Uncovering and correcting biases in neuroimaging studies

Recent work [1] on neuroimaging has demonstrated significant benefits of using population graphs to capture non-imaging information in the prediction of neurodegenerative and neurodevelopmental disorders. These non-imaging attributes may contain demographic information about the individuals, e.g. age or sex, but also the acquisition site, as imaging protocols and hardware might significantly differ across sites in large-scale studies. The effect of the latter is particularly prevalent in functional connectomics studies, where it’s unclear how to sufficiently homogenise fMRI signals across the different sites. A recent study [2] has highlighted the need to investigate potential biases in the classifiers devised using large-scale datasets, which might be imbalanced in terms of one or more sensitive attributes (like gender and race). This can be exacerbated when employing these attributes in a population graph and lead to disparate predictive performance across sub-populations. This project aims to uncover any potential biases of a semi-supervised classifier that relies on a population graph and explores methods to mitigate such biases to produce fairer predictions across the population.

[1] *Parisot, S., *Ktena, S. I., Ferrante, E., Lee, M., Moreno, R. G., Glocker, B., & Rueckert, D. Disease Prediction using Graph Convolutional Networks: Application to Autism Spectrum Disorder and Alzheimer’s Disease. Medical Image Analysis, 2018

[2] Larrazabal, Agostina J., et al. "Gender imbalance in medical imaging datasets produces biased classifiers for computer-aided diagnosis." Proceedings of the National Academy of Sciences, 2020.

Project timezone: C