Efficacy of different dynamic functional connectivity methods to capture cognitively relevant information

Area of Science:

• Neuroscience
• Brain Imaging Analysis
• Functional Connectivity

Background:

• The human brain's dynamic nature necessitates studying time-varying functional connectivity (dFC).
• Numerous dFC estimation techniques exist, but their comparative efficacy in capturing relevant dynamics remains under-investigated, especially for resting-state scans.
• A lack of well-defined ground truth hinders the evaluation of dFC methods.

Purpose of the Study:

• To evaluate the efficacy of various dynamic functional connectivity (dFC) estimation techniques.
• To determine which dFC methods best capture cognitively relevant modulations induced by external tasks using a multitask dataset as ground truth.
• To compare framewise and window-based dFC approaches.

Main Methods:

• Evaluated two framewise methods: dynamic conditional correlation (DCC) and jackknife correlation (JC).
• Assessed five window-based methods: sliding window correlation (SWC), SWC with L1-regularization (SWC_L1), moving average DCC (DCC_MA), multiplication of temporal derivatives (MTD), and delete-d jackknife correlation (dJC).
• Efficacy was measured by the ability of dFC metrics to segment multitask scans into cognitively homogenous periods.

Main Results:

• All window-based dFC methods performed well for window lengths ≥ 30 seconds, with SWC, SWC_L1, and DCC_MA showing slightly superior performance.
• For shorter window lengths (≤ 15 seconds), DCC_MA and dJC yielded the best results.
• Framewise methods (DCC, JC) did not achieve high accuracy in dFC estimation.

Conclusions:

• Window-based methods are generally effective for capturing task-related dFC dynamics.
• The choice of dFC method and window length impacts the ability to detect cognitive modulations.
• DCC_MA and dJC show promise for shorter window analyses, while SWC and SWC_L1 are robust for longer windows.