Brain Matters: Exploring Bias in AI for Neuroimaging Research

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.

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