Estimating multivariate similarity between neuroimaging datasets with sparse canonical correlation analysis: an...

Area of Science:

• Neuroimaging
• Multivariate Statistics
• Biostatistics

Background:

• Neuroimaging studies increasingly combine multiple data modalities or measurements.
• Intra-modal studies often focus on data differences, necessitating methods to measure similarities.
• Canonical Correlation Analysis (CCA) is limited with high-dimensional data.

Purpose of the Study:

• To adapt sparse CCA (SCCA) for high-dimensional neuroimaging data.
• To develop a framework for finding multivariate image-to-image correspondences in intra-modal studies.
• To investigate similarities in cerebral blood flow effects of antipsychotic drugs.

Main Methods:

• Modification of sparse CCA (SCCA) for high-dimensional neuroimaging.
• Independent estimation of optimal variable subsets.
• Analysis of information in multiple SCCA transformations.
• Application to Arterial Spin Labeling (ASL) data.

Main Results:

• The modified SCCA framework facilitates application to high-dimensional neuroimaging data.
• The method enables the discovery of meaningful multivariate image-to-image correspondences.
• Demonstrated effectiveness in investigating drug effects on cerebral blood flow.

Conclusions:

• The proposed SCCA modification is a valuable tool for intra-modal neuroimaging studies.
• This approach enhances the characterization of multivariate similarities in complex datasets.
• The framework offers new insights into the effects of pharmacological interventions on brain function.