Artificial Intelligence user interface preferences in radiology: A scoping review

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence

Background:

• Artificial intelligence (AI) is increasingly utilized in radiology to support clinical processes amidst workforce shortages.
• The development of AI in radiology necessitates an understanding of user interface preferences for effective integration.

Purpose of the Study:

• To conduct a scoping review investigating existing literature on the preference of artificial intelligence (AI) interfaces within a radiology context.
• To identify current research trends and gaps in understanding user preferences for AI tools in medical imaging.

Main Methods:

• A systematic scoping review was conducted using Arksey O'Malley's and Levac's framework.
• Searches were performed across four databases (MEDLINE Ovid, Scopus, Web of Science, Engineering Village) with inclusion criteria focused on radiological AI user interface preferences.
• Three researchers were involved in paper selection to ensure reliability.

Main Results:

• Six papers met the inclusion criteria, employing diverse methodologies including observational studies, simulated user testing, and diagnostic accuracy studies.
• AI user interfaces were evaluated, with some preference for heatmap image overlays and highly detailed interfaces.
• Limited literature exists on AI user interfaces, with a lack of research on current interface preferences, both pre- and post-deployment.

Conclusions:

• The current research landscape reveals a lack of standardized methods for assessing AI tool design and user preference in radiology.
• While some preferences emerged (e.g., simple, non-complicated designs), findings regarding ideal user interfaces are not clear.
• Further research is essential, focusing on end-user involvement, particularly radiographers, and establishing standardized outcome measures for AI interface evaluation.