Dynamic fluctuations in ascending heart-to-brain communication under mental stress

Figure 1.. Exemplary stress elicitation images. A: sample image from stress load condition 1. The displayed video is an excerpt from the documentary “The Reality of Van Life” (Different Media 2018). The video was presented for 5 min. B: sample image from stress load condition 2. The experimental condition consisted in watching the documentary from (A) simultaneously to performing a digit span task of memorizing 3-digit sequences, for approximately 5 min. C: sample figure from stress load condition 3. The experimental condition consisted in watching the documentary from (A), performing the digit span task from (B), and performing the red box task of memorizing 4 × 4 patterns, for approximately 5 min.

Figure 2.. Self-reported stress level for three stressful conditions. Each data point corresponds to the reported stress level per subject for each for: 1) stress load condition 1: documentary, 2) stress load condition 2: documentary + digit span task, 3) stress load condition 3: documentary + digit span task + red box task. ****P < 0.00005 from Wilcoxon signed-rank test.

Figure 3.. Group-wise distributions of RR, SAI, and PAI median and variability for each experimental condition. Each data point corresponds to the measured autonomic marker per subject at each of the four conditions. A: RR, SAI, and PAI median. B: RR, SAI, and PAI variability as measured through median absolute deviation (MAD). Time-varying autonomic indexes were z-score normalized for the whole experimental protocol duration before computing median and MAD values. **P < 0.005, ***P < 0.0005, ****P < 0.00005 (Bonferroni-corrected significance at α < 0.00833). PAI, parasympathetic activity index; RR, R-to-R peak interval; SAI, sympathetic activity index.

Figure 4.. Topographic maps of Friedman nonparametric test for the paired brain-heart interplay variability on the ascending (top) and descending (middle) direction from the SV-SDG model among experimental conditions (rest and three mental stress conditions). In the bottom, topographic maps of Friedman nonparametric test for paired EEG power variability among experimental conditions (rest and three mental stress conditions) are shown. White electrodes indicate P < 0.0056. The variability measure was the median absolute deviation. Complementary results on SV-SDG and EEG power medians are in Supplemental Fig. S1. EEG, electroencephalogram; PAI, parasympathetic activity index; SAI, sympathetic activity index; SV-SDG, sympathovagal synthetic data generation.

Figure 5.. Multivariate analysis results. A: the five most informative brain-heart interplay features according to the MRMR regression criteria. The regression was performed on the group median stress level: rest = 0, stressor condition 1 = 1, stressor condition 2 = 4, stressor condition 3 = 5. B: the five most informative brain-heart interplay features according to the MRMR criteria to classify low vs. high-stress levels. Low stress referred to rest and stress load condition 1; high stress referred to stress load conditions 2 and 3. C: regression model output from a fivefold cross-validation to the stress level using the best five markers presented in A. The statistical comparisons correspond to paired Wilcoxon tests on the estimated stress level score computed with the regression model. D: binary stress level classification using the five most informative features in B. Yellow and orange circles indicate low and high-stress conditions, respectively. **P < 0.005, ***P < 0.0005, ****P < 0.00005 (Bonferroni-corrected significance at α < 0.00833). MRMR, minimum redundancy maximum relevance.

Figure 6.. Exemplary participant during the experimental protocol. Parallel fluctuations in (A) SAI → β and (B) SAI → δ are displayed. SAI, sympathetic activity index.