VISUS Health IT • 08.05.2024 • Stories

PACS as a data warehouse

  • The PACS as a data repository

The volume of data in radiology is already enormous. A cranial scan 20 years ago needed just 25 images—today it needs between 120 and 5000 images. This potentially increases the quality of the finding but pushes radiologists to the limits of their capacity in terms of time and attention span. If non-radiological data relevant to the finding are also added as well as data from AI, information management—the core task of radiology in the treatment process—becomes a major challenge.

One that can only be solved in the long run with the help of smart software. Radiologists are increasingly viewed as the data brokers of medicine, distributing treatment-relevant information to the right clinical locations. To be able to do this diligently, they must also be able to read, understand, analyze, and distribute the data. And to stick to the same metaphor: It needs a digital data warehouse that structures, summarizes, presorts, and makes legible existing information so that radiologists can complete their tasks.

Combining image and finding
Such a warehouse should in future be even more powerful with our JiveX Enterprise PACS. Our strategic aim here is to incorporate and display images and other case-relevant information—from nuclear medicine, AI systems, or wherever—in the PACS so that radiologists are helped in the reporting process with useful additional knowledge. Radiological images, results from non-radiological examinations, and AI assessments should also be linked to the finding.

And even though the linking is done in JiveX itself, any additional data useful for the finding are not generated there. At least, this is not essential. The example of AI illustrates the interplay very well. When using AI for the radiological finding, we rely on a close cooperation with specialist providers, whose software is so deeply integrated into JiveX that automatically exchanging images and findings is possible. In concrete terms, radiological images are sent directly from the PACS to the AI, which in turn automatically returns its results to JiveX. The results are then directly integrated—depending on their meaningfulness—into the image used for the finding. The radiologist can, for example, be made aware of an available finding as soon as they open a study. This speeds up the diagnostics process, and potentially makes it safer, because the risk of overlooking something is greatly reduced by AI.

Standardization as the foundation for integration

So that such a scenario works in practice, it comes down to one thing for the AI solutions: The software must be based on conventional communication standards so that it can be deeply integrated into JiveX, which then acts as the primary diagnostic tool. What is also relevant is which diagnostic questions can be answered using AI. The situation on the AI market is currently such that manufacturers have often settled on easily answered but not necessarily frequently asked questions in radiology. This includes breast, thorax, bone, and prostrate. This focus on low hanging fruit is yet another limitation for the use of AI in radiology; the range of applications will be expanded in future. It can be assumed that AI will soon become an indispensable assistant in radiological diagnostics.

AI-based language models will also be of use in helping to overcome the limitations of auditory material in the digital context, an area with enormous added potential for all those involved in healthcare. In short, the benefits of technologies such as large language models lies in converting unstructured spoken language into a structured context, which makes the available data of use in the daily routine. This will bring with an enormous time saving while increasing the quality of the outcome. Another area of application is using AI to overcome language barriers.

If the integrative function that radiologists possess—such as when coordinating tumor boards—is considered, it becomes clear how important support services such as this are. And how important it is that the additional information is sorted and curated in one location—in the PACS—to relieve the burden on radiologists.

You might also be interested in

Radiology meets AI

AI will be a partner for employees in healthcare

That AI is increasingly establishing itself in radiology is generally and rightly seen as a success and an important step toward the future. Ultimately, healthcare will—and must—become digital to be able to deal with future challenges. But still the question must be allowed: are our expectations of AI realistic? Is it really going to be a “best buddy” in radiology or just a “false friend”?

Best Buddies in radiology

Best buddies in radiology?

Only a very few, if any, radiologists are worried that AI will make their job redundant. But does this mean that AI should be promoted to a radiologist’s “best buddy”? Who knows. There are supposedly people who find conversations with Alexa the same as talking to a friend.