Artificial intelligence (AI) holds the promise to help transform radiology, from workflow optimization to potentially improved patient outcomes. Implementations are complex and can affect multiple departments within an organization. Successful AI platform implementations require careful coordination and expectation-setting across multiple clinical and technical stakeholders.
During the recent Society of Imaging Informatics in Medicine (SIIM) 2024 Annual Meeting, Graeme Pirrie, Global Product Management, Bayer Radiology Digital Solutions, presented a TECH Talk on best practices for building alignment across multiple stakeholder groups when planning and implementing a Radiology AI Orchestration Platform. Pirrie recommended that “the first step is to agree on a short-term, focused mission across the stakeholders to leverage the value propositions of Radiology AI systems and bring those opportunities into the existing infrastructure.” Successful AI platform implementations require creating a shared vision across physicians, administrators, and technologists, and also require careful planning and clear communication between these stakeholders.
Developing a Common Mission for AI Radiology
The AI Radiology market is growing dynamically with more than 720 FDA-approved AI applications for radiology and approximately one hundred new algorithms coming out every year. Pirrie noted that developing a common mission for Radiology AI provides the “intelligence to wade through the hundreds of algorithms and identify the ones that are going to be most useful.”
He recommended that stakeholders reach consensus on the following criteria before implementing AI:
Intended Use of AI Solutions
- Triage, screening, detection, diagnosis measurements, image quality, etc.
- Specific qualifications for AI report interpretation
Key Success Metrics or KPIs
- Anticipated clinical/financial outcomes
- Align on base case
Alignment on Patient Cohort or Use Case
Proof of Concept Pilot
- Tasks, Assignments, Durations, Dependencies
Once clinical and technical stakeholders agree on a common mission, Pirrie recommended AI Orchestration to enable the mapping of intended use of AI devices to a specific patient cohort and required imaging. “It is really important that we're able to be able to consume those AI algorithms and bring those to practical use in our radiology workflow,” Pirrie said.
He added, “There has to be intelligence in weeding through hundreds of algorithms to be able to select the ones that are going to be most useful aligning on the patient cohort that can be looked at through the lens of the intended use as well.”
Finding an Orchestration Partner
For IT departments, managing and integrating medical imaging solutions efficiently is crucial. A platform partner can help integrate AI solutions smoothly into existing Health Information Technology (HIT) systems, such as PACS and RIS, reducing implementation complexities. This partner can also provide a defined path to help select the right AI apps, including setting tangible milestones that will let you know when you are moving to the next phase of the IT implementation journey.
When selecting a platform partner, Radiology Departments should look for a provider with imaging and informatics domain experience as well as deep knowledge in dealing with the output of modalities and how to get reports out of disparate data in a normalized way. In addition, Pirrie recommended that a partner should have project management excellence, and have radiologists on staff so that you can have peer-to-peer conversations and optimize the use of these great AI tools that are out there in the environment.
For more information, Pirrie recommended reviewing: Unlocking the Value: Quantifying the Return on Investment of Hospital Artificial Intelligence - Journal of the American College of Radiology, Prateek Bharadwaj, MSca, Lauren Nicola, MDb, Manon Breau-Brunel, MScc, Federica Sensini, MScc, Neda Tanova-Yotova, MScd, Petar Atanasov, MScd, Franziska Lobig, MSce, Michael Blankenburg, PhD. JACR
Open Access Published: March 16, 2024.
2024-12-11
