Spatial vs. Temporal Features in ICA of Resting-State fMRI – A Quantitative and Qualitative Investigation in the Context of Response Inhibition

Lixia Tian, Yazhuo Kong, Juejing Ren, Gaël Varoquaux, Yu‐Feng Zang, Stephen M. SmithPLoS 2013 · DOI

Abstract

Independent component analysis (ICA) is a standard tool for identifying functional networks in the resting brain. While spatial maps are frequently analyzed, the temporal information inherent in ICA—specifically the amplitude of network fluctuations—is often underutilized. This study investigates which ICA features correlate with behavior, using Stop Signal Reaction Time (SSRT) as a benchmark for response inhibition. By analyzing the amplitude of resting-state networks (RSNs), temporal correlations between networks, and spatial maps, we demonstrate that temporal amplitude is a robust predictor of cognitive performance, offering a complementary perspective to traditional spatial analysis.

Introduction

In neuroimaging, Independent Component Analysis (ICA) has become a cornerstone for defining Resting-State Networks (RSNs). These networks represent brain regions that exhibit synchronized activity even in the absence of an explicit task. Traditionally, researchers have focused on the spatial maps generated by ICA to understand how the topography of these networks varies across individuals.

However, ICA also yields temporal information: the specific time-series of activity for each identified network. Despite its potential, this temporal dimension is frequently overlooked. This research addresses a critical gap: which specific features of ICA—spatial maps, the amplitude of network fluctuations, or the synchronization between different networks—best correlate with human behavior? To test this, the study utilizes "Response Inhibition," measured via the Stop Signal Reaction Time (SSRT), as a behavioral metric.

Methodology

The study involved 34 healthy male participants who underwent 8-minute resting-state fMRI scans followed by a behavioral Stop-Signal Task. The analytical pipeline centered on Group ICA and dual regression to derive subject-specific features.

Spatial maps of resting-state networks (RSNs) identified via ICA at low (A) and high (B) dimensionality.

The researchers extracted three primary measures:

  1. Amplitude: Calculated as the standard deviation of each network's time-series, representing the vigor of network fluctuations.
  2. Temporal Connectivity: The correlation between different network time-series using various methods, including inverse covariance.
  3. Spatial Maps: Voxel-wise intensity within the identified networks.

The study also performed a sensitivity analysis by varying the ICA decomposition dimension (27 vs. 70 components) and testing different preprocessing steps, such as spatial smoothing and Z-transformation of the maps.

Results and Discussion

Amplitude as a Behavioral Marker

A key finding of this study is that the amplitude of network fluctuations at rest is a significant predictor of behavioral performance. Specifically, higher amplitudes in motor control, visual, and dorsal attention networks were significantly correlated with shorter SSRTs. This suggests that individuals whose attention and action networks fluctuate more strongly at rest are more efficient at inhibiting prepotent responses during a task.

The Role of Temporal Connectivity

While the temporal correlations between RSN pairs (functional connectivity) were found to be biologically meaningful, they were less sensitive than network amplitude in predicting SSRT. This highlights that the "vigor" of individual networks may be more indicative of certain cognitive traits than the synchronization between them.

Inverse covariance (ICOV) matrix representing temporal correlations between RSN pairs.

Spatial vs. Temporal Information

The voxel-wise analysis of spatial maps largely mirrored the amplitude results, with significant correlations found in the medial/lateral visual and motor networks. Interestingly, the study found that Z-transforming spatial maps—a common practice—actually reduced the sensitivity for detecting behavioral correlations. This suggests that the raw BOLD amplitude captured in the original maps contains vital biological signals that normalization might obscure.

Conclusion

This investigation demonstrates that the temporal information captured by ICA is not merely secondary to spatial maps but provides a powerful, complementary window into individual differences in cognitive performance. By "listening to the rhythm" of resting-state networks through their temporal amplitude, researchers can gain deeper insights into the neural underpinnings of behavior. Future work should expand these findings to more diverse populations and a broader range of cognitive domains.

Summary updated

Abstract

Independent component analysis (ICA) can identify covarying functional networks in the resting brain. Despite its relatively widespread use, the potential of the temporal information (unlike spatial information) obtained by ICA from resting state fMRI (RS-fMRI) data is not always fully utilized. In this study, we systematically investigated which features in ICA of resting-state fMRI relate to behaviour, with stop signal reaction time (SSRT) in a stop-signal task taken as a test case. We did this by correlating SSRT with the following three kinds of measure obtained from RS-fMRI data: (1) the amplitude of each resting state network (RSN) (evaluated by the standard deviation of the RSN timeseries), (2) the temporal correlation between every pair of RSN timeseries, and (3) the spatial map of each RSN. For multiple networks, we found significant correlations not only between SSRT and spatial maps, but also between SSRT and network activity amplitude. Most of these correlations are of functional interpretability. The temporal correlations between RSN pairs were of functional significance, but these correlations did not appear to be very sensitive to finding SSRT correlations. In addition, we also investigated the effects of the decomposition dimension, spatial smoothing and Z-transformation of the spatial maps, as well as the techniques for evaluating the temporal correlation between RSN timeseries. Overall, the temporal information acquired by ICA enabled us to investigate brain function from a complementary perspective to the information provided by spatial maps.
BibTeX
@inproceedings{tian2013spatial,
  author = {Lixia Tian and Yazhuo Kong and Juejing Ren and Gaël Varoquaux and Yu‐Feng Zang and Stephen M. Smith},
  title = {{Spatial vs. Temporal Features in ICA of Resting-State fMRI – A Quantitative and Qualitative Investigation in the Context of Response Inhibition}},
  booktitle = {PLoS},
  year = {2013},
  doi = {10.1371/journal.pone.0066572},
}
Figure