A New Approach to Partial Conjunction Analysis in Neuroimaging

2025-08-05

AI Summary

This study addresses the key problem of identifying brain regions significantly activated by a specific cognitive task in at least γ subjects within multi-subject neuroimaging analysis. We propose CoFilter, a method that formulates this as a partial conjunction hypothesis testing problem. CoFilter constructs partial conjunction p-values and integrates simultaneous inference techniques to overcome the excessive conservativeness of conventional multiple testing corrections (e.g., Bonferroni) in high-dimensional neuroimaging data. Compared with existing approaches, CoFilter substantially improves detection sensitivity and result reproducibility. Extensive validation on simulated datasets and large-scale real fMRI data—including the Human Connectome Project (HCP)—demonstrates that CoFilter yields more accurate activation localization, stronger cross-subject stability, and superior computational efficiency. By unifying theoretical rigor with practical effectiveness, this work establishes a novel paradigm for population-level statistical inference in neuroimaging.

Technology Category

• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Computer Vision: Multi-modal Vision
• Machine Learning: Multi-instance/Multi-view Learning

Application Category

• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

The problem of identifying the brain regions activated through a particular cognitive task is pivotal in neuroimaging. This problem becomes even more complex if we have several cognitive tasks or several subjects. In this paper, we view this problem as a partial conjunction (PC) hypotheses testing problem, i.e., we are testing whether a specific brain region is activated in at least $γ$ (for some pre-fixed $γ$) subjects. We propose the application of a recent advance in the simultaneous statistical inference literature to activation localization in neuroimaging. We apply the recently proposed CoFilter method to neuroimaging data to discover brain regions activated in at least $γ$ subjects. Our proposal has two distinct advantages. First, it alleviates the conservativeness displayed by the traditional multiple testing procedures in testing PC hypotheses by eliminating many of the conservative PC $p$-values. Second, it is especially suitable for several high-dimensional studies, each of which examines a large number of null hypotheses. We also compare the performance of our proposal with existing methods for testing PC hypotheses through extensive simulation studies on neuroimaging data and a real dataset.

Problem

• Identifying brain regions activated in cognitive tasks
• Testing partial conjunction hypotheses for multiple subjects
• Improving statistical methods for high-dimensional neuroimaging data

Innovation

• Applies CoFilter method to neuroimaging data
• Reduces conservativeness in multiple testing
• Suitable for high-dimensional hypothesis studies