Novel Group ICA Incorporating Time-Frequency Information for Longitudinal Brain Network Analysis

This R21 application aims to develop a novel group independent component analysis (ICA) algorithm, incorporating time-frequency information (GtfICA) to capture the temporal dynamics of functional networks (dFN) in longitudinal studies. The GrfICA will provide a comprehensive view of the spatiotemporal organization of brain activity by addressing the limitations of traditional functional connectivity methods. Our sample includes n=1000 cognitively normal participants aged 18-85 from the NKI-RS study, who were followed up to 3 years. All participants had both single- and multi-band resting-state fMRIs at each visit. We will identify age-related changes and assess the test-retest reliability of longitudinal changes in dFN between multi-band and single- band data. This is crucial for understanding the consistency and reproducibility of results across different imaging protocols that often occur in long-term longitudinal studies. In Aim 1, we will develop an independent component analysis utilizing time-frequency features and its extension to multi-subject analysis (GtfICA). We will evaluate their performance using extensive numerical experiments. Using NKI-RS baseline data, GtfICA will identify age-related changes in the dFN. We will assess the replicability of the identified age-associated dFN patterns across different fMRI acquisition sequences. In Aim 2, we will extend GtfICA to harmonize single- and multi-band longitudinal data collected longitudinally. We will incorporate the sampling rate changes in the GtfICA and quantify and test the longitudinal changes above and beyond the sequence changes. Also, we will evaluate the test-retest reliability of the longitudinal changes in dFN between multi- band and single-band data. Finally, we will evaluate the stability of the longitudinal changes in the dFN when the acquisition sequences change over time when only partial information on sequence changes is available. The outcome of this proposal is a crucial step toward accurately quantifying and testing the effect of aging and neurodegenerative disease on brain function across different image acquisition protocols and longitudinal changes. This study will enhance the generalizability of the inferences on dFN. The outcome of this proposal will be the basis for a future new R01 project. This subsequent research will evaluate longitudinal changes in the dFN in ADRD relative to aging in multiple clinical populations. Project Number: 1R21AG093534-01A1 | Fiscal Year: 2026 | NIH Institute/Center: National Institute on Aging (NIA) | Principal Investigator: SEONJOO LEE | Institution: COLUMBIA UNIVERSITY HEALTH SCIENCES, NEW YORK, NY | Award Amount: $390,122 | Activity Code: R21 | Study Section: Analytics and Statistics for Population Research Panel A Study Section[ASPA] View on NIH RePORTER: https://reporter.nih.gov/project-details/11309395