Back

The impact of study design choices on significance and generalizability of canonical correlation analysis in neuroimaging studies

Pigeau, G.; Costantino, M.; Devenyi, G.; Bussy, A.; Parent, O.; Chakravarty, M.

2022-12-16 bioinformatics
10.1101/2022.12.14.520275 bioRxiv
Show abstract

This technical note describes the effects of different data reduction methods and sample sizes for neuroimaging studies in the context of canonical correlation analysis (CCA). CCA is a multivariate statistical technique which has gained increasing popularity in neuroimaging research in recent years. Here, we investigate the parcellation methods impact on elucidating neuroanatomical relationships (based on cortical thickness) with known risk factors related to Alzheimers disease risk using data from the UK Biobank. The cortical thickness values were parcellated using four common methods in neuroimaging (atlas-based parcellation, spectral clustering, principal component analysis, and independent component analysis) and results from CCA were compared. The results show that the choice of parcellation technique impacts the strength and significance of the correlation between the brain and behaviours. Principal component analysis and independent component analysis result in the strongest correlations. Additionally, we show that regardless of parcellation technique, smaller sample sizes of participants result in inflated correlation strength and significance.

Matching journals

The top 7 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.