Functional MRI Signal Complexity Analysis Using Sample Entropy

Maysam Nezafati, Hisham Temmar, Shella KeilholzFrontiers in Neuroscience 2020 · DOI

Abstract

Resting-state functional magnetic resonance imaging (rs-fMRI) is a powerful method in neuroscience that utilizes the blood oxygenation level-dependent (BOLD) signal to analyze neural activity. This study examines the complexity of regional brain activity by calculating the Sample Entropy (SampEn) of whole-brain BOLD volumes across distinct networks, cortical regions, and subcortical regions. Our findings reveal that different brain regions exhibit distinct levels of complexity and that entropy significantly decreases during task performance compared to the resting state, suggesting a transition toward more ordered neural dynamics during goal-directed behavior.

Introduction

Traditional functional MRI (fMRI) research primarily emphasizes functional connectivity—the correlation of activity between disparate brain regions. While this provides a map of the brain's communication network, it often overlooks the intrinsic temporal "texture" or complexity of the signals within those regions. The brain is a complex system that must balance spontaneous activity with the need to maintain low-entropy states for vital functions. This study addresses this gap by using Sample Entropy (SampEn) to quantify the regularity and complexity of BOLD signals, providing deeper insights into how brain structures are organized and how they adapt to cognitive demands.

Methods

The researchers utilized high-quality data from the Human Connectome Project (HCP), selecting 100 unrelated healthy individuals (ages 22–36). The dataset included two resting-state fMRI scans and two working memory (N-back) task fMRI scans per subject.

The analytical core of the study is the Sample Entropy (SampEn) algorithm, which measures the unpredictability of a time series. SampEn is particularly suited for fMRI data as it remains robust even with relatively short datasets (200–1000 time points). The workflow involved standard preprocessing (motion correction, registration to MNI space), parcellation into seven functional networks and 63 cortical/subcortical regions, and the generation of whole-brain complexity maps. Statistical comparisons between resting and task states were performed using paired t-tests and two-factor ANOVA.

Results

The analysis revealed a significant "task-order" effect: global brain entropy consistently decreases when transitioning from a resting state to performing a cognitive task. Specifically, mean cortical entropy dropped from 0.1912 at rest to 0.1815 during the task ($p < 0.05$). Subcortical regions followed a similar trend, decreasing from 0.1981 to 0.1897.

Figure 1: Whole-brain maps illustrating the spatial distribution of Sample Entropy during resting state (A) and task performance (B). Higher entropy values (red/orange) are prominent in frontoparietal regions at rest, with a visible global decrease during the task.

Regional variations were also pronounced. Subcortical structures, such as the thalamus and basal ganglia, exhibited significantly higher entropy than cortical regions in both states ($p < 0.05$). This suggests that subcortical regions may act as high-traffic integration hubs, processing diverse and rapidly changing inputs.

Figure 2: Box plots comparing average entropy in cortical and subcortical regions at rest and during task performance, highlighting the significant decrease in complexity during the task.

Furthermore, all seven functional networks showed a statistically significant decline in entropy during task performance ($p$-values ranging from $10^{-23}$ to $10^{-33}$). At rest, the frontoparietal network—associated with executive function—showed the highest entropy, whereas the dorsal attention network became the most complex during task performance.

Discussion and Conclusion

The observed reduction in entropy during task performance suggests that goal-directed behavior coordinates neural activity, making it more predictable and "ordered" compared to the spontaneous fluctuations of the resting state. The higher complexity of subcortical regions underscores their role in multi-modal integration.

While fMRI's BOLD signal is a hemodynamic proxy and acts as a low-pass filter, potentially losing high-frequency neural dynamics, these findings demonstrate that SampEn is a reliable and reproducible measure of brain state. Future research should explore how these entropy signatures vary in clinical populations, such as those with ADHD or Alzheimer’s disease, and whether individual differences in complexity can predict cognitive performance.

Summary updated

Abstract

Resting-state functional magnetic resonance imaging (rs-fMRI) is an immensely powerful method in neuroscience that uses the blood oxygenation level-dependent (BOLD) signal to record and analyze neural activity in the brain. We examined the complexity of brain activity acquired by rs-fMRI to determine whether it exhibits variation across brain regions. In this study the complexity of regional brain activity was analyzed by calculating the sample entropy of 200 whole-brain BOLD volumes as well as of distinct brain networks, cortical regions, and subcortical regions of these brain volumes. It can be seen that different brain regions and networks exhibit distinctly different levels of entropy/complexity, and that entropy in the brain significantly differs between brains at rest and during task performance.
BibTeX
@inproceedings{nezafati2020functional,
  author = {Maysam Nezafati and Hisham Temmar and Shella Keilholz},
  title = {{Functional MRI Signal Complexity Analysis Using Sample Entropy}},
  booktitle = {Frontiers in Neuroscience},
  year = {2020},
  doi = {10.3389/fnins.2020.00700},
}
Figure