Abstract
Intrinsic brain activity is characterized by large-scale spatiotemporal patterns that underpin functional connectivity and cognition. Quasi-periodic patterns (QPPs) and complex principal component analysis (cPCA) have emerged as reproducible methods for capturing spatiotemporal network interactions in resting-state functional magnetic resonance imaging (rs-fMRI). However, these methods remain sensitive to methodological factors such as scan duration, repetition time (TR), and frequency band selection. This study systematically evaluates how these parameters influence the stability and reliability of QPP- and cPCA-derived functional connectivity patterns across multiple datasets. Using five independent rs-fMRI datasets, we evaluate the impact of scan length on pattern reliability, explore the effects of TR on spatiotemporal patterns, and compare the sensitivity of different frequency bands (Slow-5, Slow-4, infraslow) in capturing network dynamics. Our findings reveal that while both QPPs and cPCA detect intrinsic network activity, their reliability varies with acquisition parameters. QPPs exhibit greater stability in shorter scans, making them suitable for individual-level analyses, whereas cPCA provides a broader representation of phase-coherent fluctuations but shows greater between-subject variability and benefits more from longer, group-level acquisitions. Additionally, frequency band selection significantly influences the temporal structure of extracted patterns: in our analyses, Slow-5 (0.01-0.027 Hz) tended to emphasize more recurrent, synchronized network configurations, whereas Slow-4 (0.027-0.073 Hz) more often revealed transitions between connectivity states. These results provide critical insights into optimizing methodological choices for dynamic functional connectivity analysis, enhancing the interpretability of spatiotemporal patterns in both basic and clinical neuroimaging research.