by Michel Rozpendowski
Operations Engineer at Osimis
The potential for artificial intelligence to streamline processes and deliver a seamless dataflow is increasingly being assessed by Radiologists and Clinicians for their day-to-day work. With more and more AI companies appearing on the market, there is greater choice and more accessible functionality. AI in radiology is not only a buzzword, it’s fast becoming an indispensable tool for the radiologist.
While the many benefits offered make AI attractive, the integration of associated workflows can be complex for the management of the IT infrastructure at hospitals or other health service providers.
As with the introduction of any new technology, there are some considerations that need to be explored and addressed.
Let me share a roadmap to illustrate how adding yet another tool to the radiological workflow can be a smooth process AND secure that seamless dataflow healthcare providers need to focus on patient care.
Choosing the right AI algorithm is crucial. While the clinical value should be determined by a radiologist or specialist (but that makes for a whole different article, like tis one from Prof. Dr. Erik Ranschaert), it is essential that the PACS or IT admin is part of the journey from the beginning too. It will be their duty to ensure the smooth and efficient integration of the workflow.
The perfect AI algorithm is not just the one with a proven clinical value. It should also meet the institutions’ requirements for IT security, GDPR compliance and efficiencies in deployment and usage. In my experience, these are the key points to consider.
On choosing your parameters
First and foremost, you have to consider the seamlessness of the integration (from the PACS back to the PACS/RIS) and its ease of use. Radiologists and doctors dislike having to get used to endless new IT tools, so it is essential to take the pain points out from the get-go.
Once the strategic decision to adopt AI workflows in the hospital is made, PACS and IT admins should encourage the deployment of a vendor‑neutral AI infrastructure with a fully transparent integration.
Such integration should respect two criteria:
The simple but significant advantage of the PACS integration is the user experience of the radiologist or clinical specialist remaining unchanged. Exams are visible in the PACS, on the workstation or via the usual viewer. The results provided by the AI are added as an extra series or PDF report to the original exam so, there is no need for new learning with the latest software tool. No need to overcomplicate serious matters. Try to keep the complex simple.
Network security and the transmission security of personal data, in line with local regulations (GDPR), are an important aspect for investment in any software solution deployed in the cloud. While some AI algorithms could run on an on-premise server, most of them are cloud based and need to respect the principles of GDPR.
The most obvious implications are:
Managing GDPR compliance and IT security with an increasing number of software vendors could become quite cumbersome over time. Installing a vendor-neutral complaint hub while addressing a variety of radiological workflows would give the IT and/ or PACS admin peace of mind to best serve their internal customers in their day-to-day routines.
Important, too, is the flexibility and openness of the system. Each institution is different and has its own workflows. One could need a specific filtering of the studies, another could require some notifications (HL7, FHIR, email, ...) or specific actions on DICOM tags before AI processing.
In addition, a single component that makes it possible to regulate different workflows (i.e. different AI providers) would save time for the PACS administrator during configuration, but also for the data protection officer and the purchasing department...
The monitoring of all those radiological and AI workflows is another key point.
Reliability is maintained and every step of the process can be monitored if necessary. This also supports efficient troubleshooting in case of problems. A dashboard functionality allows users to monitor data transfers and track user metrics.
AI brings the need to fetch DICOM data out of the hospital. Why would you see that as a challenge, when you could consider it an opportunity? It could maximize the return on investment in an appropriated IT infrastructure that does not only cover AI workflows but also covers other challenges within the modern radiology workflow, such as:
“A blockquote here that’s worth highlighting. Go on, I dare you to not read it.”
The solutions provided through the OSIMIS platform allow automation and streamlining of the radiology workflow based on DICOM standards. Our integrated modular workflow engine allows it to easily communicate with third parties, such as RIS, HIS, EHR, AI algorithms or other tools deployed at or outside the institution. We want to build and be the bridge between the medical images' server (“PACS”) and other relevant software applications, such as third-party AI algorithms, so our users can manage, monitor, and share transferred data, fully secured and GDPR-compliant.
Whatever the chosen solution, the comfort of the doctors should be kept at the forefront of any considerations. If the tool is simple to use, reliable and easy to troubleshoot by the PACS admin, workflows will be more efficient and patient care will improve. Smarten up health tech for better healthcare – the technology is ready to serve the higher purpose.
