Knowledge, Attitudes, and Implementation Barriers Regarding Artificial Intelligence in Radiology Practice: A Cross-Sectional Survey of Radiologists and Radiographers

Abstract

Background: Artificial intelligence (AI) tools are increasingly available for clinical radiology, yet routine implementation remains limited by governance, technical, and workforce-related constraints. Objective: To assess AI knowledge, attitudes, and perceived implementation barriers among radiologists and radiographers in private healthcare centers in Lahore, and to identify predictors of intention to adopt AI in routine practice. Methods: A multi-center cross-sectional survey was conducted from January to June 2025 using a structured questionnaire assessing participant characteristics, objective AI knowledge (0–10), attitudes, and barrier severity (Likert 1–5). Group differences were tested using chi-square and t-tests, and multivariable logistic regression estimated adjusted odds ratios (AORs) for high adoption intention. Results: Of 312 participants (182 radiologists, 130 radiographers), current AI tool use was reported by 44.2% and was higher among radiologists (53.8% vs 30.8%, p<0.001). Mean knowledge score was 6.1±2.1, with higher scores among radiologists (6.8±1.9 vs 5.2±2.0; mean difference 1.60, 95% CI 1.16–2.04; p<0.001). The most frequent major barriers were ethical/legal liability (62.8%; mean 4.1/5), lack of AI knowledge (58.0%; mean 3.9/5), and trust/black-box concerns (55.8%; mean 3.8/5). In adjusted analysis, prior AI training (AOR 2.45), good knowledge (AOR 2.12), and higher trust (AOR 3.08) increased adoption intention, while high legal concern reduced intention (AOR 0.62). Conclusion: AI adoption intention was common but constrained by liability concerns, capability gaps, and trust, supporting implementation strategies centered on training, local validation, and clear governance frameworks.