Members of the iCAIRD radiology team collaboration recently had a paper published in Insights Into Imaging outlining their work on AI in breast screening mammography.

Clarisse de Vries, Samantha Colosimo, Moragh Boyle, Gerald Lip, Lesley Anderson & Roger Staff reported on their study surveying the views of breast screening readers in the UK on how to incorporate Artificial Intelligence (AI) technology into breast screening mammography.

 

Study overview

This study involved design of an online questionnaire which was distributed to UK breast screening reader community. Those surveyed were asked to answer questions indicating their degree of approval of four different AI implementation scenarios for mammography reading, and rank 5 AI representation options in terms of which evidence they considered necessary to support implementation of AI into their medical practice.

Results

87 nationally accredited respondents were invited to participate in the study, with 73 completing the survey in full. Respondents approved of AI replacing one of the initial 2 human readers in mammogram interpretation / diagnosis, and objected to AI replacing all human readers. Regarding implementation of AI as a triaging tool / companion reader, participants were divided in opinion. Most mammography screen readers’ regarded national guidelines, studies using nationally representative dataset and independent prospective studies as essential evidence. They additional highlighted concerns and the need for additional validation.

Overall, survey participants supported the introduction of AI to mammography reading as a partial replacement of human readers, with the development of further supporting evidence and national guidelines considered a crucial requirement prior to implementation.

Full article available here