Abstract
Developing fair and unbiased models is important for good scientific practice and clinical utility. This paper delves into the specific biases associated with artificial intelligence (AI) in neuroimaging research, and highlights the structural issues that underpin them. We propose a range of mitigation strategies, encompassing both behavioural and technical considerations. By recognising these challenges, we can encourage more accurate and equitable insights into neuroimaging research.
Type
Publication
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics
Sophie Martin
PhD Student
My research interests include neuroimaging, artificial intelligence and interpretability.
Francesca Biondo
Senior Research Fellow
James H Cole
Professor of Neuroimage Computing
My research interests include neuroimaging, ageing and neurodegenerative diseases