Workflow integration is the game-changer for medical imaging AI

Why PACS integration benefits radiologists and how to ensure it

color blob

The potential extra burden of AI

Medical imaging technology is increasingly sophisticated, for a large part driven by the growing suite of artificial intelligence (AI) applications. Yet AI’s potential to improve diagnosis speed and accuracy may be overshadowed by possible hurdles it may create for physicians.

A radiologist having to exit their regular working window (or even room) to log into a new system and upload a scan, then wait for the analysis, and finally sit down for reporting, may not experience the time gain and satisfaction of AI-powered decision support.

During the current covid-19 pandemic, multiple AI models are being built to address clinical challenges. In response, The Royal College of Radiologists (RCR) released guidelines on the implementation of these AI solutions for medical imaging analysis. The document sets the standard of integration of AI in the reporting workflow “in such a way that it does not add extra burden to radiologists.” In summary, RCR took the following stand:

“AI image pre-analysis is likely to have a very positive impact on radiologists’ future working lives if properly integrated into the reporting workflow.”

Workflow integration of medical imaging AI is essential – not only during the corona crisis. In other words, a well-performing AI model may not be sufficient to deliver value to healthcare practitioners. Why is that – and how does integration work exactly?

Note: The RCR recommendations focus on the radiology information system (RIS) and the picture archiving and communication systems (PACS). The radiology workflow further presents additional technical components. To make this article specific, we look at PACS integration.

Benefits of PACS integration for the radiologist

Radiologists are understandably reluctant to leave their PACS window and even get used to another user interface for reporting scans. Interruptions take time, decrease efficiency, and cause frustration. For an AI solution to provide diagnostic decision support, without becoming a burden, it must integrate into their existing protocols, systems, and workflow.

An AI tool that is well-embedded into the pre-existing infrastructure offers these benefits:

No disruption

The AI results are ready and visible within the original diagnostic series before the radiologist starts reading the study. There is no need to manually upload or forward studies in order to get them analysed. The benefit here is threefold:

  • Increased efficiency, as a result of the reduced waiting time. The time gain is, in fact, the main appeal of AI for many radiologists.
  • No bias of a radiologist deciding which scans to analyse with AI.
  • Reduced workforce requirements, as there is no additional planning or administrative staff required to find the relevant scans and upload them to the separate AI station in time for the reporting session.

Multiple users

The results are accessible to the entire radiology department. The chest, for example, is part of several imaging pathways. Thus, findings such as lung nodules may be encountered by different radiologists, general or thoracic. Having the results of a chest-focused AI solution available to the whole team supports collaboration in this case.

Remote access

A well-embedded AI tool enables remote working. The team members who have access to the PACS also have access to the AI solution, on and off-site, during and outside work hours. Whilst this benefit has been expedited up the priority list due to the corona crisis, the demand is set to increase. The high radiology workload and the evolving types of tasks that do not require a physical presence in a hospital (e.g., reporting on screening scans) are strong arguments in favour of more flexible working, such as from remote locations.


If enabled by the PACS vendor, integration allows the radiologist to interact with the AI results in the PACS viewer. Now, this requires manual dictations, but with interactivity on top of the diagnostic series, radiologists can modify the AI output  – for example, by adjusting the provided measurements – to automatically update the reports. We expect this PACS capability to improve in time, as it presents an incentive for increased PACS-based reporting.

The creation of a robust AI model alone cannot check all the boxes above. The next section outlines the main considerations for delivering these benefits.


Achieving seamless integration

Software integrations are sometimes referred to as ‘seamless’. Technically, the term describes the type of addition to a system that does not cause errors or complications. Think of online payment gateways. If you enter your details, click ‘Pay’, and you’re done, the payment application is seamlessly integrated into the website. If you are sent to another page to complete the payment, it is not.

To create a smooth integration of an AI solution into the radiology workflow, two aspects are crucial:

A thorough understanding of the radiologists’ needs 

This is possible by working with a network of radiologists and involving them in all main stages of the development process.

  • Before the development of the AI solution, radiologists can help define the need for a clinically useful feature and the way to fit it into their workflow. If the aim is to improve their overall efficiency or provide highly accurate results to support their diagnostic decision, we must first understand how radiologists are currently working and what their reporting setup is. This understanding must also apply to diverse settings, varying per country and type of hospital.
  • During the development process, radiologists provide invaluable input for designing a user interface that is easy to use and complete.
  • After deploying the AI product, the feedback channels must remain open. Real-world use can always enrich the research or test setting in which the product is built.

Careful planning and collaboration with PACS vendors around deployment 

We recognize the dependency on the PACS vendors facilitating the integration process, which, depending on the company, may become a limitation. Teaming up with PACS vendors is a prerequisite for success. Beyond the technology, this requires solid organisational capabilities from the side of the AI vendor. (For more info, one of our go-to sources for insights into PACS and AI is the ‘PACSman’).

An example of seamless PACS integration: Veye Lung Nodules

Workflow integration is a key consideration in the development of our AI solutions. Veye Lung Nodules, our pulmonary management assistant, detects and quantifies even subtle abnormalities on chest CT scans, without disruptions for the radiologist reporting on the scan. The process of Veye Lung Nodules delivering its results reflects our approach to seamless PACS integration:

  1. The study of the patient is acquired and sent to the PACS.
  2. Veye Engine queries the PACS and automatically retrieves new studies, and, where available, the most recent prior.
  3. Veye Lung Nodules analyses each retrieved study in the background.
  4. Results are sent back to the PACS, into the original study, using the DICOM standards.
  5. The radiologist starts reading the scan, either as an additional series or through the GrayScale Presentation State (GSPS), if supported by the PACS. The latter provides the option of toggling the AI results on and off.
Veye Lung Nodules workflow integration
Veye Lung Nodules workflow integration

The ‘umbrella’ engine

Optimal workflow integration can speed up the adoption of AI systems and, ultimately, solve big clinical challenges and contribute to better healthcare. But as the AI offering grows and hospitals are assessing multiple AI solutions, is integration sustainable? Historically, the increase in the number of technological developments for radiologists resulted in the creation of multiple systems, which must be linked, simultaneously open and require separate log-ins.

One way to minimize the challenges of adopting several AI applications is by deploying a single ‘umbrella’ engine, with access to the various AI modules, and integrated into existing workflows. We are not quite there yet. However, at every level, seamless workflow integration matters. Perhaps more than the AI.



About Catalina

Catalina Barzescu is Content Manager at AidenceAfter graduating from the Erasmus University Rotterdam with a master’s degree in media and journalism, she made her way to MedTech via e-learning, travel, and humanitarian aid. Her interests are storytelling and data science with a purpose. The articles she signs reflect the work and ideas of the various teams at Aidence – medical, tech, regulatory, and business.

Connect with Catalina on

Join DeepHealth at the National Congress of the Italian Society of Medical and Interventional Radiology to learn more about our portfolio. Book a meeting


Book a demo

    * Required fields.
    For more information on how to unsubscribe, our privacy practices, and how we are committed to protecting and respecting your privacy, please review our Privacy Policy.
    This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.