JMIR Publications

ABSTRACT

Background:

Artificial intelligence (AI) promises to significantly impact daily radiology practices. Numerous studies have already been conducted that anticipate this potentially disruptive innovation. So far, these studies have mainly focused on single topics, like ’trust’, or investigating perspectives of single stakeholder groups, like radiologists. This study aims to explore future directions for AI in radiology by incorporating perspectives of a heterogenous group of stakeholders on a broad spectrum of moral and economic topics.

Objective:

The study aims to co-create and reflect with a broad range of stakeholders on viable implementation scenarios for scalable AI applications in radiology in the Netherlands, thereby identifying potential opportunities and frictions, with a focus on moral and economic considerations.

Methods:

To inform the workshop design, a non-systematic narrative literature search was performed to deepen our understanding of key moral and economic considerations at play in the field of radiology and AI. Workshop participants, representing a wide range of actors including radiologists, innovators and patient representatives, were selected using purposive sampling. Data was collected in a co-creation workshop. In three subsequent rounds, mixed over three breakout groups, a total of 17 participants were asked to: 1) map what they considered important moral and economic considerations; 2) envision possible future scenarios for AI in radiology and 3) discuss opportunities, frictions and routes to success. Transcribed recordings were coded and cross-checked.

Results:

Workshop participants envision future AI-driven scenarios, ranging from extramural imaging departments for increased accessibility to healthcare, to multimodal data integration for human-centered AI-enhanced diagnostics. Seven themes emerge from the discussions during the workshop: 1) trust and efficiency of AI technologies, 2) responsibilities in clinical decision-making when AI is involved, 3) diagnosis as a one-off versus an iterative process, 4) regulations as a requirement or a restriction, 5) economic benefits or drawbacks, 6) trade-off between amount of information required and patient privacy and 7) environmental considerations.

Conclusions:

Reflecting on the seven emerging themes, we identify three overarching topics: 1) human-AI collaboration and trust, 2) governance, regulation and ethical safeguards and 3) value creation and sustainability. These topics highlight the need to balance technological advancements with ethical responsibility, institutional accountability and societal benefit. They also underscore the importance of designing AI systems that not only perform well but are also trusted and aligned with clinical workflows and patient values. These overarching themes offer a lens through which future research and policy can navigate the complex interplay between innovation, regulation and real-world implementation. Future research is needed to validate the generalizability of the results across various countries and healthcare settings.