“PACS is Dead” Revisited.

Ten years ago Don Dennison published an article in the Journal of Digital Imaging (JDI) predicting that PACS would be dead by 2018. Looking back at that article, some of his predictions were overly optimistic, for example, his suggested acceptance of enterprise imaging still has to happen at a large scale, and the emergence of digital pathology is still very much hampered by the proprietary solutions that are being offered and the poor acceptance of the DICOM Whole Slide Imaging Standard. In contrast, the Vendor Neutral Archive (VNA) has taken off, cloud PACS is being deployed at a large scale, and the DICOMWeb implementations allow for universal access to imaging studies from any web-enabled device, including mobile devices. FHIR is still struggling as vendors are hesitant to implement a solution based on a draft standard that is subject to change and new versions are not backward compatible.

However, I would argue that PACS as we knew it 10 years ago is indeed dead, let me explain:

-????????? Radiology interpretation has become a price-sensitive commodity. Administrators shop around for the least expensive service. For example, in the DFW metroplex where I live, most major hospitals have outsourced all of their reading to an external radiology group. The local PACS in the hospital, including its workstations has become obsolete. The PACS administrator’s role has changed from supporting the local PACS to resolving unreconciled studies sent to the radiology provider that are too hard for technologists to fix and troubleshoot connectivity issues.

-????????? Studies are still archived at the main hospital as the radiology groups servicing the hospital will delete them after being interpreted. The archiving is typically done in a VNA and/or in the cloud, although I would argue that few hospitals are yet cloud-based: Remember that even although many vendors claim that more than 50% of their installations are cloud-based; if one considers that the average life span of a PACS system is 6-8 years it will take that length of a time to change the landscape. There is no discussion anymore about how long studies should be archived as storage cost has come down considerably and many institutions hope to capitalize on their image vault to be used for AI training data sets and/or analytics.

-????????? The radiology groups that read the studies for the hospitals use a PACS system but its architecture looks quite different than the traditional PACS as it requires a universal worklist that can distinguish between the various institutions that are being served and it needs to ingest orders from many different sources. There are very few if any traditional PACS systems that can do this effectively which is why several radiology groups implement their version of PACS or heavily customize the traditional PACS systems.

-????????? Two years after Don Predicted that PACS would be dead, Geoffrey Hinton predicted that radiologists would be replaced by AI. This did not happen but most institutions have started to deploy AI applications, requiring sophisticated routing and prefetching through middleware, a dedicated AI platform, a marketplace provided by a PACS vendor, or embedding the AI algorithm in a modality.

-????????? EMR-driven worklists are emerging: In 2014 most PACS systems had a PACS-driven radiology worklist, and the PACS would fetch reports and other applicable information as needed. For a short time, we had a RIS-driven worklist where the RIS would fetch the necessary images for display but the emergence of the EMR with RIS functionality killed the RIS business. Many institutions have the radiologist logging into the EMR to allow direct access to any applicable patient history which then communicates the patient context to the PACS for image display. This context exchange is either proprietary, through the deprecated CCOW standard, or by the new FHIRcast standard.

-????????? Modalities are not connected to the VNA as predicted but rather to a PACS first which is merely a cache as it forwards the images to a VNA. Synchronizing the PACS with the VNA for changes is still an issue as the IOCM IHE profile which takes care of this is still poorly supported by many PACS systems.

-????????? Workflow managers and intelligent routers are emerging, as PACS systems do a poor job with intelligent prefetching of priors which can be at different locations, and images might need to be sent to an AI algorithm either in the cloud or on-prem for processing.

-????????? Modality Worklist (MWL) Providers are all over the map: It used to be that the HL7 orders were mapped into a worklist at an independent broker but as orders can be placed at different systems (EMR, department management systems for non-radiology exams) or not at all (encounter-based imaging) or after the fact such as with POC Ultrasounds there is an emergence of MWL consolidators that provide a single source for all modalities.

Yes the PACS as we knew it in 2014 is dead, it is being replaced by intelligent workflow managers/routers, VNA, and semi-custom solutions at large radiology providers facilitating AI in different locations. It will be interesting to see what the next 10 years will bring.

Herman Oosterwijk