Brain fingerprint and subjective mood state across the menstrual cycle

Abstract

Background: Brain connectome fingerprinting represents a recent and valid approach in assessing individual identifiability on the basis of the subject-specific brain functional connectome. Although this methodology has been tested and validated in several neurological diseases, its performance, reliability and reproducibility in healthy individuals has been poorly investigated. In particular, the impact of the changes in brain connectivity, induced by the different phases of the menstrual cycle (MC), on the reliability of this approach remains unexplored. Furthermore, although the modifications of the psychological condition of women during the MC are widely documented, the possible link with the changes of brain connectivity has been poorly investigated.

Methods: We conducted the Clinical Connectome Fingerprint (CCF) analysis on source-reconstructed magnetoencephalography signals in a cohort of 24 women across the MC.

Results: All the parameters of identifiability did not differ according to the MC phases. The peri-ovulatory and mid-luteal phases showed a less stable, more variable over time, brain connectome compared to the early follicular phase. This difference in brain connectome stability in the alpha band significantly predicted the self-esteem level (p-value <0.01), mood (p-value <0.01) and five (environmental mastery, personal growth, positive relations with others, purpose in life, and self-acceptance) of the six dimensions of well-being (p-value <0.01, save autonomy).

Conclusion: These results confirm the high reliability of the CCF as well as its independence from the MC phases. At the same time the study provides insights on changes of the brain connectome in the different phases of the MC and their possible role in affecting women's subjective mood state across the MC. Finally, these changes in the alpha band share a predictive power on self-esteem, mood and well-being.

Keywords: brain connectivity; brain fingerprint; depression; menstrual cycle; self-esteem; well-being.

Figure 1

FCs processing and fingerprint analysis. (A) a: the neuronal activity was recorded using a 154-sensors magnetoencephalography (MEG); b: MEG signals were cleaned by removing noise and artifacts, coregistered with the subject-specific MRI scan for source reconstruction (c); d: functional connectivity matrix estimation based on the PLM. (B) The green and the pink blocks represent the two identifiability matrices of women in two different MC phases (in this explanatory figure: the T1 and T2 phases), resulting from the correlation of the test and re-test of the individual functional connectomes, in each phase (defined according to the MC phase) separately. Hybrid identifiability matrices (IMs) were created by crossing the FCs test of the T1 phase with the FCs retest of the T2 phase and vice-versa. The hybrid IMs contain the “cycle-specific fingerprinting” score (I-cyclical, IC) of each individual.

Figure 2

Iterative model of edgewise subject identification. The success rate (SR) distributions of T1, T2, and T3 phases, obtained by adding 100 edges at each step from the most to the least contributing ones, according to the intraclass correlation (ICC) values. The T1 phase (blue) quickly reached a complete SR (100%), and a slowly progressive decline but always preserving a great identifiability power. The T2 and T3 phases (red and green, respectively), equally to the T1, quickly reach an almost complete SR but start, after a few hundred edges, an important drop and a progressive loss in subject identification reaching a success rate of ~70–75% with 4,005 edges. The shaded areas of each figure represent the null distributions obtained by adding 100 edges at a time in a random order.

Figure 3

A multilinear regression model with leave-one-out cross validation (LOOCV) was performed to test the capacity of the I-cyclical to predict self-esteem, depression and well-being in the T2 phase. Panels on the left show the explained variance by the stepwise addition of the six predictors. The significant predictor is indicated with * in bold. The middle panels show the comparison between actual and predicted clinical features. Panels on the right show residuals distribution with cross-validation.

Figure 4

Pearson correlation between cycle-specific fingerprinting and subjective symptoms. The figure shows an inverse correlation between both self-esteem and wellbeing and IC and positive correlation between IC and depression. Higher scores at the Ryff and Rosenberg tests are related to positive outcomes, whereas higher BDI scores are indicative of worse outcomes.

Figure 5

The colored areas represent the ROIs with the greatest nodal strength in identifiability and predicting subjective state. The nodal strength was calculated on the intraclass correlation matrices representing the stability between phases T1 and T2.