Clinical Integration of Artificial Intelligence for Breast Imaging

We have described an approach to integrate AI products for image interpretation in breast screening, ensuring that the choice of algorithm(s) and deployment is based on an understanding of the local breast screening pathway. This is the beginning of a new era of image interpretation, and there is much to learn as the technology moves from research and evaluation into routine clinical use. We believe that success depends on having a multidisciplinary team approach to planning and implementing technology, supported by careful monitoring of the impact of introducing a "nonhuman reader." Services should also be mindful of complying with local legislation and governance principles and must work proactively to avoid introducing or widening inequalities. It is likely that the early benefit of AI in breast screening will be an improved workflow (eg, reduced reporting times and lower false-positive rates) but ultimately the goal should be for AI to improve early diagnosis of fast-growing cancers that have the potential to be fatal. More accurate classification of lesions at diagnosis may inform treatment options and improve outcomes.48 48 Radiologists must contribute to the development of integrated systems, gaining expertise in data science to ensure they understand and support the quality of the data used in AI, with new roles for individuals with both clinical qualifications and understanding of IT.