[Paper Review] Locally-Optimized Inter-Subject Alignment of Functional Cortical Regions

Marius Cătălin Iordan, Armand Joulin+2 more|arXiv (Cornell University)|Jun 7, 2016

Functional Brain Connectivity Studies11 references3 citations

TL;DR

This paper proposes a locally optimized inter-subject alignment method that predicts functional cortical region locations by maximizing functional correlation between time courses while allowing non-smooth local deformations. It significantly outperforms AFNI and FreeSurfer baselines, achieving 24-25% overlap with ground-truth LOC (vs. 10-11%) and 26% consistency across subjects (vs. 9-11%), enabling reliable ROI transfer without localizer scans.

ABSTRACT

Inter-subject registration of cortical areas is necessary in functional imaging (fMRI) studies for making inferences about equivalent brain function across a population. However, many high-level visual brain areas are defined as peaks of functional contrasts whose cortical position is highly variable. As such, most alignment methods fail to accurately map functional regions of interest (ROIs) across participants. To address this problem, we propose a locally optimized registration method that directly predicts the location of a seed ROI on a separate target cortical sheet by maximizing the functional correlation between their time courses, while simultaneously allowing for non-smooth local deformations in region topology. Our method outperforms the two most commonly used alternatives (anatomical landmark-based AFNI alignment and cortical convexity-based FreeSurfer alignment) in overlap between predicted region and functionally-defined LOC. Furthermore, the maps obtained using our method are more consistent across subjects than both baseline measures. Critically, our method represents an important step forward towards predicting brain regions without explicit localizer scans and deciphering the poorly understood relationship between the location of functional regions, their anatomical extent, and the consistency of computations those regions perform across people.

Motivation & Objective

• To address the challenge of inaccurate inter-subject mapping of high-level visual functional regions, particularly the lateral occipital complex (LOC), due to high anatomical variability and lack of consistent landmarks.
• To improve the reliability of transferring functional region locations across subjects without requiring separate localizer scans for each region.
• To develop a method that enables precise, region-specific alignment by maximizing functional correlation while allowing non-smooth local deformations.
• To investigate the relationship between functional contrast peaks, anatomical location, and cortical computation across individuals.
• To provide a general-purpose solution for predicting functional ROI locations using only functional fMRI data and minimal anatomical priors.

Proposed method

• The method tiles the seed ROI into smaller sub-regions and independently optimizes the functional correlation between each sub-region and candidate locations on the target cortical surface.
• It uses a constrained optimization framework that limits the maximum increase in distance between adjacent sub-regions to preserve topological consistency.
• The method maximizes the functional correlation between time courses of corresponding sub-regions across subjects to identify the optimal mapping.
• It allows for non-smooth local deformations, bypassing the assumption of continuous, smooth mappings used in prior methods.
• The algorithm is applied to predict the location of the LOC using fMRI data from a single high-variability stimulus experiment.
• The method is evaluated using intersection over union (IoU) for both accuracy and consistency across subjects.

Experimental results

Research questions

• RQ1 Can a locally optimized registration method improve the accuracy of inter-subject alignment for functionally defined cortical regions like LOC?
• RQ2 How does allowing non-smooth local deformations affect the reliability of ROI prediction compared to smooth, global transformations?
• RQ3 To what extent can functional correlation alone, without explicit anatomical landmarks, enable precise ROI transfer across subjects?
• RQ4 How does the consistency of predicted LOC locations compare across subjects when using this method versus standard alignment baselines?
• RQ5 Can this method reduce or eliminate the need for time-consuming, dedicated localizer scans in fMRI studies?

Key findings

• The proposed method achieved 24-25% intersection over union (IoU) between predicted LOC and ground-truth LOC, significantly outperforming AFNI (10-11%) and FreeSurfer (10-11%) baselines.
• The method demonstrated 26% IoU for regions consistently predicted across three or more subjects, compared to 9-11% for the two baseline methods.
• Visual inspection of predicted maps showed reduced variance and higher localization accuracy, with predicted peaks fully contained within the ground-truth LOC boundaries.
• The method produced more consistent mappings across subjects than both AFNI and FreeSurfer, particularly in high-level visual areas distant from the occipital pole.
• The results indicate that functional correlation-based optimization with local deformation flexibility enables more reliable ROI transfer than global, smooth transformations.
• The method enables accurate prediction of functional regions without requiring separate localizer scans, suggesting a path toward single-experiment functional mapping.