The Financial, Operational, and Clinical Advantages of Generalist Radiology AI.

Abstract

Despite the rapid growth of Food and Drug Administration-cleared artificial intelligence (AI)- and machine learning-enabled medical devices for use in radiology, current tools remain limited in scope, often focusing on narrow tasks and lacking the ability to comprehensively assist radiologists. These narrow AI solutions face limitations in financial sustainability, operational efficiency, and clinical utility, hindering widespread adoption and constraining their long-term value in radiology practice. Recent advances in generative and multimodal AI have expanded the scope of image interpretation, prompting discussions on the development of generalist medical AI. In this context, this review proposes the concept of generalist radiology AI (GRAI) and introduces key features for its implementation. GRAI aims to <i>(a)</i> create reports based on positive diagnoses, (b) tailor reports to indications for normal studies, <i>(c)</i> compare findings with prior imaging, <i>(d)</i> incorporate patient characteristics, and <i>(e)</i> provide uncertainty-informed, interactive recommendations. By consolidating image interpretation and expanding the incorporation of patient context, GRAI has the potential to overcome the limitations of narrow AI solutions, improve financial sustainability, streamline operational efficiency, and enhance clinical utility. Appropriate development of GRAI, building on these proposed features, is crucial for realizing the full potential of AI in radiology and enhancing diagnostic performance while reducing the clinical burden on radiologists.