Cluster size statistic and cluster mass statistic: two novel methods for identifying changes in functional...

Area of Science:

• Neuroscience
• Brain Imaging
• Statistical Analysis

Background:

• Functional connectivity analysis is crucial for understanding brain networks.
• The human functional connectome involves millions of connections, posing statistical challenges.
• Standard error control methods are often insensitive for whole-connectome comparisons.

Purpose of the Study:

• Introduce two novel cluster-based methods: cluster size statistic (CSS) and cluster mass statistic (CMS).
• To control the family-wise error rate in whole-connectome functional connectivity analyses.
• Evaluate the performance and sensitivity of CSS and CMS using simulated and real fMRI data.

Main Methods:

• Developed CSS and CMS methods for family-wise error rate control.
• Utilized receiver operator characteristic (ROC) analysis on simulated data.
• Tested method sensitivity on BOLD fMRI data from subjects under normal and hypercapnic conditions.

Main Results:

• Both CSS and CMS methods successfully controlled the family-wise error rate.
• Significant changes in functional connectivity were detected between normal and hypercapnic states using CSS and CMS.
• Individual connection-level analysis with family-wise error correction showed no significant changes.

Conclusions:

• CSS and CMS are effective, data-driven, and minimally assumption-based methods for analyzing functional connectome data.
• These cluster-based statistics offer improved sensitivity for detecting widespread connectivity changes compared to individual connection analyses.
• The methods provide a robust framework for comparing functional connectivity across different groups or conditions in fMRI research.