Patient Views on AI in Radiology

Patient Views on AI in Radiology

Artificial intelligence is moving more visibly into radiology and other areas of healthcare, raising new questions about how patients view its role in care. A recent article in the British Journal of Radiology examined these attitudes in people attending hospital for outpatient imaging across a tertiary hospital network. Most participants saw AI as useful and supported its use in healthcare, but confidence fell when questions turned to trust, human interaction and autonomous decision-making. The responses show support for AI as a clinical support tool, alongside clear expectations that doctors should remain responsible for decisions and that care should retain a strong human dimension.

Support for Use Does Not Extend to Full Trust

The responses show broad support for using AI in healthcare, with 64% saying it should be used and 78% saying it would be useful. At system level, 68% thought AI would reduce waiting times. Interest was also high when participants considered predictive information, with 77% saying they would want to know about an AI prediction of future disease.

Views became more divided when the questions moved from usefulness to trust and safety. While 53% thought AI would be safe in healthcare, only 23% trusted a computer to make a medical decision, while 50% did not. Participants also raised concerns about data security. More respondents thought confidential health data would be less safe with AI than safer with it.

Must Read: AI vs Radiologists in MRI Lumbar Spine Reporting

The same pattern appeared in questions about performance. Many participants supported AI being used to check doctors’ judgement, suggesting a role for support rather than substitution. At the same time, responses did not indicate confidence that AI should act alone. This balance between acceptance and caution shaped the strongest overall finding in the questionnaire: people were open to AI in healthcare, but not to unrestricted decision-making by computers.

Human Contact and Doctor Responsibility Remain Central

The strongest views in the questionnaire concerned the human side of care. Participants placed clear value on being treated as a person rather than a number, and most said they would not be satisfied with an AI decision that did not take their feelings into account. These responses formed the factor labelled “interaction”, which captured concern about what could be lost if AI replaced direct contact with clinicians.

That concern sat alongside a very clear view on responsibility. Most participants said doctors should remain responsible for decisions involving AI. The question of autonomous use drew particularly strong opposition, separating it from other parts of the questionnaire. This was not simply a general dislike of technology. Participants were willing to support AI in several healthcare functions, but they did not support removing the doctor from responsibility for care decisions.

Responses on replacement were also decisive. Most participants did not think doctors could be replaced by AI, and many thought a computer could not compete with the experience of a specialist doctor. Even where opinion was mixed on whether AI might lead to more or fewer errors than doctors, the overall direction remained consistent: AI could assist, but it did not command the same confidence as a doctor. The findings place personal interaction, professional judgement and human accountability at the centre of acceptable AI use.

Three Attitude Patterns and Limited Demographic Differences

Factor analysis identified three latent factors linked to 16 of the 18 questionnaire items: utility and safety, interaction and comparability to doctors. These factors help explain how support for AI can coexist with strong reservations. Participants were positive about utility and safety, opposed interaction with AI in place of doctors and viewed AI less favourably than doctors when direct comparison was involved.

Scores for these factors showed marked differences. Support was strongest for utility and safety, while interaction with AI instead of doctors drew the lowest support. Comparability to doctors also scored below the neutral point, indicating that AI was generally not seen as matching doctors. Opposition to autonomous AI use remained strong and stood apart from the three main factors.

Differences between demographic groups were limited. Female respondents were more likely to oppose interaction with AI, more likely to favour doctors over AI and less supportive of utility and safety to the same extent as male respondents. Support for autonomous AI did not differ by gender. The analysis found no evidence that attitudes differed by age or ethnicity.

The participant group included people attending a radiology department in an urban hospital network, with most living in South London and others in surrounding parts of Greater London. The paper also notes that the questionnaire was designed to be straightforward and quick to complete, using 18 items and five-level responses. The survey gives a broad overview of attitudes rather than a detailed qualitative exploration, but it offers a structured view of how patients currently distinguish between helpful AI, impersonal AI and autonomous AI.

The findings show a clear pattern in current patient attitudes to AI in radiology and healthcare. Participants generally support AI when it is associated with usefulness, safety and system benefit, including shorter waiting times and additional information. That support weakens sharply when AI appears to reduce personal interaction, compete with specialist doctors or act without doctor responsibility. The responses do not reject AI as such. They draw a boundary around how it should be used. Within that boundary, patients appear most comfortable with AI as a tool that supports care while leaving responsibility, judgement and human contact firmly in clinical hands.

Source: British Journal of Radiology