Background: Medical imaging remains at the forefront of advancements in adopting digital health technologies in clinical practice. Regulator-approved artificial intelligence (AI) clinical decision support systems are commercially available and being embedded into routine practices for radiologists internationally. These decision support solutions show promising clinical validity compared to standard practice conditions; however, their implementation over time and implications on radiologists’ practice are poorly understood. Objective: This paper aims to examine the real-world implementation of an AI clinical decision support tool in radiology through a qualitative evaluation across pre-, peri-, and postimplementation phases. Specifically, it seeks to identify the key contextual, organizational, and human factors shaping adoption and sustainability, to map these influences using the nonadoption, abandonment, scale-up, spread, and sustainability (NASSS) framework, and to generate insights that inform evidence-based strategies and policy for integrating AI safely and effectively into public hospital imaging services. Methods: This prospective study was conducted in a large public tertiary referral hospital in Brisbane, Queensland, Australia. One-to-one participant interviews were undertaken across the 3 implementation phases. Participants comprised radiology consultants, registrars, and radiographers involved in chest computed tomography studies during the study period. Interviews were guided by the NASSS framework to identify contextual factors influencing implementation. Results: A total of 43 semistructured interviews were conducted across baseline (n=16), peri-implementation (n=9), and postimplementation (n=18) phases, comprising 7 (16%) radiographers, 20 (47%) registrar radiologists, and 16 (37%) consultant radiologists. Across NASSS domains, 56 barriers and 18 enablers were identified at baseline, 55 and 14 during peri-implementation, and 82 and 33 postimplementation. Organizational barriers dominated early phases, while technological issues such as system accuracy, interoperability, and information overload became most prominent during and after rollout. Enablers increased over time, particularly within the technology and value proposition domains, as some clinicians adapted the AI as a secondary safety check. Trust and adoption remained constrained by performance inconsistency, weak communication, and medicolegal uncertainty. Conclusions: The implementation of AI decision support in radiology is as much an organizational and cultural process as a technological one. Clinicians remain willing to engage, but sustainable adoption depends on consolidating early positive experiences and addressing negative ones, embedding communication and training, and maintaining iterative feedback between users, vendors, and system leaders. Applying the NASSS framework revealed how domains interact dynamically across time, offering both theoretical insight into sociotechnical complexity and practical guidance for hospitals seeking to move from pilot to routine, trustworthy AI integration.
