Associations of Brain Entropy Estimated by Resting State fMRI With Physiological Indices, Body Mass Index, and Cognition

Abstract

Background: In recent years, resting-state fMRI (rsfMRI)-based brain entropy (BEN) has gained increasing interest as a tool to characterize brain activity. While previous studies indicate that BEN is correlated with cognition, it remains unclear whether BEN is influenced by other factors that typically affect brain activity measured by fMRI.

Purpose: To investigate the relationship between BEN and physiological indices, including respiratory rate (RR), heart rate (HR), systolic blood pressure (s-BP), and body mass index (BMI), and to investigate whether and to what extent the relationship between BEN and cognition is influenced by physiological variables.

Study type: Retrospective.

Subjects: One thousand two hundred six healthy subjects (mean age: 28.83 ± 3.69 years; 550 male) with rsfMRI datasets selected from the Human Connectome Project (HCP).

Field strength/sequence: Multiband echo planar imaging (EPI) sequence at 3.0 Tesla.

Assessment: Neurocognitive, physical health (RR, HR, s-BP, BMI), and rsfMRI data were retrieved from the HCP datasets. Neurocognition was measured through the total cognition composite (TCC) score provided by HCP. BEN maps were calculated from rsfMRI data.

Statistical tests: Multiple regression models, p_height-family wise error (FWE) < 0.05 and p_cluster-FWE < 0.05 were considered statistically significant.

Results: BEN was negatively associated with RR (T-thresholds ranging from 4.75 to 4.8; r-threshold = |0.15|) and positively associated with s-BP and BMI (T-thresholds ranging from 4.75 to 4.8; r-threshold = |0.15|) in areas overlapping with the default mode network. After controlling the physiological effects, BEN still showed regional associations with TCC, including negative associations (T-thresholds = 3.09; r-threshold = |0.1|) in the fronto-parietal cortex and positive associations (T-thresholds = 3.09; r-threshold = |0.1|) in the sensorimotor system (motor network and the limbic system).

Data conclusions: RR negatively affects rsfMRI-derived BEN, while s-BP and BMI positively affect BEN. The positive associations between BEN and cognition in the motor network and the limbic system might indicate a facilitation of information processing in the sensorimotor system.

Evidence level: 3 TECHNICAL EFFICACY: Stage 3.

Keywords: blood pressure; body mass index; brain entropy; cognition; respiratory rate; resting state fMRI.

Figure 1.

Scheme of the brain entropy (BEN) calculation adapted from (9). (A) A time window is used to extract a sub-time series at m successive timepoints (m = 8 here) from the original time series. (B) The embedding vector matching process is performed for the embedding window length of m (B.1) and m + 1 (B.2). (C) Sample entropy (SampEn) is calculated as the natural logarithm of the ratio between the total number of matches of the window length m and that of the window length of m+1. BEN maps were calculated using minimally and fully preprocessed resting state fMRI data (respectively m-rsfMRI and f-rsfMRI) (D). Participants with mean BEN > |3| standard deviations from the group average were identified as outliers and excluded from the sample (E). BEN maps were smoothed with a with full-width-at-half-maximum (FWHM) filter of 4 mm (F), and participants with missing behavioral data were excluded from analyses (G). BEN maps were entered as the dependent variable in multiple regression models to investigate the relationship between BEN and respiratory rate (RR), heart rate (HR), systolic blood pressure (s-BP), body mass index (BMI), and total cognitive composite (TCC) scores.

Figure 2.

Brain entropy (BEN) vs respiratory rate (RR) using minimally preprocessed rsfMRI data (m-rsfMRI) (a-d) and fully preprocessed rsfMRI data (f-rsfMRI) (e-h). Age, sex, and heart rate were included as covariates in all models. Results are reported for Rest1 LR (a, e), Rest1 RL (b, f), Rest2 LR (c, g), and Rest2 RL (d, h). The T-threshold is reported for each analysis. Colormap intensity is based on T-values. The colormap limits have been set as the highest T-value found across all analyses. Negative values indicate a negative association between BEN and RR.

Figure 3.

Brain entropy (BEN) vs systolic blood pressure (s-BP) using minimally preprocessed rsfMRI data (m-rsfMRI) (a-d) and fully preprocessed rsfMRI data (f-rsfMRI) (e-h). Results are reported for Rest1 LR (a, e), Rest1 RL (b, f), Rest2 LR (c, g), and Rest2 RL (d, h). The T-threshold is reported for each analysis. Colormap intensity is based on T-values. The colormap limits have been set as the highest T-value found across all analyses. Positive values indicate a positive association between BEN and s-BP and between BEN and BMI.

Figure 4.

Brain entropy (BEN) vs body mass index (BMI) using minimally preprocessed rsfMRI data (m-rsfMRI) (a-d) and fully preprocessed rsfMRI data (f-rsfMRI) (e-h). Results are reported for Rest1 LR (a, e), Rest1 RL (b, f), Rest2 LR (c, g), and Rest2 RL (d, h). The T-threshold is reported for each analysis. Colormap intensity is based on T-values. The colormap limits have been set as the highest T-value found across all analyses. Positive values indicate a positive association between BEN and s-BP and between BEN and BMI.

Figure 5.

Brain entropy (BEN) vs total cognitive component (TCC). This association was first tested including only TCC in the model (a-d), and then adding respiratory rate (RR), hearth rate (HR), and systolic blood pressure (s-BP) as covariates (e-h). Age and sex were included as covariates in all models. Results are reported for Rest1 LR (a, e), Rest1 RL (b, f), Rest2 LR (c, g), and Rest2 RL (d, h). The T-threshold is reported for each analysis. Colormap intensity is based on T-values. The colormap limits have been set as the highest T-value found across all analyses. Positive values indicate a positive association between BEN and TCC, whereas negative values indicate a negative association between BEN and TCC.