Artificial intelligence (AI) in mammographic screening – a systematic review and assessment of medical, economical and ethical aspects

Abstract Introduction The aim of this project was to evaluate the effects of implementing aritfical intelligence in mammographic screening for breast cancer. Method A comprehensive search of articles was conducted in March 2024 across MEDLINE, Embase, Cochrane Library and CINAHL databases. Selected studies were assessed by at least two independent authors, with inclusion decisions made by consensus. The studies underwent critical appraisal, and relevant data were extracted. Due to diversity among studies, meta-analyses were not possible and therefore a narrative synthesis of data was performed. Certainty of results was evaluated using the GRADE approach. A health economic evaluation based on information from available studies on short-term outcomes and costs was conducted, and an ethical analysis performed. Result Use of AI in combination with radiologist reading in breast cancer screening with mammography provides equivalent or higher cancer detection rates compared to double reading by radiologists (moderate certainty of evidence ). Use of AI in breast cancer screening with mammography reduces the workload of radiologists compared to double reading by radiologists (low certainty of evidence ). Use of AI in breast cancer screening with mammography will, according to the short time scenarios, result in an increased cost compared to the current routine of double reading by radiologists. Discussion It is important to ensure that use of AI leads to long term health gains or other values for care, which may motivate additional costs.