A method for capturing dynamic spectral coupling in resting fMRI reveals domain-specific patterns in schizophrenia

Deniz Alaçam, Robyn L. Miller, Oktay Agcaoglu, Adrian Preda, Judith M. Ford, Vince D. CalhounFrontiers in Neuroscience 2023 · DOI

TL;DR

Researchers have developed a new framework called time-resolved spectral coupling (trSC) to look beyond simple time-series correlations in the brain. By analyzing how the frequency profiles of brain networks synchronize over time, they discovered that people with schizophrenia exhibit significantly less consistent spectral coupling in the visual network, offering a potential new biomarker for sensory processing impairments.

What Problem Does This Paper Address?

Functional MRI (fMRI) typically measures brain connectivity by looking at how the "wiggles" in blood-oxygen-level-dependent (BOLD) signals from different regions line up over time. While this "dynamic functional connectivity" has revealed much about the brain, it has a blind spot: the frequency domain.

Two brain regions might have signals that look poorly correlated in time—perhaps due to a slight phase shift or time delay—yet they might be oscillating at the exact same frequencies. Standard methods miss this. The authors address this gap by proposing that the spectral profile (the "fingerprint" of frequencies) of a brain circuit is a vital piece of the connectivity puzzle, especially in disorders like schizophrenia where sensory processing is known to be disrupted but the underlying mechanics remain fuzzy.

Methods

The authors introduced a framework for time-resolved spectral coupling (trSC). Here is the breakdown of their pipeline:

  1. Network Identification: They used Group Independent Component Analysis (gICA) to identify 47 distinct functional brain networks (e.g., visual, auditory, default mode).
  2. Sliding Windows: Instead of looking at the whole 5-minute scan at once, they used a sliding window (50 time points wide) to capture how activity changes.
  3. Frequency Transformation: For every window, they applied a Fast Fourier Transform (FFT) to convert the time-course into a power spectrum.
  4. Spectral Correlation: They calculated the correlation between the power spectra of pairs of brain networks. This tells you if two regions are "tuning" to the same frequencies at the same time.
  5. Statistical Analysis: They grouped these correlations into quartiles and used "cluster size" and "cell count" metrics to determine how stable and modular these spectral relationships were across different groups (Healthy vs. Schizophrenia, Male vs. Female).

Datasets

The study utilized data from the FBIRN (Functional Imaging Biomedical Informatics Research Network) project.

  • Participants: 314 total subjects.
  • 151 people with schizophrenia (SZ).
  • 163 healthy controls (HC).
  • Demographics: The sample included a mix of males and females (though skewed toward males), with a mean age of approximately 37 years.
  • Data Type: Resting-state fMRI (rs-fMRI) collected at seven different sites across the US using 3 Tesla scanners.

Key Insights and Ideas

  • The Phase-Shift "Aha": The authors used a "toy example" to show that two signals shifted by 90 degrees ($\pi/2$) have zero correlation in the time domain but a correlation of 1.0 in the frequency domain. This proves that trSC captures connectivity that standard methods simply cannot see.
  • Visual Network Fragility: The most striking finding was in the visual network. In healthy controls, visual circuits show highly consistent spectral coupling. In schizophrenia, this coupling is significantly reduced. This suggests that the "visual clock" in the schizophrenic brain is less synchronized or more erratic.
  • Gendered Connectivity: The study found that males generally show higher spectral coupling in the visual network than females, regardless of diagnosis. This highlights the importance of accounting for sex as a biological variable in neuroimaging.
  • Modular Differences: Schizophrenia isn't just "noisy" brain activity; the differences were highly modular. While visual coupling decreased, coupling within the Default Mode Network (DMN) actually showed increases in some contexts, suggesting a complex reorganization of how the brain prioritizes internal vs. external information.

Key Results

  • Visual Network (VIS): Healthy controls (HC) showed significantly higher connectivity rates in the upper quartile of spectral coupling compared to the SZ group.
  • Sensorimotor (SM): Similar to the visual network, the HC group showed higher coupling strength in sensorimotor regions.
  • Default Mode (DM): Interestingly, the SZ group showed higher spectral coupling between DM-DM and DM-Cognitive Control (CC) pairs compared to controls.
  • Quantitative Metrics: The researchers used $p$-values thresholded at $< 0.05$ (FDR corrected). The "cell count" analysis (measuring the number of highly correlated window pairs) proved to be a more sensitive metric for detecting group differences than "average cluster size."

Limitations and Future Work

  • Window Parameters: The study used a fixed window length of 50 TRs. The authors acknowledge that smaller windows might capture more rapid fluctuations but could also introduce more noise.
  • Eyes-Closed State: The data was collected with participants' eyes closed. This is a less controlled state where drowsiness or "mind-wandering" can influence results. Future studies might compare this to "eyes-open" or task-based spectral coupling.
  • Frequency Range: The current study looked at the full spectrum available in the BOLD signal. Future work could benefit from "filter banks" to see if specific frequency bands (e.g., very low vs. high frequency) drive these differences in schizophrenia.

Summary updated

Abstract

The results of this study reveal significant differences in the degree to which spectral power profiles are coupled over time. Importantly, there are significant but distinct differences both between males and females and between people with schizophrenia and controls. We observed a more significant coupling rate in the visual network for the healthy controls and males in the upper quartile. Fluctuations over time are complex, and focusing on only time-resolved coupling among time-courses is likely to miss important information. Also, people with schizophrenia are known to have impairments in visual processing but the underlying reasons for the impairment are still unknown. Therefore, the trSC approach can be a useful tool to explore the reasons for the impairments.
BibTeX
@inproceedings{alacam2023method,
  author = {Deniz Alaçam and Robyn L. Miller and Oktay Agcaoglu and Adrian Preda and Judith M. Ford and Vince D. Calhoun},
  title = {{A method for capturing dynamic spectral coupling in resting fMRI reveals domain-specific patterns in schizophrenia}},
  booktitle = {Frontiers in Neuroscience},
  year = {2023},
  doi = {10.3389/fnins.2023.1078995},
}
Figure