What you need to address for successful artificial intelligence adoption and enterprise imaging integration.
With so many artificial intelligence (AI) products flooding the imaging industry, the radiology space has taken on a gold rush feel. Vendors are in a frenzied battle to position themselves as the leader with the best tools available – but, there’s still a lack of over-arching standardization on how to implement and integrate these products.
In a session during the Society for Imaging Informatics in Medicine (SIIM) 2021 Virtual Annual Meeting, Florent Saint-Clair, executive vice president at Dicom Systems, outlined four major challenges facilities must address before they can successfully adopt AI algorithms and integrate enterprise imaging solutions.
“We can’t put our heads in the sand as the use of AI is going to bet more and more intensive,” he said. “This ecosystem is going to make sure the gold rush becomes a very well-structured industry.”
For more SIIM 2021 coverage, click here.
To reach that goal, he suggested you consider these obstacles.
Challenge No. 1: Insufficient Use of Industry Standards
In many instances, AI developers have developed their algorithms using basic JPG images. Products developed like this cannot take full advantage of the metadata that images contain. When you’re talking with AI developers, make sure their algorithm has been trained with and makes full use of DICOM data, he said, and clarify if they use DICOM, HL7, or DICOM Web, etc. In addition, make sure your workflow isn’t going to create an expectation that prior images will be pulled in with the primary image.
Challenge No. 2: To Cloud or Not to Cloud
Technology can be made available to you in a variety of ways – does this vendor offer it strictly in the cloud, such as Google, Amazon, or Azure, or can it also be deployed on premises? If data cannot fit in Docker in your own server, you will need to address the need for encrypted communication that can move the images safely, as well as data de-identification.
Challenge No. 3: Limited Workflow Knowledge
DICOM, HL7, and PACS workflow have been around for quite a while, but using them involves jargon that is unfamiliar to many AI developers. Before selecting a tool, check to see the algorithm has been developed with workflow in mind – not just the core technology that is part of the machine learning itself, he said.
You should also be aware that not everyone is 100-percent comfortable with using DICOM Web yet. Find out where your vendor stands and whether you will need to plan for additional hurdles when trying to communicate with them.
Challenge No. 4: Inconsistent Results Delivery
This is, perhaps, the biggest, most important challenge surrounding AI adoption and enterprise imaging integration, he said. You must be sure your data can be delivered to the right place in the right format. For example, results that should come in as a DICOM overlay can come through as a blob on top of the image if the vendor is not conforming to the right industry standards. You can also run into HL7 problems if the proper HL7 connection isn’t in place. If the system is set up correctly, he said, results can be shared seamlessly.
“Every piece of the workflow has different things that the AI vendor needs to think about before they develop their AI,” Saint-Clair said. “They need to know what to do at what point in the workflow to add value without having to re-invent workflow.”
Ultimately, he said, addressing these four challenges will help you identify your needs and what you should pursue from a vendor.
For more coverage based on industry expert insights and research, subscribe to the Diagnostic Imaging e-Newsletter here.
Considering Breast- and Lesion-Level Assessments with Mammography AI: What New Research Reveals
June 27th 2025While there was a decline of AUC for mammography AI software from breast-level assessments to lesion-level evaluation, the authors of a new study, involving 1,200 women, found that AI offered over a seven percent higher AUC for lesion-level interpretation in comparison to unassisted expert readers.
SNMMI: Can 18F-Fluciclovine PET/CT Bolster Detection of PCa Recurrence in the Prostate Bed?
June 24th 2025In an ongoing prospective study of patients with biochemical recurrence of PCa and an initial negative PSMA PET/CT, preliminary findings revealed positive 18F-fluciclovine PET/CT scans in over 54 percent of the cohort, according to a recent poster presentation at the SNMMI conference.
Could an Emerging PET Tracer be a Game Changer for Detecting Hepatocellular Carcinoma?
June 23rd 2025In addition to over 90 percent sensitivity in detecting hepatocellular carcinoma (HCC), the glypican-3 (GPC3) targeted PET tracer 68Ga-aGPC3-scFv appeared to be advantageous in identifying HCC tumors smaller than one centimeter, according to pilot study findings presented at the SNMMI conference.