even a country, connected through one or more Health Information Exchanges (HIE) which provides the registry for the imaging sources that are archived locally or in the cloud.

A picture of an enterprise imaging system is shown, which was published by the HIMSS-SIIM working group 8 years ago.

The figure shows its input imaging sources on the bottom and the top shows the source for orders and patient demographic information as well as the destination for results and presentations in an EMR and viewers, whether embedded and/or launched within an EMR or as a stand-alone.

What are some of the roadblocks that have prevented widespread deployment of EI so far:

1.?????? Lack of standards support by non-traditional image sources: It is easy to take a picture with a digital camera or smartphone but to make sure that the picture is labeled with the correct patient demographics, uniquely identified, tied to the right patient, and connected with an order for billing and tracking purposes is challenging. We need simple apps on these capture devices that can acquire the needed demographic and ordering information and upload the images using standards, preferably DICOMWeb protocols. IHE has defined profiles to address these encounter-based imaging protocols, the industry just has to start supporting these.

2.?????? Lack of standards support among digital capture systems: Digital Pathology is a good example and is probably the largest source of proprietary formats. According to some of the pathologists I talked with, they have to deal with up to 10 different image formats. Digital slide scan manufacturers try to lock in their customers with a proprietary format requiring them to use their infrastructure and image archiving and viewing system which works if a lab only uses a single vendor and does not have an intention of sharing images through an enterprise. There is a DICOM standard for whole-slide pathology imaging but it is still poorly supported.

3.?????? Scalability issues: I recently worked on an RFP for an enterprise viewer for a large enterprise US hospital which included 60,000 licenses for each one of their affiliated physicians. That number is for today, who knows, in another 10 years it might be 100,000+ licenses. This does not include the patient imaging portal which could easily be a million+. We see the same large scale in some of the government-managed healthcare systems such as Singapore, Saudi Arabia, or the Canadian provinces. Appropriate infrastructure, availability, and backup, combined with the local requirements for patient privacy and security have to be provided which is challenging as many PACS and/or VNA’s were never designed with that kind of scalability in mind.

4.?????? Legal issues regarding patient private information: As many of these systems are being hosted in the cloud, there are national rules and regulations to follow. This has been a significant barrier in smaller countries, especially for LMIC countries. I have encountered this in several African countries while deploying a cloud-based PACS system as they want the cloud hosting company to base their servers in the local country which is not always an option.

5.?????? Infrastructure challenges: If you look at the figure, you’ll see that the center part calls for an Enterprise Image Repository, Interface Engine, and Modality Worklist Services. The Enterprise Image Repository function can be taken care of by a VNA, assuming it has full functionality to support an enterprise deployment (scalability, tag morphing, unique patient indexing, etc.). As of today, several vendors can provide this functionality. Concerning the Modality Worklist services, these are often provided by the EMR, or, if a sophisticated set of rules are needed to be configured to provide the right worklist at the right modality, one can get this from a third party. The third component is the interface engine. These are mostly provided by third-party middle-ware software providers offered under various names (“Router, Orchestrator, Conductor, Workflow engine, Platform”). Its job is to get the images and related information to the right destination which often includes historical data. Most recently its job includes routing images and related information to and from an AI platform or individual AI algorithms as well.

In addition to the technical barriers to deploying EI mentioned above there also seems to be a lack of incentives, a lack of familiarity, or maybe the fear of the unknown by a typical physician. As a case in point, I always take my laptop with my imaging study to my physician and show it to him/her and only rarely they are interested in looking at it, let alone they want a copy (they often don’t have a CD reader anyway to upload the images). At my last visit, my physician told me “Just give me a copy of the report”. There is still some work to be done, maybe we need a new generation of physicians to appreciate the added dimension of “seeing is believing” instead of relying on written text. Especially when soon this written text might become autogenerated using ChatGPT which might or might not reflect everything there is to see in an image.

In conclusion, don’t let the AI hype distract you from focusing on deploying enterprise imaging. There are still some barriers to overcome, some of them mentioned above, but the result will be worth it.

?Herman Oosterwijk, The AI-Guy