Insights From the September Webinar on Enhancing Healthcare Workflows with Artificial Intelligence

The recent SIIM’s Artificial Intelligence webinar held on September 4th, discussed how advanced AI platforms are revolutionizing healthcare workflows. Moderated by Dr. Mana Moassefi, a postdoctoral research fellow at Mayo Clinic, the session featured insights from Dr. Bradley Erickson, Professor of Radiology at Mayo Clinic, and Dr. Elliot Siegel, Professor of Radiology at the University of Maryland School of Medicine. This webinar focused on AI platforms, their components, and integration challenges within clinical workflows.

The webinar began by addressing the question, ‘What is an AI platform?’ AI platforms are multifaceted systems designed to handle various tasks in healthcare, from image analysis to patient scheduling and data integration. Dr. Erickson emphasized the importance of seamless integration into clinical workflows, automatically receiving images from devices like CT or MR scanners and applying relevant AI tools. He emphasized the importance of a DICOM receiver to manage image data, along with Health Level 7 (HL7) or Fast Healthcare Interoperability Resources (FHIR) capabilities to interpret non-image data such as patient demographics or medical history.

To explain the need for workflow orchestration, Dr. Siegel drew parallels between the early days of transitioning from film-based radiology to digital platforms and today’s growing AI ecosystem. Much like how Picture Archiving and Communication Systems (PACS) revolutionized medical imaging by standardizing and organizing data, AI platforms must now evolve to integrate various AI applications smoothly. Without a central platform to manage this integration, scaling the use of multiple AI tools becomes cumbersome. As Dr. Siegel described, AI tools need to work together, much like instruments in an orchestra, to provide seamless insights. This orchestration should ensure interoperability across different systems while maintaining vendor neutrality. The term “vendor-neutral orchestration and analytics portal” (V-NOAP) was proposed to describe these future platforms that could support radiology and extend across various healthcare departments.
Both experts highlighted the growing role of multimodal AI, where non-image data such as pathology results, medications, and demographic information play a key role in enhancing AI’s performance. Dr. Erickson emphasized the importance of continuous monitoring to ensure AI tools maintain their expected performance over time, with the platform providing insights into when and why performance might degrade.

Looking ahead, both speakers agreed that AI platforms should support a wide range of functionalities, from imaging support to billing and patient follow-up, making them indispensable across departments beyond radiology. As the role of AI expands, healthcare systems will need to adopt platforms that not only enhance workflow but also ensure seamless integration with electronic medical records (EMR) and other hospital systems. In conclusion, this part of the discussion underscored the need for healthcare to adapt to a new AI-driven era. With advanced platforms designed for flexibility, scalability, and interoperability, AI can truly revolutionize healthcare workflows, bringing efficiencies and improved patient outcomes.
Next, the speakers talked about the future of AI platforms in healthcare, particularly in radiology, and how these platforms could become essential for medical practice. They described how AI could assist in identifying subtle medical issues like lung nodules or intracranial hemorrhages and the importance of presenting this information to doctors in a user-friendly way. One key point is that a centralized platform to integrate various AI applications would help streamline radiology interpretations. The idea was to create a system where AI tools are not used in isolation but work together, simulating the collaborative environment often seen with multiple specialists. They stressed that without such platforms, the efficiency of radiology reporting could not be fully realized. While some AI tools exist, the slow pace of platform development was a concern.
Standardization of AI processes was another key focus. They argued against an “AI app store” model, where users manually choose AI tools, favoring instead an automated, standardized approach where scans are processed consistently. This would reduce variability in medical practices and improve efficiency. However, they acknowledged that exceptions and special cases must be handled differently, requiring flexible tools. In the meantime, the necessity of a balance between customization and standardization in AI use was also discussed. While departments or divisions might want to personalize their tools, too much individualization can create issues, especially if AI tools are designed to work with specific protocols.

Finally, the discussion emphasized the need for long-term data storage to track AI performance, quality, and feedback, with the FDA potentially requiring post-market surveillance of AI tools. This means platforms will need to manage feedback while carefully considering data access and transparency, especially for patients. Key advice for evaluating AI platforms included assessing if the platform can handle future needs, integrate with existing systems, support standardized protocols, and manage long-term data storage, feedback, and regulatory requirements. Additionally, the importance of adhering to standards for long-term reliability and preferred implementations was highlighted. Essential features for future platforms include multimodal data support, AI application monitoring, and compatibility with existing systems like PACS and speech recognition platforms. The conversation also underscored the need for flexibility in AI platform choices, as there is no clear dominant player yet, and many radiology practices may already have underutilized platforms capable of supporting AI tools. In other words, some AI vendors and large hospital systems could declare themselves as platforms, integrating imaging and other applications seamlessly. However, it remains unclear who will emerge as the leader.

In summary, the webinar provided a comprehensive overview of the transformative potential of AI platforms in healthcare. The insights shared by Dr. Erickson and Dr. Siegel highlighted the critical need for seamless integration, workflow orchestration, and continuous monitoring to ensure AI tools deliver optimal performance. As the healthcare industry moves towards an AI-driven future, adopting advanced platforms that support flexibility, scalability, and interoperability will be essential. By addressing the challenges and opportunities discussed, healthcare systems can harness the full potential of AI to enhance workflows, improve patient outcomes, and drive innovation across various medical disciplines.