Decoding AI-adoption in radiotherapy: A multi-case qualitative comparative analysis

Purpose/Objective:
Over the past decade, Artificial Intelligence (AI) has gained significant attention in radiotherapy (RT) research. Nevertheless, in clinical practice, the number of implemented AI applications remains limited. The most prevailing AI tools are organ at risk (OAR) auto-contouring, followed by auto-planning. This study aims to identify and decode the patterns of conditions that facilitate the adoption of AI in radiotherapy using Qualitative Comparative Analysis (QCA).
Material/Methods:
A multi-case analysis was conducted across eight radiotherapy centers in the Netherlands. Eighteen interviews were
held with clinicians, medical physicists, and radiotherapy technologists. The study employed QCA to quantify qualitative data and identify patterns across cases. Key variables examined included data set characteristics, IT capabilities, AI configurability, explainable AI, stakeholder involvement, and standardization of contouring guidelines.
Results:
Auto-contouring was frequently mentioned as a successful implementation (5 out of 8 centers), and a couple of centers also adopted auto-planning (2 out of 8 centers). Twelve interviewees questioned the effectiveness of AI, acknowledging that while it saves radiotherapists time, it doesn't fully automate the workflow, and human oversight is still necessary as the model can be inaccurate. The QCA revealed multiple pathways leading to successful AI adoption. Crucially, the presence of AI configurability and stakeholder involvement emerged as consistent factors in positive outcomes. Centers that lacked high-quality
data set characteristics faced significant barriers, as insufficient or non-standardized data led to inaccurate AI performance. Conversely, when centers had data sets with standardized contouring guidelines, auto-contouring performed more effectively. Strong IT capabilities were also instrumental to support the integration with imaging systems, EHR and other critical
software. The QCA analysis indicated that even in the absence of certain factors like explainable AI features, the combination of configurable AI tools and proactive stakeholder engagement could drive successful adoption.
This suggests that customization to local needs and involving end-users in the implementation process can mitigate some of the challenges posed by limited data quality or legislative constraints.
Conclusion:
The study identifies AI configurability, stakeholder involvement, and high-quality data sets as essential for AI adoption in radiotherapy. Data set characteristics, particularly their quality and standardization, directly impact AI performance. By focusing on these key facilitating factors, radiotherapy centers can overcome challenges such as legislative barriers and varying IT capabilities. This study underscores the multi-faceted nature of AI adoption in healthcare and provides insights into practical deployment strategies, emphasizing the importance of data standardization and stakeholder engagement for policymakers and practitioners.