Systematic Review of Challenges in Implementing Artificial Intelligence in Breast Cancer Screening Programs: Towards a Framework for Safe Adoption (Preprint)

Abstract

BACKGROUND Artificial intelligence (AI) studies show promise in improving accuracy and efficiency in mammographic screening programs worldwide. However, its adoption in the clinical workflow faces challenges, such as unintended errors, professional training needs, and ethical concerns. Of note, specific frameworks for AI in breast cancer screening are lacking. OBJECTIVE This paper reports a systematic review aiming to assess existing literature and develop a tailored AI governance framework for adoption in breast cancer screening METHODS Three electronic databases (PubMed, EMBASE, and Medline) were searched using combinations of the key words, “artificial intelligence”, “regulation”, “governance”, “breast cancer” and “screening”. Original studies evaluating AI in breast cancer detection or discussing challenges related to AI implementation in this setting were eligible for review. Findings were narratively synthesized before being mapped directly against the constructs within the Consolidated Framework for Implementation Research (CFIR). RESULTS A total of 1240 results were retrieved, and 20 original studies were eventually included in this systematic review. Studies identified challenges in adopting AI in breast screening included reproducibility, evidentiary standards, technology concerns, trust issues, ethical, legal, societal concerns, and post-adoption uncertainty. Mapping these findings against the constructs within the CFIR, we recognize the complex interactions in the development and implementation of a governance framework, across the AI adoption life-cycle in the context of breast cancer screening. Action plans corresponding to the main challenges were included within the framework, aiding in a structured approach to address these issues. CONCLUSIONS This systematic review identified key themes as well as the barriers and facilitators for AI governance in breast cancer screening. Post-market surveillance is emphasized for continuous monitoring and auditing to ensure the effectiveness and ethical implementation of AI in breast cancer screening. CLINICALTRIAL na