Perceptions of artificial intelligence for radiographic image accuracy: a cross-sectional survey of radiology personnel.

Authors

Abstract

To evaluate the perceptions of radiology students, interns, and technologists regarding the role of artificial intelligence (AI) in improving the accuracy of radiographic images, and to compare these perceptions across the three groups. A quantitative, cross-sectional study was conducted at King Saud bin Abdulaziz University for Health Sciences (KSAU-HS) campuses in Riyadh, Jeddah, and Al-Ahsa, and at National Guard hospitals in Jeddah, between January and December 2025. Radiology students, interns, and technologists were recruited by non-probability convenience sampling (<i>n</i> = 81). Data were collected through a structured, self-administered online questionnaire using a five-point Likert scale. Descriptive and inferential statistics (Chi-square, Mann-Whitney U, and Kruskal-Wallis tests) were performed using Microsoft Excel and JMP, with significance set at 0.05. Of 81 respondents, 91.4% self-rated their AI knowledge as good to excellent, although 60.5% reported never having used AI tools in radiological practice. Most respondents (79.0%) had not received AI-related training, and 86.4% agreed that combining AI with human expertise increases diagnostic accuracy. AI was widely regarded as an effective diagnostic aid (87.7%), although trust in AI-processed radiological images was considerably lower (30.9% agreement, 42.0% neutral). The overall composite AI-attitude score (median 3.85, IQR 3.54-4.15; mean 3.82 ± 0.53) showed no statistically significant differences in this sample across the subgroups examined, including gender (<i>p</i> = 0.063), professional role (<i>p</i> = 0.193), prior AI use (<i>p</i> = 0.086), or formal AI training (<i>p</i> = 0.178); effect sizes were small. Most respondents (82.7%) disagreed that AI will replace radiologic technologists. Participants held positive perceptions of AI as a supportive tool and believed that its integration with human expertise could improve radiographic image accuracy and reduce diagnostic errors, without viewing AI as a replacement for radiologic technologists. These findings reflect participants' perceptions rather than objectively measured diagnostic performance. Understanding how radiology students, interns, and technologists perceive artificial intelligence supports its integration into clinical imaging workflows, informs targeted training programs, and helps align professional expectations with the realistic role of AI in improving radiographic image accuracy.

Tags

Topics