AI in Radiology: Hype vs Reality, Imaging AI, Workflow, and Clinical Adoption | Dr. Ben Fine

Is radiology AI finally living up to the hype — or are we still waiting for the revolution Geoff Hinton promised back in 2012? In this episode of Imaging Informatics Unplugged, Jason sits down with Dr. Ben Fine, a radiologist with a deep background in imaging informatics, AI deployment, and enterprise imaging strategy.Ben shares a refreshingly honest take on where radiology AI actually stands today: only about 1% of imaging workflows are meaningfully augmented by AI tools, and real ROI is still rare — driven more by FOMO than outcomes. But he argues the corner is being turned, as the field shifts from narrow deep learning models to foundation models capable of assisting with full radiology reports.The conversation digs into real-world AI deployment lessons from the AIDE Lab at Trillium Health Partners, where Ben’s team developed a pre-deployment evaluation methodology for PACS-integrated AI tools that has since become best practice. They also explore the Swiss cheese model of human-AI collaboration, why operational workflow use cases (like protocol automation and procedure nomenclature mapping) often beat flashy diagnostic AI for ROI, and what effective AI governance looks like for health systems.Whether you’re a PACS Admin, Imaging Manager, Radiologist, or Healthcare IT professional thinking about enterprise imaging and AI in radiology, this episode is packed with practical, colleague-to-colleague insights.Learn more at nagelsconsulting.comKEY TOPICS COVEREDAmara’s Law and the realistic 3-year outlook for radiology AI — why we’re finally at the inflection point after years of overhypeThe AIDE Lab at Trillium Health: how pre-deployment evaluation of PACS-integrated AI tools became a best-practice framework across OntarioThe Swiss cheese model of human-AI collaboration — why AI and radiologists fail in such different ways, and how to design systems that catch what neither misses aloneOperational AI use cases that deliver real ROI: procedure nomenclature mapping, CT/MRI protocol automation, and bone mineral density (BMD) workflow assistance• Ontario’s centralized diagnostic imaging repository (OCINet/DIR) as an untapped opportunity for population health AI and opportunistic screening at scaleAI governance frameworks for health systems — applying a pharmaceutical-committee model to the selection, validation, deployment, and monitoring of AI toolsThe future skills that matter: why domain expertise combined with AI fluency — not just soft skills — will define the next generation of imaging informatics professionals