Ethical AI: A qualitative study exploring ethical challenges and solutions on the use of AI in medical imaging

Abstract

Artificial Intelligence (AI) is being rapidly deployed in clinical practice in medical imaging settings worldwide. AI applications have the potential to transform this discipline and provide better patient outcomes. However, many ethical challenges exist when implementing AI in clinical practice. This study aims to explore these challenges and suggest ways forward. This study was supported by the European Federation of Radiographer Societies (EFRS), together with the European Society of Radiology (ESR) through the EFRS Research Hub at ECR 2024. Ethics approval was in place before data collection. All professionals within the medical imaging AI ecosystem who were registered congress attendees were eligible to participate. This qualitative study employed semi-structured interviews. All interviews were audio recorded after informed written consent by study participants. Transcribed data was analysed using a content analysis approach. In total, 43 professionals took part in this study. The sample included radiographers, radiologists, medical physicists, health informaticians, and business and IT specialists. Respondents recognised many ethical challenges in the clinical use of AI, such as data protection issues, lack of governance frameworks, potential inequalities in healthcare delivery, lack of diverse data, accountability issues in case of erroneous use, and lack of explainability. They also expressed additional concerns on staff deskilling due to overreliance on technology, AI education gaps and sustainability. Participants proposed that teamwork, continuous monitoring of AI tools, close collaboration with industry, rigorous legislation, and updated academic curricula could help address these ethical challenges. This study highlights the need to consider different ethical issues before AI implementation and to carefully introduce customised solutions to minimise risks. • Data protection, lack of governance, over-reliance on AI, and accountability were noted as ethical challenges of clinical AI use. • AI training, post-market surveillance, governance, co-production, and multidisciplinarity could address these challenges. • Risks and challenges must be efficiently addressed to ensure responsible AI implementation in clinical practice.

Ähnliche Arbeiten